Is this the inevitable outcome of frontier labs who own their hardware? the GPUs and datacenters are the major cost. The inference and training a higher tier value proposition, if the company gets nervous that the investment in hardware won't pay off - renting it becomes a major topic of conversation.
A frontier model team having to fight their board on whether to monetize the datacenters directly or continue to invest in AI work is going to have a hard time.
> And Google is a major shareholder in SpaceX, so they certainly have incentive to juice the valuation of the IPO.
Google own 5-6% of the shares of SpaceX. SpaceX is seeking a valuation of $1.77T which means Google's shares would be worth $88.5B-$106.2B. I'm not a skeptic of AI/LLMs but this makes me deeply suspicious of these circular deals. What happens when the music stops?
I want to make a comparison with a car rental business and say that it would be like valuing Hertz entirely on the basis of the number of cars they own, as opposed to how many they rent out, but cars have a much longer depreciation period, if there are no customers they’re not costing you more money, unlike your computer which you are using for training and sucking up massive amounts of energy, and those cars do maintain decent value even after they’re of little use to the car rental company, unlike the compute here.
> Time on 4 year old H100 servers costs more now than when they were new (!!)
There are several confounding factors.
We’ve seen massive inflation since then. So some growth in cost was expected.
More importantly, the current Tech industry almost always starts by selling things at a loss. The increased cost could simply be the industry choosing to not subsidize that particular service anymore.
But also, I don’t think that’s a realistic comparison. Rented out GPUs are likely not a similar use profile as compute used for training LLMs. The latter is likely closer to the cryptocurrency GPUs that are running at full tilt 24/7.
> Rented out GPUs are likely not a similar use profile as compute used for training LLMs. The latter is likely closer to the cryptocurrency GPUs that are running at full tilt 24/7.
This is untrue.
H100's are used for training (well were, but are now outdated because B100/B200s are much faster).
Most of the reason people rent H100s is for smaller training runs.
If you are doing inference you usually buy managed capacity at Baseten or something, and that is often priced differently (although it comes down to an extra margin on longer term H100 prices basically).
Inference utilization is often actually higher than training now because so much effort has been spent on optimizing that stack.
Everything is a temporary situation on long enough timeframes, especially if it’s exponentially growing. Moore’s law which dictates that compute depreciates quickly has been slowing down a lot in the last few years, coupled with the explosion in demand we’ve found ourselves in a prolonged shortage situation. The bubble will pop, but if you predict when correctly, you will be a rich man.
The key question is on direction of LLMs. Right now, LLMs are taking over human jobs. If the cost of silicon+power < cost of human being doing the same work, what rational reason is there to employ a human being?
If this applies to SWEs, lawyers, business analysts, many research scientists, .... this situation could persist for a long, long time. While capital costs less than the inputs of labor (nominal food, housing, etc.), there is no need for labor.
The key question is about continued progress in models, and of the tooling around them:
- Plateau: Old silicon obsoletes in due course
- Rise quickly: Old silicon maintains value for a long time
The overwhelming majority of the labor force remains service, manual labor, and other such stuff that LLMs will have no real effect on. So the economy will be fine, but I do agree with you from a different angle. The entire goal of LLMs seems self destructive. If they're successful then the endgame is completely removing the barriers to entry to producing software and other digital tech. But if we do reach that endgame then the value of tech is going to plummet because there will be absolutely no barriers to entry to compete, or even just individuals homebrewing up what they need on demand.
Like imagine there was something you could buy where you insert some lumber, give it some passable description of furniture, and it outputs it. And you paid $20/month for access to this. And this was all being bankrolled by the furniture industry? I mean, sure guys - it's much appreciated, but I don't think I've ever seen anybody so enthusiastic about digging their own grave. I think it's already obvious that the gazillion dollars of API calls isn't going to materialize - it seems the handful of companies that trialed that are already reversing course hard. And in the future where LLMs are successful, that'd be even more true.
Llms either reach the point where they can quickly design and build physical robots to take on that service industry or they stop exponential growth.
Both of those are devastating for their valuation. Stopping growth means open modes catch up in a year or so. Continuing means end of the current economy.
Yes, the plan seems to be anti human in the extreme. Why do you need the plebs if they can be entirely replaced by AI? But the question then becomes why does the AI (and before that their security detail in a post money world) need billionaires?
This likely is the tertiary reason as to why llms are so heavily kneecapped. Granted, at this point, projects do exist to remove those arbitrary restrictions, but the effort that goes into it suggests it is a real concern.
Short term, money physically exists and gets spent, so if you wave a magic want of oversimplification and transition all labour to AI instantly, all the money currently in bank accounts and wallets gets spend on the same businesses it was already getting spent on, a lot of which gets spent on stuff from other businesses who have in this scenario also replaced all their labour with AI.
Eventually, perhaps quickly, all this money ends up in the hands of shareholders and landlords. There's a lot of both in the economy; famously retirement funds, but smaller-scale shareholders and landlords also exist. I wouldn't want to guess what the distribution looks like, probably highly variable between countries not just social classes (the definitions of which themselves can vary between countries).
Long term, money exists as a convenient fiction to help us organise transactions of goods and services: while it may be physically possible to eat gold and banknotes, you're not getting any real nutrients out of it when you do. So in a world where goods and services come from machines, the options are too broad to forecast: humanity could be relegated to the same role and economic stature as other primates (both in and out of zoos), or we could get universal UBI denominated in machine labour credits which lets each of us live better lives than the most extravagant billionaires live today.
I don't know. It just seems odd because money was used as an abstraction of labor and if labor disappears it seems like money has no fundamental value. If you can't pay people to do something (because machines are doing all the labor). Then people have no money and money has no value to people. Industrialization resulted in transition to service-based economy but this new wave of machines are being said to replace service work.
I'm just trying to understand if suppose you have fully robotic farms and fully automated slaughterhouses and fully automated McDonald's, who is McDonald's selling anything to and how do these people supposedly buying fully-mechanized burgers have jobs? Something just doesn't add up about this in my head about how this equation balances.
UBI ultimately seems like socialism with extra steps. Mostly is comes across as billionaires desperately begging for an alternative to being nationalized.
It's not rational relative to the short-term incentives of a typical corporation or investment vehicle. PE, VC, fund managers aren't paid to give a fuck about the social contract. Literally not in their job description.
> Is wanting low unemployment in our society not rational?
Only conditionally on there being bad consequences for high unemployment.
I don't particularly trust politicians, but there's a whole host of hypothetical scenarios about futures where work is essentially optional. Unfortunately, they're all either in the sci-fi or religion sections of the book store:
Despite people occasionally investigating UBI, the efforts to research UBI seriously have the same problems that Marx had with literal Communism, in that there's an obvious difference between any partial transition as compared to a global transition, and we don't have a completely disconnected parallel world to be a petri dish for us to test the economic outcomes on.
> Capitalism allows individuals to take decisions in a free market.
Capitalism provides a set of incentives that shape how people make decisions. Anyone can be selfish, but selfishness in capitalist society has a particular shape. To ignore the external incentives when looking at human behavior is horribly naive and shortsighted, but is frequently done by capitalism-apologists who seek to disregard any criticism of their favorite incentive system.
Are current datacenter deployments structured in such a way that the memory can later be moved to newer GPU dies? Or is it all packaged together as on consumer graphics cards?
I assumed the latter and therefore that the memory is depreciating along with the GPU cores it's soldered onto PCBs with.
... or is it a different argument being made, perhaps that depreciation for GPUs has slowed because rising demand will keep them in service longer?
Removing RAM chips off old cards is uneconomical, until it isn't. With things going the way they are, if you've got a card with soldered on RAM that could be transplanted to a newer card, I think you'll start seeing that happening.
Depreciating doesn't just mean it could depreciate in value relative to the performance of newer GPUs, but also that its lifespan is limited by reliability issues and failures.
I also feel that the GPU/NPU value does not lose money as fast anymore.
What I am wondering though is how long can you run such a system at basically full load without interruption before it starts to just physically degrade.
If I have a H100 and I let it run for 4 years at full throttle does it still have the same theoretical value as it had at the start or are the chips just burning out.
I think I remember that back when the cards used for crypto mining were sold en masse on ebay the advice was to stay away from them because they are more likely to fail?
Temperature is a big factor, as well as current density.
But there's also the # and magnitude of thermal cycles (which translate into mechanical stress, leading to metal-fatigue like effects on contact points etc), attack from chemicals in the air, cosmic radiation, ESD damage & more. Some may matter, some not.
That's why "new" > "used" in case of electronics. Especially since you don't know the (ab)use history of used parts.
Others say that moderate load means a lifespan of ~5 years
Not sure what that means but I would assume that a datacenter will start replacing a node once the error rate hits a certain threshold without really investigating why it failed, so the practical lifespan may be shorter than 5 years even if it would technically still be usable enough
> I also feel that the GPU/NPU value does not lose money as fast anymore.
That's because the rate of improvement in silicon manufacturing has been continually declining for a few decades, which has a compounding effect. Just compare the technological improvements in successive decades. 1976->1986->1996->2006->2016->2026.
That's why "in real terms" performance has only been very slowly improving if you compare apples to apples (and not e.g. apples to oranges by reducing precision, like nvidia tends to do, or by comparing chips with x W to an MCM with x*2 W and saying the latter is much faster). The "just halve the number of bits in each generation" strategy has also run out now, there's no more bits to halve.
My local inference rig now costs three times what I bought it for. If I'd gotten the max ram I could at the time I would have made $10k after selling the excess to my current spec.
How someone can look at an asset class thats appreciated an order of magnitude in the last two years and say it will depreciate in value when the tailwinds are even stronger now is beyond me.
Undervolting is not running at max utilization by definition almost.
…but the real question whether you want to undervolt your asset if you’re renting it out is why bother? You probably expect to replace it anyway after it’s spec lifetime, for sure want to replace it when a more efficient solution is available since datacenters are power and volume constrained and customers care about performance much more than hardware longevity (otherwise they’d buy instead of rent).
Why do you think it’s a waste? If you’re buying GPUs to rent them you’re almost buying a bond. If you’re leasing them, it’s even more obvious that you’re collecting the spread. The GPUs have a financial lifetime after which the business doesn’t pencil and they get sold for peanuts so you can put a better bond in your volume-power.
Consumer GPUs/CPUs tend to be operated at higher clock rates and voltages, because they need to win benchmarks. If you ever bothered to pay attention to how data centers operate their hardware you would notice that they have always gladly sacrificed 10% of performance if the total cost of ownership is reduced.
Since this entire sub-thread is in the context of used 3090s or consumer GPUs in general, you've failed to bring up anything relevant yet again.
Here is your strategy:
1. Increase power consumption by 50%: This costs you more energy to run the GPU, it also costs you more energy to cool the GPU, it ruins the GPU and since you hit power limits of your infrastructure earlier, you will have fewer GPUs in total.
2. Increase maximum performance by 10%: This is hardly noticeable, since the standard inference use case primarily involves taking advantage of the high memory bandwidth of a GPU. This means prompt processing will be 10% faster, or maybe your segmentation model that ingests video runs at 33 fps instead of 30 fps. You're optimizing for winning a benchmark with what will be used hardware in the future, that's asinine.
3. Throw away old GPUs or sell them for peanuts when they still sell for $1000 on the used market if they are in good condition and for $400 if they are damaged. I think the mistake here is obvious. If your GPUs are sold for peanuts, it's because you didn't take care of them.
Your business strategy is obsolete and based around the idea of pre COVID excess hardware capacity before there was massive AI demand where throwing out hardware made sense, because Moores' law was in full swing. Even Google is still offering their v2 TPUs from 2017 even though they've been long since obsoleted. Now in 2026, there isn't enough memory for consumers and people are snatching up all the hardware they can get their hands on. There were some big initial energy efficiency wins from implementing smaller data types that are no longer possible now that fp4 is the smallest possible floating point type that still makes sense and even if you go smaller, you can go down to two bits at best. The parameters are starting to become so small that 2:4 sparsity is becoming unattractive, because it adds one bit to the parameters.
2:4 sparsity for fp4 means 4+4 bits are compressed to 4+1 bits, but 2 bit parameters mean 2+2 bits are compressed down to 2+1 bits.
If you understand even a little bit about hardware, you notice that the tensor core hardware has already been optimized to the extremes and that there isn't much more you can pull out of it. Unlike CPUs there is hardly any control flow in matrix multiplication. The tensor cores implemented in Nvidia GPUs might be a little bit less efficient than an NPU/TPU based implementation (think Google), but there are no more obvious micro architectural improvements here. With CPUs the micro architecture has become so complex, that there may be ways to increase performance further, but for GPUs and NPUs, there is not much left other than process scaling. Further gains require better manufacturing processes from TSMC. TSMC introduced 3nm in 2022 and only started producing 2nm in 2025. That's a three year gap where barely anything happened and all the gains came from going from bf16 or half precision floating point, to fp8 and fp4.
Burning through hardware at high power consumption and mediocre performance increases is clearly not the way to go.
Will it continue to appreciate to infinity? Maintain its value forever? Or will something else happen?
The same argument you’ve made would work for tulip bulbs, dotcom prices, or whatever. Prices go up until they don’t. Exponentials don’t last forever and the intrinsics of technology assets depreciate: things wear out and are also replaced with better things.
> if there are no customers they’re not costing you more money, unlike your computer which you are using for training
So are you using the computers or not? I'd argue that if you're using them for training, then it's not wasted capacity. And if you're not using them, then you can turn them off, so you're not sucking up energy.
I don’t know but this dude at my son’s school has a 32GB RTX 5090 and it’s worth more than what he paid for; and he did the same trick with the RTX 4090 before that.
Until shortages are the rule, these assets are appreciating
"depreciating" is not being used in the right sense.
There is depreciation, which is taking the purchase price and dividing it across N number of years (typically 5). That's the D in EBITDA and is mostly used as a profitability calculation.
The depreciation of a GPU also gets mucked up in the current GPU financed market as well. DDTL loans. The people running the GPUs often don't even own the GPU, they lease it, so there is nothing for them to depreciate (D).
The analogy that a GPU is like a used car makes zero sense. There is no oil or tires to change on a GPU. They don't wear out in the same way that a rental car would. They are housed in climate controlled locations with clean power. They just don't fail the way that is portrayed in the press.
Useful life of a GPU is based on profitability. When does opex cost more than profitability?
Some companies, like mine, also have support contracts. Anything goes wrong with the GPU (or any part of the system), Dell comes and fixes it at no extra charge. We just migrate customers and workloads to hot spares while the parts are replaced.
As for compute going down in value... the 122TB of enterprise nvme and 2GB of ram in each server that I bought 2 years ago is now worth vastly more than I paid for it. I'm also renting my GPUs out for more money now due to supply being so tight and demand being so high.
Compute is about to come an appreciating asset in the near-term, and it some ways it already is.
The frontier labs are shifting from pricing grounded in the price of compute, to pricing grounded in the intelligence provided, or more specifically the economic value of that intelligence downstream.
The margins on that allow them to pay a hefty premium on compute and still come out ahead.
As they buy more compute at high prices, they're also pricing out competition from cheaper models. It's already become materially more difficult to get compute to run open weight models at competitive prices as a result of frontier labs in the last year.
In the short term, compute becomes an appreciating asset.
In the medium term, everyone ramps up production. Huawei and other Chinese companies work really hard to develop in-house alternatives. At some point, the hype cycle will peak and less money will flow into datacentres (yes, this will happen. It always does. Even for technologies that change society. The bubble always bursts).
The question is not if this will happen. It will happen. It's just a question of when it happens and how big the magnitude of the cycle is.
Not necessarily. The GPU leases Spacex has made are month to month, so they are taking on all of the risk. If demand goes down, they're the ones stuck with the assets.
I run a consumer AI product and the current reality of trying to get compute vs what it was 6-12 months ago is enough to justify it to anyone who has the money.
I think OpenClaw created a mania that was completely unfounded (Apple Silicon is worth dirt compared to literally anything from NVIDIA including consumer GPUs), but the prediction of compute becoming scarce was correct
Opus 4.7 has all the signs of a smaller model distilled from a newer pretraining run... except a smaller price.
Flash 3.5 raised in price pretty meaningfully over Flash 3
GPT 5.4 got a small price bump over gpt-5.3-Codex/gpt-5.2, then gpt-5.5 doubled pricing over gpt-5.4
Even open weights isn't immune: Kimi K2.6 was originally priced higher despite openly being 2.5 + more post-training, same with GLM 5.1 vs 5
-
All while rental prices are spiking month over month, and NVIDIA Inception discounted prices for buying are higher than undiscounted prices for buying 6 months ago...
Let's say it does all collapse. How would we know it's the 5-6% stake (which in my mind doesn't make them a "major shareholder") that was a circular deal that was the fall of the house of cards vs some other segment?
This might not be true. Someone was comparing Nvidia's production rate with known data center capacity, and they do not match. Their conclusion was that people (possibly even Nvidia) were hoarding GPUs- in the very short term this might be a good strategy, but GPUs go EOL fast. There are other stories about paused datacenter builds that match with this.
TSMC is definitely fully allocated, based on current 40 wk lead times for FPGAs..
All that means is that there's a bottleneck at the data center layer. When he says "dark GPUs" he's saying that there are no dark DEPLOYED GPUs.
This is a reference to the 1990's dot com bubble where internet infrastructure companies overbuilt network capacity, leading to the term "dark fiber". That was an indicator of a bubble because it showed that capacity was larger than demand. OP is saying that this is specifically NOT happening in the case of GPUs yet, indicating that demand still outstrips supply of compute.
>GPUs go EOL fast
We are seeing the opposite of what was expected, GPUs are actually getting more valuable because demand is so great, something that basically never happens. Even older chips have become more valuable.
>paused datacenter builds
It doesn't seem that datacenters have been paused because of lack of demand for AI, it seems mostly that there is a lot of pushback by cities to build these things and also there is a shortage of power to run them.
IMO none of these things point to a AI being a bubble (over-hyped, demand does not match the stated value). It mostly points to the opposite, there is massive demand for AI and every layer of the supply chain is struggling to keep up with that demand.
Adding to this, a lot of fiber installed in the 1990s is still dark. Multi-wavelength XYZ and other improvements mean the same fiber from 35 years ago can carry 100 or 1000x what it was originally designed for. Also, like Solar, all the cost is in labor. When they designed the Seattle/King County fiber network, they knew they would never have access/permits to go back and add more, so instead of running a single 12 fiber bundle the size of your pinkie, they ran 3 x 1024 bundles the size of your arm through the hollow bridges that span I-5 freeway and snakes through Seattle underground. Almost all of that sits dark today despite being in a very busy area, simply because fiber technology keeps getting better.
Yea, fiber is great. They can do hundreds of terabits/s per fiber today, and petabits/s is not far away. Bandwidth is fundamentally cheap enough that my ISP offers 50Gbps residential service!
> ...GPUs are actually getting more valuable because demand is so great, something that basically never happens. Even older chips have become more valuable.
Huh, anybody want to buy a GTX 680? Or even a formerly-SLI'd pair?
> IMO none of these things point to a AI being a bubble (over-hyped, demand does not match the stated value). It mostly points to the opposite, there is massive demand for AI and every layer of the supply chain is struggling to keep up with that demand.
Yes, the demand is there for the currently unsustainable price. Lets see what happens when the dumping of money into AI stops and the companies are forced to increase prices a lot.
Don't you think that under excess demand, production will ramp, competition will become available etc? These posts read like we're all out of fresh silicon or something.
Supply will catch up, it will just take 3-5 years, with the price rising the whole time. Basically a worse version of the Covid supply disruption where I sold my car for more than I bought it for years later.
The physical world can’t be patched overnight, and cutting edge manufacturing takes a long time. Fortunately we are in a very peaceful low tension world right now and no one would try burning down or blowing up one of those extremely important, irreplaceable fabs.
> IMO none of these things point to a AI being a bubble (over-hyped, demand does not match the stated value).
I agree the demand is there, but hyperscaler capex is what now? 3% GDP? This is an absurd amount of money and people who question whether the ROI is there have a point just because of the order of magnitude of this spend number.
My read is that xAI built a lot of compute for their own use, but they didn't get any adoption so they are reselling the unused capacity to recoup at least some of the costs. So calling it a good bet is kind of misleading
Don't you mean gas turbine purchases and questionably legal operation? But yeah I feel exactly the same way. The AI part of xAI looks like a mess but it seems that they still managed to score a massive win.
> Don't you mean gas turbine purchases and questionably legal operation?
The point is it’s running. They built fast before the backlash got organized. Now everyone has to deal with delays and thoughtful permitting processes.
Sure, they brought in artillery and a small freelance militia to shoot at the unionized workers, but the point is, the survivors are back working the mines...
The "backlash" is the poorest residents one of the poorest large cities in America trying to fight for their right to clean air.
Your point might end at "it's running", but holistic thinkers have no problem considering the how they arrived there, given what it's doing to these folks for marginal benefit.
It's not like xAI would go under if they had chosen a less populated location and waited to get permanent power.
> "backlash" is the poorest residents one of the poorest large cities in America trying to fight for their right to clean air
Sorry, I'm referring to the national pushback against datacenters being built in peoples' backyards. xAI didn't face backlash. At least not organised enough to stop them. Their competitors, today, are facing backlash sufficiently powerful to stop new datacenters from being put down.
The point is they're in a business no one would claim is particularly profitable but claiming a valuation like they're in a totally different business - one where they're not even top 3.
Its not that there isn't value in that business, but it's not the AI business either. Its the one where Oracle is laying off staff to try and avoid a revenue crash on future commitments.
Both Google and Anthropic would be trying to can this sort of rental arrangement as fast as possible since it's a mind bogglingly expensive way to get something you already do in house.
It isn't normally particularly profitable but given their lucky timing they appear to temporarily be doing quite well. When their tenants eventually vacate either they make a move to reenter the race for the cutting edge and get lucky or else they revert to a "boring" cloud rental business with near cutting edge hardware. That seems like an extremely favorable mode of failure to me.
This feels highly revisionist: they bet on becoming a frontier lab and were aiming for AGI.
If they were speculating on compute, it seems highly unlikely they'd have spent the operating costs for the last 3 years of model development and deployment instead of just getting even more compute.
And while there's no challenging the underlying proposition "AI has value", right now 95% of corporate users are still at the "throw everything at the wall and see what sticks" level in terms of model usage compute.
It's sheer brute force, tons of waste, seems like very little thought going in to fitting the implementation to the problem.
The value of compute can drop significantly in the event of users figuring out how to optimise for their particular need. And yep, there are wasteful applications that can burn whatever compute is available, but how much demand for that is there when it's properly priced?
Extreme example. Generating novel 4K VR video on demand. I'm certain there's a market for it, at $10/hour probably quite a healthy one, at $100/hour not so much.
Indeed, that hardware was bought on old RAM, SSD, etc pricing. These are now 5x the price.
To reap massive profits before depreciation is just plain smart. LLM space, model generation is just plain crowded now too. And everyone thinks a crash is coming.
They could also build out their own end-user infra, but letting someone else which already sells direct to the public do so, is sensible.
I know of the desire to show profit for the IPO, but my point is, this is a good move on its own.
The compute is useless if nobody is left to pay for the compute, once all the AI companies die, from all that debt getting called in, once everyone realizes it's a scam. (AI isn't a scam, but the financial deals and promises of unrealistic recoupment are)
The thing I've never understood about the ai investment model is the upside. What's the point of valuations that only make sense if you've built a digital god, when at that point you've literally got a digital god. I can't imagine the tangible value of money being high in that scenario
In fact, for all these companies to do what they're going to do, they need a massive, massive massive amount of data centers, a highly improbable number of data centers that need to be built in an highly improbably short amount of time.
And the capitals about to dry off in about a year. So it's a race between these improbable timelines on data center construction, with capital evaporating.
- iran destabilization will shift middle east capital to military spending and infrastructure repair
- Ukraine war similarly is triggering an EU buildup and reduction in us dependency
- all the IPOs indicate the companies themselves know the private investment is coming to an end so they need the retail investors to keep the boondoggle moving
Nah, man, it's all fine, they're just going to take down the entire global financial system doing this crap, and by global, I mean <<everyone's>> pensions are going to take a hit, even "fully funded" pension systems.
> Both Alphabet/Google and Amazon have invested recently into Anthropic and are doing all sorts of financial chicanery
bko didn’t say there isn’t circular financing going on. They’re just saying this isn’t an example of it. They’re right.
It’s a potential conflict of interest. And if the agreement is fake—if Google cancels without paying the cash—it could be market manipulation. But the influencer space likes to latch onto jargon, and the one it’s overapplying right now is circular financing.
The comment you’re responding to and the comment above it are about circular financing. It’s reasonably to assume that’s the same chicanery you’re talking about; expecting everyone to watch a random video to understand your comment is unreasonable.
I listed a bunch of data points that make no sense (profits spiking 50% in a non-Christmas quarter for companies) and weren't directly tied[1] to the circular financing.
They were unrealized gains on non-marketable equities. It’s clearly disclosed and done according to GAAP. It’s put under other income precisely so analysts can strip it out when modelling long-term trends.
Like, yes, if SpaceX goes to zero Google would have to realize losses and probably lose a quarter or two of GAAP profits. (But not cash flows. Cash-flow wise, it may wind up being positive due to tax effects.) It’s a risk factor, of course, but far from making no sense.
None of which is particularly relevant to the deal at hand other than in raising a potential conflict of interest among related parties.
When I said "it makes no sense", I didn't mean "the accounting math doesn't work out". I meant "raising a potential conflict of interest among related parties".
This whole AI financing this is the motherlode of "potential conflict of interest among related parties".
And people who are obtuse enough to ignore this because it's not illegal right now will discover 5-10 years from now that laws are written in blood (or massive bankruptcies).
> whole AI financing this is the motherlode of "potential conflict of interest among related parties"
Sure? Lots of things are potential conflicts. In the Google and Anthropic deals, I'm not seeing evidence of problems.
And that's saying something, because we have a lot of evidence of actual circularity or related-party deals being done with no arms-length anything across AI, in many cases in ways that definitely do see like they are illegal.
> Sure? Lots of things are potential conflicts. In the Google and Anthropic deals, I'm not seeing evidence of problems.
And that's where my video comes in. Google and Amazon are very likely juicing up their share price. Of course, in this day and age we can't prove it's a pump and dump anymore...
The music would have a risk of "stopping" if these deals were backed by a speculative entity. However AI actually has real value/revenue, and is not a speculative product (i.e. people aren't buying tokens to resell them, a token is "consumed" at moment of inference)
Enron collapsed due to legitimate fraud. To imply Enron is an apt comparison requires assertion that AI companies are actually cooking the books. Is that what you are saying?
The ARR were fine but showing skewed quarterly profitability numbers by slowing down research due to hitting compute capacity suggests otherwise.
I am certain Anthropic spent less on building the next model this quarter if they make it to profitability due to the shear fact that they don't have enough compute.
Which solves the profitability problem with relative ease momentarily.
Also just to confirm, AI subscriptions are definitely being sold at a loss how big I don't know but these models are much harder to run.
API is definitely being sold at a decent profit.
So if you rate limit users and do usage billing + lower research costs which is a money pit temporarily.
(Proof is the fact that we don't have a new pre training run since 4.5 yet, they used to do one every 2 releases)
4.9 will probably be the same.
Next model Mythos doesn't seem to have a successor yet and was trained previous quarter most likely, they don't seem to have pre trained another one just improved Mythos if at all.
As much as I am into AI these attempts to show that there can be a profitable quarter seem like cooking the books, even if we assume no shady dealings otherwise.
Unless one of the Labs can say for certain training is going to stop they can't be profitable and I don't think training can stop because marginal gains is all they have.
8-12 months behind narrative for Chinese labs literally is going to kill the company that stops training first.
If we assume only a 3-6 month gap once China has more compute, then well then even if they keep training the lack of ability to arbitarily scale data centers in US, will kill them first.
DeepSeek V5 might actually just end the AI race for good.
Also given Mythos is atleast a 10x model compared to Opus, then it's pricing is likely going to be 10x as well so well token prices are likely never coming down, especially if these companies want to IPO.
Why would V5 kill the AI race? Do you believe that there are diminishing returns on model intelligence when applied to real-world tasks?
I think there are accelerating returns: i.e. a models are still not good enough to be “drop in” remote workers, but once that threshold is passed, the value of each token of inference has a far higher multiplier.
This justifies the buildup. However not everyone agrees that model intelligence will continue scaling thus they assert that eventually the economics will hit a wall.
>Also given Mythos is atleast a 10x model compared to Opus, then it's pricing is likely going to be 10x as well so well token prices are likely never coming down, especially if these companies want to IPO.
I don't know why people say this when cost per unit of intelligence has been going down continuously over the past few years. When Opus 3 was first released, its API cost was $15.00 per million input tokens and $75.00 per million output tokens. Opus 4.8. which is significantly better, is $5.00 per 1 million input tokens and $25.00 per 1 million output tokens
Assuming 2-3 years from now when V5 is out China would have mostly caught up in compute, and honestly that's it China can scale up compute a lot faster than US maybe a few countries can match it, or help match it but won't happen while US Iran thing is going on.
Further the human costs in the loop for AI training are insanely low or atleast substantially lower outside of US, so sure without the Nvidia upcharge I think everyone else who can use Compute from China is at an advantage.
If the assumption is AI is scaling issue then China will win because they can do infrastructure. Maybe if US wasn't in a trade war with rest of the planet there was some hope but I don't think so.
Once Deepseek figures out the new compute and can get it on par with Nvidia's clusters even if by using 4x the energy(cause they can). I don't think OpenAI or Anthropic can maintain a lead, if they don't have a lead the pricing difference will kill the AI race.
The best case scenario is OpenAI and Anthropic are dead in 2-5 years once China is caught up.
The worst case scenario where AI is not a productive boost is that well the thing pops.
Either way I don't see how this works out. Sure US govt could bomb China that's always an option.
>The ARR were fine but showing skewed quarterly profitability numbers by slowing down research due to hitting compute capacity suggests otherwise.
I have to say, I find this really puzzling. We know for a fact that Anthropic are making bank on metered inference. That's their biggest source of profitability, we are seeing software companies start to majorly adopt coding agents over just the last few months.
Right as the biggest driver of enterprise adoption is accelerating, and it's tied to their biggest profit vector, you find it suspect that their profits are increasing significantly?
Also, can you clarify what you mean by "slowing down research" exactly? Do you mean they're not doing big pretraining runs? Less compute available for researchers? Scaled back RL?
>Also just to confirm, AI subscriptions are definitely being sold at a loss how big I don't know but these models are much harder to run.
Maximum usage of AI subscriptions is a loss, but do we actually know how that nets out? Has anyone done any research to try to figure that out?
> can you clarify what you mean by "slowing down research"
He is claiming that they have been investing less in R&D and that this is juicing their numbers in an unsustainable way given how close the competition is to catching up. His evidence is the content and cadence of model releases recently. (I'm not taking a position one way or the other, just clarifying for you.)
> Maximum usage of AI subscriptions is a loss, but do we actually know how that nets out?
They almost certainly don't have to care. All the enterprise accounts use the API pricing AFAIK and that appears to be profitable and is expected to be the vast majority of the usage in the medium to long term (if it isn't already).
> API is definitely being sold at a decent profit.
Where do you get this from?
Enterprise plans are being cancelled or limited all over the place (Uber, Microsoft). I doubt Anthropic would be leveraging a loss leader with their consumer plans, while catastrophically hemorrhaging customers on the enterprise.
They are either operating at a loss (possibly a minor one), or a minor profit (which is chasing customers away).
If they were comfortably profitable they wouldn't need to participate in the circular deal circus.
It would be insane, if they can't serve the models at a profit sure at current GPU prices the profit might be 10% or lower.
But at realistic gpu prices it would have been close to 30-60% based on how big the models actually are and how much they have optimized the stack to serve them.
1T parameter models like Kimi K2.6 can be served for 1/10 to 1/5 of the price of opus 4.8 for perspective.
Sure opus is 2x the size and hosting might be non linearly scaling so still it should be around 50% margin at regular gpu prices.
If it isn't I would be very surprised.
Also for enterprises we joke but Google is not paying same rates as us there are big massive enterprise discounts. I have heard upto 20-30%...
OpenAI is supposedly even more generous.
I don't think API is being sold at a loss at the end of the day even if the API profits are marginal 10-20% because of insane GPU prices now.
Please address the primary point first: Selling some product does not disprove speculation.
In the case of Enron, people were obviously speculating in its stock, and that remains true regardless of why it collapsed later, or even whether it collapsed at all.
I say "first" because if you still can't agree that speculation in AI stocks even exists, then it's pointless to discuss what people might be doing to exploit or encourage it.
Speculation exists for every security. However wrt revenue numbers, Anthropic/OpenAI’s revenues are largely made of companies/individuals purchasing tokens. Enron’s was accounting which stated future potential revenue as current earnings. They are not the same. Enron pulled off a lot of shady schemes to hide their accounting practices. All of the “circular deals” AI labs are doing are publicly known and clear to see, so its not like anyone who knows what a circular deal simply knows something everyone doesn’t.
Also to be more specific about our point of disagreement, I think we are referring to speculation in different domains. When I brought it up, I am referring to the fact that any companies whose revenue is driven by a speculative bubble (like what precipitated the 2008 crisis) would be at risk of massive losses "if the music stops". Anthropic/OpenAI aren't flipping assets. It is true that VC funding is based on speculation, but their core business model is producing massive revenue growth on selling tokens.
It's an interesting point that the token revenue will presumably survive a crash in stock prices. But (IIUC) much of the new infrastructure is funded using stock is it not? So it seems like token revenue theoretically surviving doesn't address the risk to the rest of the economy here. And if the economy takes a large enough hit then presumably so will token spend because someone has to pay for that after all.
Sure their actual immediate revenue is driven by concrete numbers but when the rest of the economy is reorganizing itself based on their projected future revenue is the former observation still relevant?
That is true, if all the new data centers don’t produce revenue then there will be a crash. However you’d have to bet that the models won’t stop getting better, or if they still keep getting better, that somehow better models does not translate to increased productivity. Would it be wise to look at how AI has progressed over the last 5 years and make that bet?
It's possible to question the accuracy of the projections without disputing that the numbers are expected to go up. It's not that the new data centers wouldn't produce any revenue but rather that those numbers are where unfounded speculation could be happening. If and when those numbers fail to materialize (or when investors revise their projections) would presumably be the point at which the music stops.
Recall that the exchange earlier called into question the similarity or difference to enron. Sure, the current revenue numbers don't appear to be cooked but if the future revenue numbers are unrealistic and everyone is using those future numbers to make their decisions then isn't the end result roughly analogous? Blatant fraud not withstanding of course.
Note that I'm not claiming the above to be the case. Merely illustrating the commonality and acknowledging the possibility.
Remember when nvidia asked us to stop calling them enron because unlike enron they actually admit to doing all the things enron did so it's not illegal?
Circular dealing or round tripping is a form of cooking books and sometimes results in accounting fraud. Especially when circular revenue is booked without cash flow growth. Do you see cash flow growth on any side of these transactions.
The value of AI companies is speculative just like the railroads were. Railroads also have real value. But you have to have everything ready to use those railroads to make money, or they're just steel bars in the dirt and a big loud heavy thing that moves along the horizon. Too much speculative investment in the railroads (in part) led to the panic of 1873, because just having a promise of a return isn't the same thing as having the return.
(note: Allied Irish Banks and Anglo Irish Bank are different organizations with the same initials; the latter is the massively fraudulent one run by Sean Quinn who did eventually see a small amount of jail time)
Unfortunately, the entire US economy is being propped up on AI stocks. If they are allowed to crash, the consequences would be extreme all across the board. See the recent worming into index and pension funds. If they collapse now, a lot of regular people are going to get wiped out.
Should the government bail them out or somehow stop the collapse? Arguable. Will they anyway? Almost certainly. These companies have engineered themselves into a position where being allowed to fail would wreak catastrophic damage to the national (and global) economy precisely so that the taxpayer will be left holding the bag if and when it all comes crashing down.
Capitalism is rotten to the core and there's no fix for it.
> These companies have engineered themselves into a position where being allowed to fail would wreak catastrophic damage
Where is this assumption of malicious intent coming from? This has all been fueled by a global AI hype that might or might not prove to be justified in the end. The overall economic situation looks (IMO) quite similar to that of the railroads in the US and those did ultimately fail and were nationalized(ish).
The current situation is hardly limited to the US and capitalism. China also appears to be actively reorganizing their economy around AI.
And you would have been massively wrong. People have been complaining about quantitative easing since post GFC, and if you took the figures at face value, those would imply inflation was nearly 100% between the end of GFC and before the pandemic. Whatever you thought about the post-pandemic inflation, the period between GFC and pre-pandemic definitely did not see the level of inflation implied by those figures.
CPI works by asking how much people pay for rent. If home prices raise 20% in one year (not at all unreasonable in various times in the last ten years), it takes a long time for that to be reflected as many people have their rents fixed, some people have rent control, some landlords will only raise rents on new tenants, etc.
for existing, well established renters, sure. For new renters, or people who move, or people who face housing insecurity, or people who want to buy homes, it's much higher and not accurately reflected in CPI.
> "The idea that inflation and the money supply are linked is one of the most dumb one in folk economics"
"folk economics" implies it is by untrained people.
Milton Friedman's famous quote of "inflation is always and everywhere a monetary phenomenon" shows that he deeply believed the relationship between inflation and money supply, and one certainly cannot call Friedman a "folk economist" considering he won the Nobel prize in economics and was a professor at the University of Chicago.
Note: I am not saying he is right or supporting his belief. I am merely stating that such a belief is not a "folk economics" belief. This belief is still very prevalent in the freshwater schools of economics. [1]
As a personal anecdote, at Ronald Coase's 100th birthday party, I personally got Gary Becker and Richard Posner debating a very related topic (whether and by what degree the velocity of money of fluctuates and whether helicopter drops of cash would have been better during the early days of the money supply collapse in 2008/2009 than just giving money to the banks). In a room full of Nobel Prize winning economists in 2010, there was a very rigorous debate on the topic.
> "folk economics" implies it is by untrained people.
The problem is mostly its appropriation by untrained people though.
> Milton Friedman's famous quote of "inflation is always and everywhere a monetary phenomenon" shows that he deeply believed the relationship between inflation and money supply
Creationists theoreticians believe in creationism too. The problem arise when their theory reach the mainstream… (Influential people inside the Swedish Central bank making a fake Nobel prize to promote these ideas didn't help of course…)
Lagged processes are one of the most fundamental concepts in economics. If merely recognizing the possibility that one could be at play here is throwing you for a loop, you need the simplified monetary model more than most.
Where's the hell is the lag on these graphs though!? The money supply grows both before, and after the inflationary spike. (And the fact that it stops increasing when inflation is high is not surprising at all, by the way, high inflation make the central bank raise interest rates, which reduce credit, which is where money comes from).
We are basically dealing with the fallout of the 2008 GFC bailout to this day.
The fiat economic system is irreparably broken, and we are circling the drain. Another bailout is _probably_ inevitable. But the cycle sure as hell isnt resetting and we are speeding towards something... what it is is unclear though, and when is also unclear.
The part people cant wrap around is the scale of it and the time it takes to go through the super cycle. Theoretically, it all started with the Dot com bubble, which indirectly cause the housing bubble, which caused the GFC. Which caused whatever happened in 2019, which caused QE in 2022 under the guise of COVID, which is causing whatever the hell is happening now.
Capitalism has become uncorked, and money is irreversibly flowing to the top at an increasing rate. The logical next stage is that like 75% of the world's population is literally not even part of any economy. And that doesnt really make any sense
Sigh, no. Money is not flowing; company valuation might be, but that's temporary and only works if the company keeps delivering insane amounts of value.
So the founders sell plenty of stock while the price is high and then when their valuation crashes sure they "lost" half their net worth but the other half is still there.
I'm not saying that's what's happening, just making it clear that company valuation not being permanent is not a valid argument against money flowing to the top.
yeah I intuitively have felt something like this has been happening, too. And finding the evidence is such an immense task, and feels way out of my current energy level.
When COVID was ongoing there was a term floating around I liked, "Psychosis" was it. The spell is like that of, denial? Terror & shock?
Trauma might be better?
Looking at trauma responses and how to detect it in humans is an interesting perspective to look at all this with. Personally, if I look at it from "people are afraid, traumatized, defending themselves" and use that to extrapolate how most people (the masses, the non-rich) would act and also the rich - that points me to why theres such a sudden hastening of action and pace of wealth up towards the top in the name of AI & war.
You seem to imply that with this deal their shares are worth 88B but without it they're worthless.
It's very hard to know how much the deal actually increases SpaceX market cap, but unless Google exits their SpaceX position soon it doesn't even make much sense as a circular deal.
Executive compensation is often based around share prices, so this can be worth quite a lot to the people making these decisions without any long term upside for the company.
If you want to understand how companies behave you really need to look at things from the perspective of people making the decisions.
I don't see Alphabet share price changing much just because of SpaceX being valued 2T instead of lets say 1T (being extremely generous). In fact this deal will hurt their profits, which is more likely to hurt Alphabet stock price than the valuation of an asset that they hold.
> I don't see Alphabet share price changing much just because of SpaceX being valued 2T instead of lets say 1T
Half of Alphabet's revenue increase last quarter came from marking up unrealized gains in their Anthropic investments.
I'm not saying Alphabet is doing this to juice the share price, but I want to point out that they don't have to sell shares to post banner earnings results and see a 10% jump in share price overnight.
They're worthless to the S&P 500 which requires four quarters of profitability. SpaceX is running at a $5B loss. Google 'buys' $10B of compute every year...
There's no realistic way for the music to stop. The demand for LLMs is staggering and the big providers are charging full freight for inference. They might not make back the money from training but these data centers are definitely going to be fully utilized for at least the next 5 years.
> the big providers are charging full freight for inference.
Except they're not. Anthropic's claims of temporary profitability line up exactly with when SpaceX is giving them discounted compute, OpenAI's such a shitfest they threw the CFO off the glass cliff for daring to push back against the IPO. "Profitable on inference" is an unsubstantiated rumour.
Just look at the copilot changes. Demand switching to other providers immediately when prices rise, and there's not even certainty that the new copilot prices cover costs.
> They might not make back the money from training
This is an understatement. With all the datacenter buildout, they need trillions. For the investors get their money back and the bubble to not implode, they functionally need to unemploy everyone in the US.
Unemploying everyone was what openai described as their success condition when it was founded a decade ago. There was a q&a on their website that said "How will you know when you have reached AGI? When the system performs most or all economically valuable work." Lots of people thought they were joking, or it was marketing, but they were 100% serious from the first.
I think they've since changed that definition, but the reason for it is their agreement with Microsoft which stops granting MSFT IP rights to OpenAIs tech once they reach "AGI". So, it's in OpenAI's best interests to be able to claim "AGI" ASAP.
The pricing on Open router is clear. Anthropic, OpenAI, and Google all garner a massive premium over deepseek and qwen. There's no other realistic explanation except that they're making bank.
Why do you think Chinese companies can do that? It's government subsidising price they do it with literally every ibdustry.
Home grow a bunch discount them federally, let them wipe the foreign markets.
If AI is threatened by china why would US NOT do the same? If they did they're in a much stronger position to do so than china. Cheaper energy, more cash, stronger industries.
Infrastrucure is thr kind of thing that only a foolish US admin would let fall apart to their advesary.
This is just silly. Deepseek has published so much regarding speeding up and making cheaper inference and people are still harping on the government subsidies thing.
So what's all the project Stargate stuff? Subsidies only work when China is doing it?
Deepseek is actively sacrificing performance for cost, which is very clear in their latest model releases. They are not attempting to get to number 1 in benchmarks, and they say it clearly in their own publications.
Furthermore, being open weight, anyone can sell qwen and deepseek compute, not just Ali and deepseek themselves.
> Home grow a bunch discount them federally, let them wipe the foreign markets.
US is doing the same and was doing that for decades now. American companies operate on loss for astonishing amounts of money and consider it completely normal. One gotta love complains about Chinese companies selling under price coming from American tech industry.
And yet they are not profitable on an ongoing basis, and aren’t even claiming to be.
The supply is currently constrained because 50+% of data center plans were cancelled as a result of the impossibility of the buildouts happening in a timely fashion, and subscriptions are charging a small fraction of the actual cost of inference, leading them to all bleed money, hence the rush to IPO to get one last infusion, since many of the past investors have publicly stated they aren’t putting any more money in until they see an ROI.
Companies are hitting their budgeted limits for AI tokens less than half way through the year and reporting that they aren’t seeing enough benefit to substantially increase that budget, and so they are scaling back use and asking people to be prudent rather than token maxxing.
In the meantime subscriptions still exist in the form of chatbots and it’s easy to exceed the inference cost of the provider by simply using your daily, weekly, and monthly limits.
The reality is that we just don’t seem to be at a point now where people are willing to pay full price for the perceived value. Perhaps we’ll get there within another generation or two of hardware and software improvements.
Can 5% of the population even pay for that? Some kind of huge increase in prices for compute and inference and companies maintaining large bills for AI assistants for key employees or teams (1000-2000$) seems most likely to me.
Anthropic is a five year old startup, if they can be profitable that quickly in the AI space, even if only temporarily, I'm not really seeing the problem?
These companies are going all in and growing rapidly, because they want to dominate the market and since it is difficult to differentiate between competitors, even being third place is a terrible place to be in the consumer facing AI space.
The demand is finite. There is clear evidence that it has limits. When costs become great, the consumers set limits, create budgets and seek alternatives. Consumers are still figuring out where the cost/benefit lines are, and we can all see that the lines at least exist.
> the big providers are charging full freight for inference
They're not and it's not clear why you seem to believe that. The immense capex for buildouts, training costs, etc. are not rolled into inference costs. Moreover, companies are already rapidly starting to re-evaluate token spend.
Data center operators are in the business of selling electricity. They do not command large PE multiples. This is an even worse business, because xAI decided to also be the bagholder for the NVIDIA graphic cards. Not to mention they finance an unreasonable number of 20-somethings on way too large salaries with shitty opinions and no AGI delivered.
This take clearly has a bone to pick. But ignoring that, the first sentence is just not reflective of the reality here—xAI is making a killing on renting out its GPUs, way more than "just power". The dynamics that normally make infrastructure providers have slim margins don't apply when demand far outstrips supply; the situation right now is closer to monopoly pricing power.
It will likely take a few years for supply to fully catch up, which means xAI will eat well for a while.
I can see a world where a few data centers come on line this year and reduce margins a bit, but it's crazy to think the margins will go to "cost of electricity plus a few percent" anytime soon.
In the article, it states that the two deals will cover the entire cost of SpaceX's AI buildout in 18 months. OpenAI and Anthropic would kill for that kind of cashflow.
xAI is a failure of an AI company from a consumer perspective. They invested a large amount of money into owning their own infrastructure, while driving away consumers with their right-wing or "alt right"-ish branding and having a reputation of X users abusing the AI services.
Turns out there was another company with a much better reputation for which the compute is a better fit. Now that the data centers are being put to use, they actually make them a little bit of money instead of losing money.
Datacenter operators who rent space are selling electricity. SpaceX is selling a fully built datacenter with compute designed for a specific purpose. They’re operating at a higher level of the value chain and can charge accordingly.
What's their novelty or moat to maintain the value chain? And why do we only see google, who already owns it, raising their hand to rent at these prices?
I’m not sure they need novelty or moat. AI compute resources are so scarce that inference providers will buy whatever is available. SpaceX sells inference hardware in bulk, with a proven track record of running inference and training workloads at scale.
What? They make money from their own inference and models too, which they can train effectively for free by funding their operations with rental income from their last gen datacenter.
SpaceX and Tesla used aggressive vertical integration, manufacturing simplification, and reuse to radically lower the cost of building rockets and EVs. It's not unreasonable to speculate they might be able to do the same for hyperscale compute.
They're not any sort of bag holder. They're going to make back what they spent on these data centers in a year.
It's a fairly sweet deal for everyone involved. Anthropic/Google get to sell more tokens and xAI gets a war chest for another bite at the apple. I don't have much confidence that they'll do anything with it but that doesn't mean these deals don't make sense for them.
* Company valuations around LLMs are not realistic
Both can be true, much like they were during the Dotcom bubble. The internet turned out to be a pretty real thing. A couple examples below might feel familiar in the next couple months/years.
> Blucora (then InfoSpace): Founded by Naveen Jain, at its peak its market cap was $31 billion and was the largest Internet business in the American Northwest. In March 2000, its stock price reached $1,305 per share, but by 2002 the price had declined to $2.
> Broadcast.com: A streaming media website that was acquired by Yahoo! for $5.9 billion in stock, making Mark Cuban and Todd Wagner multi-billionaires. The site is now defunct.
> eToys.com: An online toy retailer whose stock price hit a high of $84.35 per share in October 1999. In February 2001, it filed for bankruptcy with $247 million in debt. It was acquired by KB Toys, which later also filed for bankruptcy.
> GeoCities: Founded by David Bohnett, it was acquired by Yahoo! for $3.57 billion in January 1999[20] and was shut down in 2009.
> MicroStrategy: After rising from $7 to as high as $333 in a year, its shares lost $140, or 62%, on March 20, 2000, following the announcement of a financial restatement for the previous two years by founder Michael J. Saylor.
Here's my theory about the dotcom bubble. The market correctly identified the internet as hugely valuable and correctly identified search engines as being able to capture a large share of this value. Consequently early search engines, chiefly Yahoo!, obtained (merited) high valuations. What Yahoo! did with their stock (they IPOd in 1996) was go on a big startup buying spree. This is what actually started the bubble: invest in some random dotcom crap company in the hopes that Yahoo! swoops in, buys and you get a big payday.
What caused the crash was Yahoo! being unable to do anything with their acquisitions and Google coming out with a better search engine, undermining Yahoo!'s core product. Google basically pulled the rug from under the dot com bubble.
The situation we're in now with LLMs is different, if I'm right we're actually pre-bubble, the bubble hasn't even started yet.
I was expecting this comment. You know the answer. A scam will keep scamming.
There are also legitimate companies from the dotcom bubble era like amazon, microsoft, and intel. They all were vastly overpriced during the dotcom era. Probably also now lol.
I am of the same mindset as you, but you also have to look at PE multiples of Cisco in 1999 and Nvidia today. One being the "ammunition" supplier in the battle for the Internet, and the other supplier in the battle for AI.
Cisco was over 400 at one point and Nvidia is around 30. Not quite the same.
Other players today:
- Digital Realty 48x
- Equinix 75x
- CoreWeave (still losing money)
There is likely a bubble of some type here, but I don't think this is the same as the Dotcom bubble.
The circular financing aspects in the current era are really obscuring some of the financials. There are also very legitimate companies offering very real products. The big issue today is that things feel a lot more obscured and interconnected, which makes it hard to discern shit from gold. Does not help when the gold and shit are swimming in the same circles and shaking hands with all the same people.
Google also just announced a new equity raise of $80B. I have no idea if doing this via equity vs debt is trying to suck some of the wind out of the IPO Market for Anthropic and OpenAI but it’s going to be interesting to see how the markets deal with all the new equity being floated. Someone isn’t going to hit their raise targets and the later IPOs may be the ones holding the bag.
Given all the rules changes related to IPOs that also just went I to effect, I can only assume their hope is that the public is holding the bag when the music stops and that they think it is going to happen quick.
If the S&P 500 dropped 20%, that's about a year's growth. Long-term investors who bought before that would be poorer than they thought they were, but they're not worse off than they started and there wouldn't be any particular bill to pay. If they're a long term investor then they can wait for it to come back. (A similar argument could be made for larger drops.)
The real suffering comes from whatever effect there is on the rest of the economy due to a recession, more layoffs, etc.
No, asset values are not like energy. There's no conservation rule.
When stocks get bid up, market valuation goes up far more than the amount of money that changed hands. Most of the market cap appears "out of thin air." It's just what people think it's worth.
And when the stock goes down again, it goes back where it came from.
The investors who bought stock at too high a price lose some of the money they put in, but there are others who never paid that price.
But thats the point. Your last sentence is the problem:
Investors proping up stuff by 20%, 401k and etf etc. regularly invest, investor drop out.
Who loses? 401k and etf.
Money was transfered.
Same shit happen to my company share: Price jumps 40%, company has to buy them because of employer benefits, I auto buy them, price falls back by 40%, what happened?
Investors extracted money out of the company and me.
In fact [fiat] money does appear out of thin air (well, created by banks when they originate loans) - and has to to support a growing economy. Unfortunately, for various reasons, rather too much has been appearing, and has been funneled to the already wealthy.
This doesnt fix the systemic issue. Most people put their money in a target fund and leave it alone. Those target funds are at risk of being forced to buy these over-inflated assets. The incentive to do this is there because those target funds and naive investors exist.
The diverse investment is the reason that funds will be forced to buy these worthless stocks. It's a direct transfer of money from the working class to the extreme capital class.
If you're good with that, I'll send you my PayPal so you can get me my 5 bucks. It's a tiny fraction of your overall cash flow, whats the big deal?
When you destroy pensions by crushing organized labor, create 401k incentives, and place your new captive audience by default into a certain investment class, a whole lot of people are going to leave it there. Whether the provider forces anyone to do anything is irrelevant, it creates second order impacts that ultimately lead to what is perhaps the greatest attempted fleecing in market history
Pensions also invest in stocks, though the difference is that pensioners have no control at all. Of course, they also have guarantees of a certain level of income.
I think the main problem with the 401k is that not enough people actually contribute to one. Or they don’t put in enough.
But I very much doubt the average person who’d invested enough over several decades in a 401k feels like they got fleeced.
Until someone can come up with a better option though...
Note that a pension plan that invests for you blindly is no better - either the returns are so bad that they are a scam, or they are investing in stocks anyway and so you get the same results but less control. Similar for things like social security, they are either worse options or you need to pump stocks.
This is what financial capitalism and "democratizing finance" has meant in practice. Rich people have access to different types of investments, and by the time those trickle down to common investors the juice has all been squeezed out. Whatever the trend is, by the time you hear about it the market has already been arbitraged by faster investors with more resources.
We are not going to come up with a market-based solution to fix income inequality. The solution, as much as people in the dwindling middle class resist it, is a strong social safety net coupled with a hard reset on taxation and housing policies. Nobody should be homeless, nobody should be allowed to starve, but you might have to accept that your 401K goes down in exchange for a government guarantee of housing and food.
This is hard for people to accept because they currently have equity in their home or a 401K to save them from starving. But those are transient, individualistic solutions. You can lose your house. You can lose your 401K. Society should be taking care of each other in a broader way than letting everyone accumulate a little, private pile of money.
I didn't even mention venture capital because the win rate is so low. When you have billions already, then maybe you can buy $10M lotto tickets that get pitched to you. If you're a regular guy and want risk exposure like that, you can buy penny stocks.
Everyone is so fixated on the winners, that they completely forget (or aren't even aware) that there a many many times more losers.
I think people understand that there are losers. What they are complaining about is that the losers can dump $10M on a lotto ticket and not feel any pain when it disappears. If all those with money are are placing huge long-shot bets and cashing out when they win then what does that say about the state of the system, and markets in general? I don't know exactly, but I don't think it's good.
When money is cheap you take it. Google sees all the capital waiting to pour into these AI IPOs, and correctly assumed they could tap into that with little dilution.
It makes good financial sense for a company to sell shares when the price is high and do stock buybacks when it's low. I guess they think the price is on the high side?
Also, selling shares puts them in a better position to survive a downturn (more cash, less debt).
In a real competitive market it would never make financial sense to do stock buybacks because competition is so fierce you need to invest it all in R&D and sharp prices for your customers. See the Chinese EV market.
Stock buybacks are also a tax trick.
They're just holistically evil and should have never been made legal.
Google is also issuing a bunch of debt this year. It sounds like they need a lot of capital and want to keep a particular debt/equity ratio, rather than having a strong opinion on their share price.
When the music stops, all the AI companies fall, except Google. Google remains the world's largest advertiser and a cloud provider. They actually make money, own their own hardware, etc. They can survive a stock market bust and still come out victorious, because they still have a product people want to buy (ads). The rest don't.
The sci-fi SpaceX S1 talks about asteroid mining and other imaginary chimeric stuff like space data centers... while 80 to 90 of the case is about AI. But their AI case is like BMW bragging about their thriving auto business...while renting all their car factories to Toyota.
It’s funny because that is a guy with enough sense to both see what is going on and also not short it, because he knows that none of this actually matters with regard to stock performance for a properly frothy investor class.
It's not interesting to say "this is a bubble!" I've heard that about virtually everything (and in many cases it's likely true). What is interesting is pointing out the mechanics that make the bubble pop.
This is precisely what makes the movie the Big Short interesting: we see that people did identify, within a reasonable time frame, when people would start defaulting and how that would cascade into a true crisis.
It's pretty clear that while the fruits of AI are quite useful, the entire thing is rife with very questionable financial engineering... but I still don't know what it is that makes all of this break. For example, it's obvious that the SpaceX IPO is a massive wealth transfer program, but it's not obvious that it will immediately end in a crash. Given how irrational the stock market has been, I don't see a reason it can't continue to be irrational for long after the bag has been handed over to the retail investors and retirement funds.
If you sum the valuations of the company from its individual parts, no one sane would value it more than half a billion. But look at TSLA a P/E still at 370.
Who is the smart and who is the idiot?
The one who invested in it $40 or the one who was saying that even at $40 it was already too expensive?
Retail investors are currently being set up to hold that bag, and presumably the companies themselves will get government bailouts, so the taxpayer gets hit coming and going.
It's not even subtle at this point, what with the attempt at S&P rules changes, the insane valuation, the attempt to change the trade-through rule, and more.
The DDR5 will be registered DIMMs. The GPUs will be 600W paperweights with a custom form factor. Similarly the NICs and other PCI-E accelerators. The motherboards also adopt custom form factors to fit in racks. The hard drives will be using SAS connectors. The flash will be in E1.S form factors.
The server CPUs that you want for a home desktop or small server, high clock SKUs, will be in high demand.
Any savings for someone willing to build a system from second-hand server hardware will be eaten by using adapters or sourcing a rack.
I'm not saying you won't be able to make a slightly outdated frankenserver with more compute than you need, I'm saying that's not going to bring down prices for Grandma's machine that she needs working to check on her retirement account.
I dont think so. These entities and the hardware they own would be bought for legitimate AI use long before they'd hit the open market. AI is very useful, and even profitable at the inference level. It's just an open question whether this monumental amount of spend for research is worth it.
It's not just that there's a circular deal it's that they're prevalent. And worse, with frontier labs IPOing seeking astronomical valuations that means a lot of the public is now exposed too (even if they don't all get fast-tracked into eg; the SP500).
The problem is the valuations assume astronomical growth... that is likely impossible for all of them to simultaneously achieve. Which means something's got to give.
Google rents from SpaceX enough to show profitability, so that SpaceX can IPO and make googles early shares worth more than enough to pay for the renting they're doing.
Great deal for Google but they end up basically just paying spacex to pay them back, right?
I believe you've described "investing with a hope for a profitable return" which is usually the point of investing.
Circular investing is a thing that is happening with all of these companies related to language models. Google hoping for a ROI isn't a great example of that.
In accounting terms it’s not an investment it’s an operating expense or in your example a personal expense, but if you leased a property and operated a business (ie an AirBnB for an apartment) it could be considered as part of an investment as it’s a means to make a profit.
Buying the 5 percent stake is investing, but is paying them to be sure they can IPO normal? It reminds me more of Microsoft paying apple or Google paying Firefox or something.
Answering that question requires determining how much of the valuation is predicated on growth in AI spending from Google->xAI, but not counted as a forecasted expense for Google, and similar for other deals.
Circular deals aren't bad; what's potentially bad is if those deals are misinterpreted by active investores.
This, along with many other recent deals, shows that there is no real competition between these mega companies. They're at this point only orchestrating the market (or should I say scheming) to build an oligopoly and move as much resources and money to the hands their little group through circular deals.
Someone told me this isn't "fraud". (Was in another one of these hacker news thread where a guy called all this Brilliant Financial Engineering). How is this not unethical at least, it befuddles me.
Maybe we've come to celebrate unethical behavior and its become so normalized that we forget to ask ourselves what should be allowed.
Government hands Wall Street another bailout to the tune of trillions of dollars. Wall Street executives and hedge funds use funds to enrich themselves as usual. Main Street and tax payer get fisted again. These massive data centers go bust. Get gutted during bankruptcy and foreclosure proceedings Public deals with the fallout with no help from government.
Bubble bursts, somewhere between 2008 housing crisis and the dotcom bust.
Really dependent on if there are any OTHER structural problems to compound a fast re-valuation of tech stocks. There's plenty of noise about banks holding large amounts of bad private credit debt. There could be a lot or only a little collapse. There's so much uncertainty and the combination of war, high oil prices, and uncertainty about tarriffs that the market struggles to value anything as international fear drives investment into the US and high prices confusing whether growth is growth or just inflation.
Definitive peace in Iran combined with some sort of sobering AI news signaling the end to the infinite growth party could crush the markets.
The post-information age has never felt so well-named as it does lately. Investors dumping billions into completely unproven and, largely, undesired tech. Why? Because the Valley doesn't have anything else to sell, seemingly.
Either way, as always, we'll do it the American Way: Privatize the profits, socialize the losses.
>The post-information age has never felt so well-named as it does lately. Investors dumping billions into completely unproven and, largely, undesired tech. Why?
Eh. There's too much money. Covid response involved printing a lot of money and it all ended up somewhere. The chaos of the current administration has made everything considerably harder to price and the coincidental rise of the LLM has put us in strange situation that is legitimately difficult to price things correctly.
> There's plenty of noise about banks holding large amounts of bad private credit debt.
This is still only big enough to cause funny banking collapses not actual 2008 scale financial disasters. Banks hold a lot of bad debt, but it's isolated from consumer accounts. Might not want to hold equity in SoftBank though.
> There's so much uncertainty and the combination of war, high oil prices, and uncertainty about tarriffs that the market struggles to value anything as international fear drives investment into the US and high prices confusing whether growth is growth or just inflation.
The big concern lies in what the Trump admin will do. Things could end up merely a bad recession, like the Dotcom and Telecom bubble.
Or they can attempt to keep the bubble going once it collapses, crashing interest rates, and doom the US economy.
On other hand private corporate credit freezing might take down lot of business that need credit lines to operate regularly. Even the not so bad zombie companies. Tightening up and not being able to revolve credit anymore could lead to bankruptcies.
>This is still only big enough to cause funny banking collapses not actual 2008 scale financial disasters. Banks hold a lot of bad debt, but it's isolated from consumer accounts. Might not want to hold equity in SoftBank though.
Banks are lending to these private funds that are packaging questionable loans into securities (as opposed to banks giving loans or companies issuing bonds). This is the post-2008 place for people to get highly leveraged loans and they probably need to be better regulated.
But yes it doesn't seem like private credit alone will cause problems, the concern I'm trying to outline is a few of these things happening at the same time causing a kind of collapse.
TACO uncertainty is strangely propping up asset values as there's always a credible thought that whatever is happening is pretend or going to be reversed soon. And the expectation that the fed isn't independent any more and will make decisions to prolong the bubble resulting in a bigger crash ambiguously far into the future. Few want to start shorting because they have no concept of how long the market can stay irrational or if 20% inflation might be around the corner instead of a popped bubble.
> I'm not a skeptic of AI/LLMs but this makes me deeply suspicious of these circular deals. What happens when the music stops?
A financial crash that will make the 2007ff crisis look tame in comparison. That is why Anthropic, OpenAI and SpaceX (which xAI belongs to) are all going public soon and why NASDAQ bent the rules to include them... the current owners all want to raid pension savings worldwide [1] to get their payday before the bubble inevitably bursts.
And when it bursts, you can bet that the vultures will use their fresh cash to buy up assets at fire-sale prices. For the truly rich, a boom-bust cycle is only one thing, an opportunity to achieve extraordinary profit.
It's hard for me to see this being bigger than the great recession unless there's some vulnerabilities in the banking system we're not aware of. However, the amount of money that's being spent is going to demand a large return that I'm not sure will be made whole given the scale of investment in a time frame they want
> It's hard for me to see this being bigger than the great recession unless there's some vulnerabilities in the banking system we're not aware of.
The scenario I see is write-offs. At the moment there are hundreds of billions in IOUs being passed around, much more in liabilities than Lehman had back then in 2007. Compounding that is the frankly insane valuation - it's as clear as day that at least one of the major AI shops will go bust, they all run at a (huge) loss and sooner or later, one of them will run out of cash before achieving market dominance.
Unfortunately, OpenAI and Anthropic are valued at almost 1 trillion $ - backed by nothing but the hope on the winner surviving and achieving the classic VC-backed near-monopoly. The staff can be poached, they don't hold much in IP like patents, the servers and GPUs are mostly owned by third parties like AWS, Microsoft, Google or Oracle - once the cash runs out, they can't sell any assets for even some runway extension because there are no assets. Even the model weights and training data aren't worth much - all competitors already have training data sets of their own, it does not make sense to acquire further data, and model weights are being rendered obsolete by the constant churn of open-weight models particularly from China.
SpaceX is valued even higher, but unlike the other two candidates, they still at least got a viable business even if the entire AI BS bubble collapses, Starlink is a money printer and there's no alternative in sight that matches SpaceX and their reusable rockets.
Now, if either of the three even experiences a large drop in valuation for whatever reason, it's not just experienced VCs that can readily afford (and expect) investments to fail, but this time a lot of "everyday" investment vehicles (such as pension funds) will have to issue write-off losses, and now that they are publicly traded, that may also trigger stop-loss cascade orders further dropping prices, and retail investors will probably join in on the mass panic. That's the #1 risk IMHO.
The #2 risk is that after a collapse, the service providers (i.e. the ones owning the servers) will be sitting on a ton of hardware that has nowhere near recouped its cost. AWS, MS and Google can probably repurpose most of the hardware for their own use and rent out what remains, but they will have to eat significant accounting losses, provoking again a drop in their stock price, but this time with even more blast radius as all three of them are established stock index (and thus ETF) members that a looooot of people have exposure to. But someone like Oracle? They might actually get fried for good.
And the #3 risk is further downstream, particularly relating to NVDA. They have enjoyed years of insane profits because they are the only ones making high-performance AI chips. When demand for new chips collapses due to the event(s) I just described, they can easily shift their TSMC production slots back to GPU wafers and sell these to gamers - but at a far lower profit than before, which again can trigger stock price drops and write-offs.
I won't go further downstream - TSMC and their suppliers are IMHO pretty safe because there is just so much pent up demand from everything not AI, and the construction companies building datacenters don't have too much of a blast radius when the big guns stop expansion projects.
The concrete scenario I'm really, really afraid of: all three succeed with their IPOs, maybe they all survive a year and get included even in S&P 500. The existing shareholders and insiders all slowly dump a lot of their vested stock onto the public market, which in cleartext means into the dozens of billions of $ of retirement contributions. One day, the bubble bursts for whatever reason. The stock markets drop in a panic sell-off, either triggered by stop-loss orders or because retail investors are a herd of sheeple (just like in the 1st covid lockdown). Eventually, circuit breakers on the stock markets will trigger (just like they did in the GME post-apes collapse) and trading will pause, but it will resume until the markets have adjusted to the new valuation... and once the dust clears up, there will be a lot of blood on the floor. Possibly even riots, depending just how much retirement assets just got wiped out.
> A lot of people are emotionally unprepared for a world where the music doesn't stop.
I've been wrong before. However, when was the last time this business model made sense -- that facebook, SpaceX and others, all just pivot from their market niche to general purpose AI datacenter providers.
How on Earth does this make sense?
What happens in a few years when DeepSeek runs on the chinese chips like the Huawei Ascend at a fraction of the cost ?
These are all very high value added companies going into comodity AI hosting and they're all going to make a killing?
The only assumption I am making is that NVIDIA and others dig thier claws into western governments and make decade long contracts for even greater surveillance. Trillions of dollars worth.
> AND magically gets 4x cheaper energy they are in a much weaker stance for the long haul.
By energy you mean electricity,the nuclear, solar and hydro-- the kind China has installed new 2TW capacity over the last decade while US installed 0.2TW capacity in the same time?
The level of denialism when faced to confront hard realities of the world around us never ceases to surprise me. Alas AI capabilities continue to rip through expectations and the next goalposts are moved.
As far as business models go I think REIT for GPUs looks much more durable than Frontier Lab these days. The AI economy has a few big parts:
- Raw materials: Silicon, electricity
- Data centers: turn raw materials into compute
- Model vendors: turn compute into tokens
The frontier labs are competing in the idea that their tokens are worth more $/mtok than the others. If you look at the cost/quality Pareto curves, yes OpenAI and Anthropic are in the corner of expensive & good. But you need a log scale on price to look at these charts because the Chinese models are almost as good for a small fraction of the price. For this business model to be sustainable they need to keep innovating faster than everybody else AND for the quality difference to stay meaningful. Neither of those seem like sure things or frankly even likely to happen.
In contrast, further down the supply chain, folks supplying compute and raw materials both seem to be providing solid services that will be useful in the long term.
AFAIK the Colossus data center was more or less "brute force" when it comes to building a datacenter fast -- ie it was very expensive and they cut some pretty ugly corners (their generators are "temporary" to get around regulations, but I don't see how they can possibly be "temporary" and they create a massive amount of pollution compared to other power sources). The reason I point that out is the article mentions that SpaceX is "better" at building huge datacenters. I think this might be an overstatement -- they're just more willing to throw massive amounts of money and bend as many rules as possible.
>but I don't see how they can possibly be "temporary"
Fairly simple. You put 16 turbines that don't require permit for a year. After a year has passed you put another 16 pulled from another site and move the initial 16 to the former empty one.
Then you lobby hard to make sure that the authorities read temporary per turbine serial number and not total installed capacity.
Colossus is the world's largest single, unified GPU cluster, all GPUs acting as one coherent supercomputer rather than fragmented pools or multi-site setups. They spun it up in a fraction of the time by all estimates. It's not something you can just throw money at and reproduce the results.
Per Jensen Huang:
"As far as I know, there's only one person in the world who could do that; Elon is singular in his understanding of engineering and construction and large systems and marshaling resources; it's just unbelievable. A supercomputer that you would build would take normally three years to plan and then they deliver the equipment and it takes one year to get it all working."
..."it took 19 days to get Colossus from hardware installation to beginning training, the fastest by far anyone's been able to do that."
Regarding on site generators. Meta, OpenAI (Microsoft/Oracle) and others are also using on-site gas turbines, generators, and "behind-the-meter" power plants to keep up with the power demand. This has become an industry-wide strategy driven by grid constraints, with natural gas as a fast-deploy option.
It would be great if the grids could keep up with demand, if other options would be considered capable of producing the ongoing demands (ie. more renewable, nuclear, etc) but they're not, and companies are not going to just wait because then they're as good as done.
> Elon is singular in his understanding of engineering and construction and large systems and marshaling resources
Why would I believe a rich guy hyping his company's temporarily magical product when he hypes another guy who is a proven liar and flagrant fraudster? The cool thing is how the Twitter purchase was "on hold" due to bots and now it's mostly bots. But if you own the company making the software that powers the bots, I guess that's ok.
Jensen is smart enough to know he's glossing over the many shortcomings of an ultra-rich loser because it benefits him in the markets. I have no respect for that.
Colossus took an old manufacturing plant where they made rangetops and stoves and turned it into a functioning 100,000-GPU datacenter in 122 days. That's a truly insane turnaround time!
I’m not anti data center, but Colossus 1 is the posterchild for the anti data center crowd. The grievances with Colossus 1 might be a major cause why other data centers have such high opposition.
Polluting power generation, straining local power and broken promise after broken promise to fix the situation. And regulators caring more about helping xAI than mitigating the problems.
> In comparison, SpaceX/xAI are incredible at building datacentres on time. The original Colossus 1 datacentre was built in 122 days. Musk's empire does have a huge advantage in really understanding how to plan, build and execute enormous infrastructure projects quickly
Without even mentioning that it was done illegally and the air pollution they are creating with gas turbines is wildly irresponsible
I think the point is not difficulty but rather recklessness. I don't think other large players lack the ability or the deep pockets to do the same, but they might be lacking on the recklessness department for whatever reason. That reason, whatever it may be, might be the interesting part here.
No it's not. That statement assumes other corps care. They don't. If it was easy, everybody would be doing it. The fact that not everyone doing it is not because everyone else is not out of the goodness of their hearts
I laughed at brand risk. I think every player in AI is well past the brand risk. Only posturing you see is being too dangerous. And even that is for hype not actually stopping them. Eventually they push the models out despite that. It really is weird market branding wise.
Just goes to show how desperate they are for compute. It's definitely a brand risk for them, but failing to keep the treadmill going is an existential risk.
So we know what they are renting these GPUs for. I'm really curious about the input costs of their power generation. Is there actually enough margin in these deals for xAI to cover their depreciation cost?
Edit: from the footnotes:
> Colossus actually runs largely on its own on-site gas turbines, which comes out even cheaper: at a simple-cycle heat rate of ~10,000 Btu/kWh and Henry Hub gas at ~$3.50/MMBtu, the fuel bill is only around $90mn a year.
OK, that's crazy. How can I get into renting GPUs to hyperscalers?
As an investor it's really important for me to see this kind of bearish sentiment widely. It's actually a positive for the stock in my opinion, not a negative.
Makes sense. Very difficult to catch OpenAI and Anthropic now since their flywheel of generate revenue, use revenue to buy more compute, train a smarter model with more compute, made it hard to compete.
Being able to supply compute makes more sense for SpaceXAI if you can't compete in SOTA LLMs anymore.
I have the pro account for ChatGPT, Claude, Gemini, and Grok.
They all have various strengths and weaknesses. My favorite is still ChatGPT, then Gemini/Claude, then Grok.
Grok often feels 1-2 generations behind the competition in general use, but it has three things that I love:
1. It seems to be the best at understanding current events. Maybe due to X integration, or some other tool call optimization in the backend? I don't know, but I often ask about things going on, and the other models have outdated info, give unhelpful answers, etc.
2. It is generally the least sycophantic for personal things. Anthropic is getting here too. ChatGPT and Gemini are working on this, but previous models in those families would almost never say anything negative about what I am doing. Sometimes I need career advice, personal advice, etc and I like the tone of how it responds. I think Claude will be caught up soon.
3. For professional work, there are certain topics that other models would refuse to engage with. At my last company we had an enormous amount of legal users. When a deposition would need a summary on certain topics, most models would refuse. Grok would not. I understand the need for safety and I don't blame the other model providers, but for some professional use cases you NEED a model that is capable of handling sensitive subjects.
I recently worked with NRC dataset, specifically about nuclear reactor events and status reports(example: https://www.nrc.gov/reading-rm/doc-collections/event-status/...). Public data that just needed some cleaning. Several time Claude API would refuse to engage. Because of that I can't trust Claude to clean production data sets.
> 1. It seems to be the best at understanding current events. Maybe due to X integration, or some other tool call optimization in the backend? I don't know, but I often ask about things going on, and the other models have outdated info, give unhelpful answers, etc.
That makes sense, but occasionally you ask about an issue where it's clearly received political instruction from the commissar and it acts totally lobotomized. But it's true that Gemini will often blithely state that something could never happen and you'll say "what do you mean, that just happened" and then it comes back apologizing after running a Web search.
We saw this too with Gemini specifically. My favorite example - we built a hallucination detector (given the input, does the output make any false claims) in Gemini, and after the Seahawks won the Superbowl in February, it would consistently flag that as "not possible".
It was mind blowing the first time I got a refusal, and retorted "yes you can" and had that work, but now it's just another reason to move to a different model.
Almost too much so, it often feels like opus is pushing back for the sake of pushing back. The way old models used to add disclaimers to every message regardless of content
People are weird about their cars and make major errors in judgement as a result (e.g. we tolerate incredibly high rates of people getting killed because they were "hit by a car", as though the driver had nothing to do with it). Pushing back on that is absolutely worthwhile.
All 4 of these still regularly insist that I am a genius and everything I say is brilliant. Grok definitely pushes back more than the others, but I don't like how sycophantic they all still are.
I don’t want to open up that whole can of worms but Grok on any vaguely philosophical or political topic is a scaredy cat and has a very hard time staying factual if it could make Musk or the conservative movement appear negatively.
I almost exclusively use claude for all my professional and private needs. In my experience it's really good at adhering to my wishes in regards to sycophancy and pushing back. If you really want to you can tell it to systematically push back on anything where pushback makes sense until it continues with the flow of conversation.
In my first therapy session, the answers were too long and contained multiple questions, spawning multiple threads of conversation. I told it to tone it down and only ever ask one question back, maybe two, if they are related. The answers got too short. I told it to make them "slightly longer" again and reached a sweet spot.
The conversation is yours to form! You need to find the "system prompts" and guidelines to give it that work for you.
My favorite was ChatGPT, and I still use it often, but it becomes way too 'hair splitting' argumentative too often over very minor non controversial topics. Like it's always going out of its way to "well actually..."
Grok used to be really really bad ~8 months ago or so, but it's gotten better.
ChatGPT team needs to turn down the 'disagree just because' factor by a lot.
1. It seeks to manipulate the information you see and your lens to the world. This is already partially true from independent and major publications.
As soon as we hand over searching out information to social media algorithms and LLM tools, we abandon our ability to see reality outside our direct vision.
Grok's ownership has already demonstrated capacity to influence major world elections and other events. You cannot trust it with this sort of information gathering and reporting.
I guess the benchmarks disagree, but whenever I need to find specific information that does not easily show up with a web search, I try chatgpt, gemini and grok. Grok surfaces what I was looking for more often than the others.
Things like "find the github repo from 2017 that does $vague_thing".
Good question. You can actually see the searches it runs (momentarily) so testing could determine if it's using public search engines or a private system.
Eh. It was a leading model for a few weeks, it was a real effort, but they never built a real revenue model around it. It wasn't SaaS, it wasn't for governments, it couldn't get B2C payments. Made it hard to justify the training cost to stay at the frontier.
And they are planning (well "planning" if you believe Elon) to start building their LLM over from scratch, which means they need a HUGE ass training data center, i.e. not a data center for inference to do so.
It's a general problem of defining yourself in negative terms. Being "un-{thing I don't like}" doesn't say what you are. It only excludes one possibility while leaving behind an infinitude of mostly crappy alternatives to try to choose from.
Having a positive set of beliefs annoys people and and can make them feel judged, but at least it provides a vector that points somewhere definite in possibility space.
I don't think this is good news for the AI industry.
If they can't build enough capacity where their best option (and they're signing multi-BILLION dollar contracts) is an unproven 'datacenter in space' technology, we are toast.
- near term costs will go up (demand is greater than supply)
- tokens shift from all you can eat (TOKENMAXXX) to ROI-driven
- engineers with real orchestration skills rule and shift to lower cost optimization (deep seek)
- Frontier AI unit economics collapse
- roughly 1B revenue per month is a good look for SpaceX
- provides some credibility to otherwise non-existent xAi business.
- pos. News for Google due to their share in SpaceX.
- can be interpreted as leasing versus building for Google, a nice Hedge on compute capacity.
- saves Capex for Google at time of horrendous GPU, memory costs.
It is only a bridge until 2029. What will be the value of the SpaceX data centers by then? Fine print matters on deal with Google. MS made billions up front payments to Coreweave. SpaceX has no upfront payments, has to stem Capex alone. Very favourable for Anthropic and Google.
I suspect that this is the start of a play for SpaceX's orbital datacenter project - if they're really planning on launching as many satellites as they've said (and Starship is going to massively lower the cost of launch), they won't be able to fill them with Grok. So perhaps it's best to become the infrastructure provider to the other AI Labs.
Is there anything to read on how the economics of an orbital datacenter make any sense? Because I don't really see how blasting a server into space solves any of the typical issues associated with datacentres beyond easier access to solar.
> anything to read on how the economics of an orbital datacenter make any sense?
I'll do a write-up at some point. But the core drivers are launch cost, permitting delays for terrestrial datacentres and interest rates.
The balance is between, on one hand, the financing cost of the permiting delays against, on the other hand, the cost of launching radiators. (Chips are light. Solar panels without glass cladding are surprisingly light, too. The weight of an orbital datacenter is almost entirely in its radiator.)
The math high-level works with Starship (6 flights/year), 3+ year financing delays and a 10 kg/kW radeiator (assuming 6% financing cost). Of course, there are devils upon devils in the details. But directionally, we're seeing pushback against terrestrial datacenters. And from what I can tell, advanced heat pipes may be the unlock to get radiators down to 5 to 6 kg/kW, at which point I think even New Glenn's $300/kg projected prices become competitive.
It all goes out the window if launch costs don't come down, interest rates go above 10%, terrestrial datacenters start getting built quicker, or demand for this category of compute collapses.
"Well, you didn't want a data centre in the field near your town, so instead we'll rain astrocentere debris across the western hemisphere and set off a Kessler syndrome cascade. Thank god we didn't have to wait for a permit."
I get the point, but if society cared about globally distributed pollution more than about money, we've have transitioned to renewables and EVs a decade or three earlier.
A lot of bribes have been spent to buy that delay you know, the first global meetings about addressing climate change happened in 1992 and the petrostates have been stalling since then.
I started doing some numbers around the scale of tokens per second we can generate with figures like 300 million watts and I really don't understand the destination anymore. I see that Anthropic is somehow constrained in the news, but that doesn't line up with headlines. Everything seems off by 3-4 orders of magnitude here. I realize there are some users of AI who can burn a million tokens like it's nothing, but these facilities can produce trillions (10^12) per day.
I feel like there is some use case planned here that isn't to be known about until it's way too late to do something about it. Or this is a very serious bubble. One of the two or some really horrible blend.
I haven't done any math, but if you have individual users that can burn millions of tokens a day, then it does not take very many of them (at SaaS scale even limiting to power users) to hit trillions. And even fewer to run into problems specifically with time of use.
Is this HN user runako’s comment[1] from 2 days ago turned into an article?
I guess it’s very possible multiple people are coming up with the same idea at the same time but given this was submitted by the author it seems kinda rude not to mention it.
Sure, but it's very different from a regular datacenter REIT. It's supply and demand: a regular datacenter REIT competes with tons of other datacenter REITs. xAI has something to offer that basically nobody else can: datacenter capacity ready to use for LLM inference. When you are a monopoly you can charge exorbitant prices. There's no shame in being "just a datacenter REIT" when you can charge 10x what it costs you to run that datacenter.
I’m so confused by conflicting headlines and studies. One set says gpus sit idle most of the time, then there are apparent capacity problems with Anthropic. So what’s the deal?
What are those two groups selling? Blowhards like Ed Z. are selling their opinions so they need people to believe them. Anthropic et Al are selling inference for today dollars, and need compute to run that on. Your guess as to who's actually right, and who's selling smoke and mirrors though.
It’s a vertical company they did compute very good , their top model bounce between top tier and -1 -2 gen. They were top tier only once though on paper briefly . If tomorrow thy will hit top tier , that do have know how to expand . They can even buy back from Google or anthropic if they agree.
Weren't we just talking about how SpaceX is valued based on some profits from starlink + tons of speculation?
Yet when we learn of this new $26B in yearly revenue (2.2B/month from Google and Anthropic)the conversation does not return to that discussion. It transforms into:
"xAI's tech sucks"
"Google/SpaceX is Structurally Bad for the Economy"
etc
This is called motivated reasoning. We get new information and instead of the obvious thing, updating prior conclusions, we just find a different way to react negatively. The negative reaction will be achieved. The narrative here is completely polluted by people who dislike Elon/SpaceX.
Think two things can be true at once. They should be using their capital to achieve their speculative price. Instead, they are using their capital to achieve a modest ROI, thus invalidating the speculation AND proving they have tech issues in what the speculation is around.
Elon says Grok models are being trained right now. (Unless I missed an update.) For whatever reason these training runs are not using xAI's full GPU capacity. Short of a miracle or time machine it sounds like there is nothing more they can do to advance their mission.
Where did you get 2-3B from?
Colosus 2 GPU's alone were 18B
Total cost including construction, power and water treatment facility might be close to 25-30B.
I'm not OP but that was the cost of the initial facility if I remember correctly when it was first up and running, what you're describing I believe is the full cost after all expansions/etc
There is a shortage, they are short lived assets. It's a blip and unrelated to their long term profitability and valuation. They can't make a long lived business of building and renting out compute at those margins.
It was definitely a smart business move. It should be troubling to any shareholder than xAI is unable to utilize this infrastructure as renting it out to competitors.
T1 companies have longer depreciation cycles, they have customers that will use the dated hw for non-frontier work. They can make the capex more justifiable and have flexibility to be more creative about its use. A frontier lab really needs the best hw available at full capacity.
Respectfully, I tend to think of tier 1 data centers as someone I'm paying for colocation services and the value they provide is power infrastructure and redundancy, network infrastructure and redundancy, cooling, and physical security.
The shortage I referred to is in GPUs, that's what really being rented here.
Even if GPUs lasted forever, they're are a depreciating asset because they become obsolete with improvements over generations.
GPUs do not last forever, either. I've read here, and heard from others, that they aren't even living up to their 5 year depreciation schedules under production load, closer to 2-3 years.
I use AI all the time. I hope AI isn't short lived. It might be if they can't figure this shit out, or if IPOs like spacex poison public opinion against them first.
> GPUs do not last forever, either. I've read here, and heard from others, that they aren't even living up to their 5 year depreciation schedules under production load, closer to 2-3 years
People said this about GPUs during the crypto mining craze and were wrong back then too. While I can’t speak for the entire industry I can say my personal experience follows any normal intuition over solid state electronics.
Some early failures in the bathtub curve, and then you start seeing fans, heat paste, and board capacitors fail far before you start seeing any chip failures at scale.
Sure you can abuse anything you want to burn it out, but I doubt that’s what’s happening inside these facilities.
It's right in the article, there were 40bn of disclosed costs. It's still a good return, it pays for itself in 18 months, but if you build and rent data centres, then that's your business, and you're not likely to 100x in 3 years, which is the wild projection behind their valuation.
Also moves spend from capex to opex for your competitors - their access to your GPUs so don't have to wait to buy so many of their own, and I'm going to take a stab that those puppies are going to depreciate hard.
But better to make some money with it while trying to catch up than none money hoping you _can_ catch up.
I think the point is, that although at least xAI is monetizing their GPUs/datacenters, they are doing so at a REIT/rental multiplier instead of a frontier lab multiplier.
Clearly, xAI thinks this is the best way for them to extract value out of their assets.
Also, it is clear that Google and Anthropic both think they can extract more value out of those assets than they will pay in rent to SpaceX.
I got approached by a recruiter to directly train Grok on coding, so it seems like they're still trying to build a model that's better at coding than shitposting at least?
I just don't like the clever corporate shell games Musk has been playing with his related party businesses to juice revenues and keep share prices up - e.g., SpaceX buying $131 million worth of Cybertrucks from Tesla, or SpaceX buying xAI.
It's clever business perhaps, but it's terrible governance.
But then Musk will always control the majority of voting rights in SpaceX, so not like the shareholders are able to vote to remove him from the board. Being fair, it's the same share structure Zuckerberg uses to retain control over Meta, in case I give the impression that I think only Musk is doing this.
Which is why I'd never buy shares in either of them, the directors are supposed to act in the best interests of all shareholders, and well, if you can't vote on director appointments, you can't do anything when they decide to act in the best interests of a few shareholders.
Well, perhaps, but those concerns seem different enough that it seems fairly plausible different people have them. It seems hard to argue the basic point that Grok is not as good as its competition if you spend time using both. That may or may not matter from a business perspective.
> he narrative here is completely polluted by people who dislike Elon/SpaceX.
Hard disagree. It's polluted by Elon in general (pro and con), just like Tesla's idiotic valuation.
But in this case, a pivoted business model fundamentally changes the value proposition, and I'm not clear why "this space company making money on space things is now pretending to be a compute reseller and that's a good thing" is the narrative you think is preferable.
It's also beyond lame to essentially subtweet a "narrative" instead of responding to it directly. Who is "we", aside from a transparently dishonest way to pretend consensus exists?
Technology has a very short life. The difference is that a REIT might contain an office buildings that can be used for any business, but a data center is filled with carcasses that start rotting and stinking from the day of installation.
The idea that the AI data centers would depreciate in just a few years is plain wrong. The argument was that new chips would be so much more powerful and efficient that it would be cheaper to buy and operate the new chips than to just operate the old chips. Except that demand is significantly outpacing new chip manufacturing, and until it catches up years and years from now, the efficiency argument doesn't matter at all.
I was wondering, with the "10k" price tag quoted here for some earlier (!) units, how much of that is R&D + cost, and how much is just gouging the companies?
No, that's silly. Chips don't rot like produce. Some components will go bad and will need to be replaced. The owner can choose how fast to replace them depending on how prices look in a few years. The rest of the building (including things like power) will still be useful.
i think what op meant is they're instantly out of date. You're not going to be able replace every GPU in your datacenter on every new release from Nvidia and customers are going to go to whoever has the highest performing gear.
> customers are going to go to whoever has the highest performing gear
This is where we lack data. I’m skeptical of the claim. If anything will force retirement of old chips, it will be power efficiency, not customers being picky amidst a chip shortage.
Demand is so high and supply so low customers will go to anyone that has any gear, period. Anthropic is paying xAI for GPUs from 2022, not the latest Nvidia release.
what if superintelligence is a myth? There is no mathematical reason why it should exist. The exploitablity of the environment is finite. Humans seek utility not intelligence as it own end. It is entirely possible we have overinvested in compute to the point of cargo cult.
Elon is controversial, and I know the popular mood is to dunk on this IPO, but in the end he gets a lot right. I think it's very likely the Tesla deal happens, and I further thing that regardless what Google and Anthropic do long-term — SpaceX/Tesla will likely have an awful lot of Optimus robots, and those robots will need an awful lot of compute. I'd be very surprised if Elon lets some frontier lab be the intelligence layer of his robotics stack when he has all the raw materials himself. And in fact, rather than introduce model COGS, he will have had the frontier labs paid for all the CapEx he'll need to run robotic inference using his own models.
There is a lawsuit against xAI about those datacenters. They have a strong case that Musk is clearly flouting environmental laws. If they are able to get a preliminary injunction against the datacenters, then they are dead in the water. That's the only reason why they build them so quickly.
And yet Anthropic is paying xAI over a billion dollars a month for those out of date GPUs in their first datacentre (H100s being nearly 4 years old at this point).
Even A100s are still barely available on the major clouds despite being 6 years old.
Yeah most of the performance increases have mostly been from architectural improvements like reduced precision tensor cores. AFAIK FP4 is basically the limit for floating point matmuls, after which you need to switch to integer addition if you want to reduce bits, and I don’t think we’ve figured out 1-bit LLMs just yet.
It makes sense. They've long since fallen behind the big 3 in quality of their models. There's no good reason at this point to keep burning money on Grok rather than making back some of that money renting out their Colossus data center.
Not that my feelings or opinion matter, but I'm going to be devastated if this xAI grift takes down SpaceX. I am reminded that the 90s Boy Band groups Backstreet Boys and N*SYNC were the side projects of a Ponzi Scheme.
> While this doesn't include opex[2] and depreciation, if the deals continue for 18 months, xAI recoups all the capex they spent and still has many hundreds of MW of GPUs available. With the giant compute shortages likely to persist into the medium term, even older H100s are likely to be extremely useful even 18 months out.
xAI is more than half of SpaceX revenue with the Google sublease. SpaceX is looking like a datacenter REIT.
Moreover they're leasing compute - the actual infra around it is much less important - and how long does anyone expect heavily utilized GPUs to run? How likely is SpaceX to be able to re-lease this compute capacity? It will be broken down or out of date in 2-3 years.
This should be essentially ignored in the long term for SpaceX business prospects, and is low margin business that barely justifies a 10x earnings multiple let along a 100 revenue multiple for the xAI unit.
Except there is no gold. Or arguably that "gold" is prohibitively expensive to use (not just store, it deprecates fast) so the entire rush is being subsidized. Let's go! /s
If xAI is a datacenter REIT, it is a special kind that has a promise that no other datacenter provider could dream of: LEO datacenters. As far-fetched as that may sound, the biggest profit center for SpaceX in my understanding was Starlink. xAI already has extremely high-bandwidth connections from Earth to LEO available. Connecting that to solar powered orbital datacenters seems doable in realistic timeframes, especially once Starship comes online and gives them a significant boost in launch capacity.
If that ends up being viable and profitable, there is no realistic competition for decades. In this view, xAI earning a reputation as a reliable AI hyperscaler is just another tactic in that strategy.
Cost per pound? Then $/Watts/TFLOPS minus cost per pound?
Out of curiosity (since I basically never saw $/lb mentioned in any replies anywhere on this, which is hilarious; like talking about having your grain mill at 10,000ft/in the mountains since sunlight is better there): Have you ever tried a forSpace Program?
(And not only 100mi or more above, they're 17,500 mph faster--Mach 22 Datacenters in an oxygen-free, higher-radiation, insulated environment with absolutely no resources)
I have not done any calculation on the capex but I'm guessing that SpaceX has. We're also just guessing on what these "datacenters" will look like. It is silly to think of them in the same way we see the football-field-sized ultra secure facilities on land. They can be highly distributed in a way that just wouldn't make sense on land. Perhaps even incrementally built out in the same way that Starlink capacity was.
Regardless, if Google is spending just shy of 1 billion USD per month, that suggests that there is a pretty high ceiling on capex available.
Consider the PR of massive datacenters here on Earth. People complain about noise, water usage. It doesn't even matter if those concerns are valid, the PR is bad enough. That might attract other massive corps that want to outsource instead of deal with the headache of building local.
You realize that not long ago companies were exploring building nuclear power sites next to their data centers to handle the expected power needs?
I'm not saying it will work. I'm saying if it does, SpaceX will own the market for a good while.
Sure, but considering the size of the challenge it makes sense to figure out the parts that can be studied on the surface first. Challenges like procuring, challenges like setting up relationships with potential customers. You probably want to figure out everything you can so that when you move on to the hard part you aren't distracted by the rest.
Consider the alternative. SpaceX figures out how to build the datacenter in space thing but fails at the rest. That would be an expensive mistake.
Elon is brilliant when it comes to hardware. But unfortunately with xAI, he went on a firing spree, PayPal Mafia style, just like with Twitter when he bought it, shortly before doing another hiring drive, and failing to hire software engineers at scale.
The datacenter deals came after. But now, the man who promised the world an AI system that defends free speech and is “pro-human”, is instead selling to his competitors and lowering the daily app usage limits of his own Grok by an order of magnitude (really).
If you’re dealing with the world’s richest man, you can predict that money will come before other concerns despite other rhetoric. Interesting strategy though!
Edit: To be fair, they did decide that hardware was "the bottleneck" according to an interview I saw last year. But I firmly believe they underestimated the software problem (and their app was/is riddled with them).
