It has to get bigger. As soon as it starts shrinking, the next round of debt will no longer be able to cover the prior round of commitments. What is happening in AI is essentially a gigantic version of what is happening in consumer auto loans, they just keep refinancing for more and more money. Eventually there will be no one willing to lend them more, and then they'll go to the government to bail them out.
Yeah, I'm pretty sure we've already passed some sort of fiscal singularity where economic and political interests are so intertwined that there can only be growth. If there ever isn't growth, then the legislators (who are elected via corporate sponsorship) will pull whatever levers necessary to make sure that there will always be asset growth.
It's some sort of tragic positive feedback loop that isn't going to stop until the whole thing comes crashing down for everyone and we're paying $37,000 for a loaf of bread.
There's one problem with the idea of a bailout. What comes after?
Say, for the sake of argument, that the US government bails out OpenAI. Are they now suddenly not a money-burning operation anymore that will need another bail out in a period better measured in weeks?
The right analogy here is not the auto industry, but the music industry. Regulation might "win", but margins will be driven down to commodity levels. That is not the assumption that current US AI company valuations are based on.
A more useful comparison would be the share of new emissions that data centers represent, and a comparison of the growth rates of various sources of marginal new emissions (eg, from cement, or ammonia). Besides their visibility, the reason that people focus on AI data centers is that they are a new source, on top of whatever organic growth there is in existing industries. And marginal growth in CO2 emissions is important because climate models bake in certain assumptions about growth. If a wildcard source suddenly appears with exponential growth, the projections on which public policy is based can be affected non-linearly. This article uses data only up to 2024; 18 months of exponential growth can change the picture quite a bit.
>Austin is surrounded by empty flat land. San Francisco is surrounded by water and was built out many decades ago.
This^^^. The idea that housing starts are solely (or even substantially) impacted by regulation is a case of control bias, attributing disproportionate influence to minor factors we can control (like regulation) over major factors we cannot (land, labor and material costs).
I've read this like 5 times and I still can't make sense of it. Is this a variant of the "5-bedroom houses in Cairo Illinois only cost $40,000, there is no housing shortage" argument?
No---did you read the original article? The comment and the parent will make more sense. Basically the article compares Austin and San Fransisco and attributes the difference in building rates entirely to disparities in regulation. Which is obviously not an accurate framing, for the reasons mentioned in our respective comments.
Your comment is more addressed at affordability, which is largely a function of wealth disparity: as the wealthy get richer with respect to the lower percentiles, they absorb an increasing share of the home market and drive prices up.
I did; the top comment on this thread is my response to the article. I don't think you fully get the argument: San Francisco needs to upzone and do large amounts of infill development. Most of the most important areas of economic opportunity in this country are landlocked; San Francisco's challenge isn't unique.
I think you're making the same mistake the article did, focusing on the regulatory environment. My argument is that the zoning and other regulation is not the main reason Austin has more housing starts than San Francisco. The main reason is that land, labor, and materials are cheaper there. SF can upzone and infill to it's heart's content, it won't make land, labor and materials cheaper than Austin.
Your argument is conclusory. The whole question is the extent to which the regulatory environment is preventing upzoning. If you think San Francisco should remain zoned the way it is, say that, and say why.
I have no opinion one way or another how SF should be zoned. I'm just saying that how it is zoned is a second- or third-order effect on the rate of home-building. Land, labor and materials are the first-order effects. Now, maybe SF is saturated, and every square foot that could have a home on it already has one. Then you'd have to rezone other areas in order to increase the supply of land. But even if you did that, as long as the land was a lot more expensive than Austin's, the rate of home construction would still be slower. And I think it would be physically impossible for SF to zone its way to a land supply comparable to Austin. Hope that clarifies.
It sounds like Nvidia is not only supplying GPUs first to neoclouds, it is also supplying them for free if they cannot be resold:
"Furthermore, in the case of CoreWeave, Nvidia has also provided a significant financial backstop against unsold GPU capacity. Under the agreement with an initial value of $6.3 billion, “in instances where [CoreWeave’s] datacenter capacity is not fully utilized by its own customers, NVIDIA is obligated to purchase the residual unsold capacity through April 13, 2032.” In other words, Nvidia is committed to purchasing unsold GPU capacity if CoreWeave is unable to find another buyer. With an initial value of $6.3 billion, there is the potential that the arrangement could become larger over time."
I don't know how Nvidia is handling Coreweave GPU sales revenue in their accounting, but it sounds to me like it should have a pretty big asterisk attached to it. It's more like a consignment arrangement than an actual sale. And it obviously creates a huge incentive for Coreweave to over-order GPUs, since there's no risk (I doubt they're paying cash up front).
From an accounting perspective, this absolutely isn't a consignment agreement.
The sale of the GPUs by Nvidia to CoreWeave is real. CoreWeave pays Nvidia cash and becomes the owner of the asset, so it's properly booked as a sale. If it can't sell capacity, the GPUs are not returned to Nvidia.
CoreWeave is using debt to make the purchases but the backstop provided by Nvidia ostensibly helps it get better loan terms. That doesn't change the accounting.
If Nvidia has to purchase unused capacity, it simply becomes an operating expense for Nvidia.
Nvidia's exposure is the $6.3 billion backstop obligation and the equity it holds in CoreWeave.
>CoreWeave is using debt to make the purchases but the backstop provided by Nvidia ostensibly helps it get better loan terms.
According to the article, the $6.3B is a floor, not a ceiling. And it's not clear whether CoreWeave is actually paying cash or getting the GPUs on credit. If the full amount is getting booked, it's an accounting loophole that's being exploited. If GM sells Hertz a million cars, but says "Hey, we'll buy these back if you can't rent them," can GM book all those cars as actual revenue? What if Hertz only has to pay 10% up front and the rest in 5 years?
Your GM/Hertz comparison is not applicable here. Under GAAP accounting rules, GM wouldn't be able to book those as sales because it was obligated (or likely) to buy back the asset. Under the rules, this means the transaction gets treated as an operating lease. The cars would stay on GM's balance sheet and the revenue would get recognized over the lease period.
The CoreWeave-Nvidia deal is not the same because Nvidia is not buying back the asset (the GPUs). CoreWeave has title to the chips and if they're worth nothing in 5 years, that's a problem for CoreWeave and its lenders.
What Nvidia obligated itself to was buying compute capacity, which Nvidia would be able to use for its own workloads.
In the GM/Hertz analogy, this is like GM selling Hertz the cars and saying "If you can't find renters for them, we'll rent them from you at market rates, up to $x." Under GAAP accounting rules, GM would book the car sales as revenue, the commitment to rent would be a purchase obligation, and if the rentals ever occurred, GM would incur the costs as an operating expense.
There is a question of whether the CoreWeave-Nvidia deal structure is sensible economically, and how much risk is being created. But there's no GAAP accounting question here. At all.
Good explanation. But whether it's GAAP compliant or not, the arrangement incentivizes Coreweave to buy chips it doesn't need. You're assuming that Nvidia will have some business need for the excess capacity, but there's absolutely no assurance that that is the case---indeed, Nvidia is incentivized by the AI market dynamics to show revenue growth at all costs, because there are plenty of bulls who will wave away any potential future obligations as "ordinary business costs". But are they really ordinary, or is this potential obligation to buy compute actually much greater than Nvidia's actual future needs?
> ...the arrangement incentivizes Coreweave to buy chips it doesn't need
You state this as fact but this is just cynical speculation on your part.
The less cynically speculative analysis is that Neoclouds like CoreWeave are rushing to build datacenters because their whole business is based on the premise that AI is a revolutionary technology and there will be massive durable demand for AI compute for the forseeable future.
CoreWeave generated over $2 billion in revenue in Q1 and has a nearly $100 billion contracted revenue backlog. This is not an imaginary business with no demand.
Nvidia has invested a very modest amount of money in CoreWeave equity. Dividing its revenue by the number of days in a year, Nvidia generates about $2 billion in revenue in ~3 days, and $2 billion represents 0.04% of Nvidia's market capitalization.
Are there risks here? Yes. Is the circularity potentially problematic? Yes. But is it also true that some of these arrangements are being used to make hyperbolically apocalyptic claims? Yes.
> You're assuming that Nvidia will have some business need for the excess capacity, but there's absolutely no assurance that that is the case...
You're absolutely correct here, which is a source of risk for Nvidia. That doesn't change the accounting as far as GAAP is concerned though, and you aren't looking at the big picture.
The $6.3 billion backstop through 2032 is not a huge burden at all for Nvidia. Nvidia will generate about $190 billion in free cash flow this year alone.
>> ...the arrangement incentivizes Coreweave to buy chips it doesn't need
>You state this as fact but this is just cynical speculation on your part.
Whether CoreWeave actually bought chips it doesn't need is speculative; whether they are incentivized to do so is not. That is clearly the case: if you are guaranteed that any excess capacity will be bought, a rational actor will buy more than they need, as there is no risk for over-buying, but there is risk in being caught short. That isn't cynical, it's simple econ 101.
>The $6.3 billion backstop through 2032 is not a huge burden at all for Nvidia.
As noted earlier, the $6.3B is a floor, not a ceiling.
>Nvidia will generate about $190 billion in free cash flow this year alone.
Maybe. But the whole point of this discussion is trying to answer the question, "How much of Nvidia's revenue is real?"
> Whether CoreWeave actually bought chips it doesn't need is speculative; whether they are incentivized to do so is not. That is clearly the case: if you are guaranteed that any excess capacity will be bought, a rational actor will buy more than they need, as there is no risk for over-buying, but there is risk in being caught short. That isn't cynical, it's simple econ 101.
There's a problem here: you haven't actually quantified how much CoreWeave is spending versus the value of the backstop. You seem to be suggesting that for every dollar CoreWeave spends on Nvidia chips, it's getting a dollar in backstop. But that's not how it works.
CoreWeave buys chips from Nvidia; Nvidia has agreed to buy up to $6.3 billion in unused compute capacity through 2032. It's not buying back the chips, etc.
CoreWeave has raised way more debt (over $35 billion) to build out compute than what Nvidia has backstopped (up to $6.3 billion). In other words, CoreWeave is spending a ton to buy chips, build datacenters, buy electricity, etc. and Nvidia's backstop, while important, doesn't come close to backstopping all the investment CoreWeave is making to acquire its compute capacity.
If demand for compute dries up, CoreWeave and its lenders are going to be on the hook for way, way more than Nvidia is.
> As noted earlier, the $6.3B is a floor, not a ceiling.
You keep repeating this but it's factually incorrect. The $6.3 billion is the maximum.
As I've noted, $6.3 billion works out to a few days' revenue for Nvidia.
> Maybe. But the whole point of this discussion is trying to answer the question, "How much of Nvidia's revenue is real?"
This is such a strange question.
Nvidia reported revenue of ~$215 billion for FY 2026, and ~$96 billion in free cash flow.
I don't know how to put it more simply: this is real money. And gobs of it. It's not made up.
The question you seem to really be asking is: is the demand for chips that is driving this revenue sustainable, or will it collapse, leading to a massive rapid drop in revenue? That's a completely different question but just FYI: Nvidia reported $81.6 billion in Q1 FY 2027 revenue so...
>There's a problem here: you haven't actually quantified how much CoreWeave is spending versus the value of the backstop. You seem to be suggesting that for every dollar CoreWeave spends on Nvidia chips, it's getting a dollar in backstop.
I'm suggesting nothing of the kind. Nvidia is essentially guaranteeing CoreWeaves return on the chip investment. If someone offered you that deal, regardless of what the investment was, you'd buy as much as you possibly could.
>If demand for compute dries up, CoreWeave and its lenders are going to be on the hook for way, way more than Nvidia is.
True, but irrelevant. The return on investment incentive is independent of how much other debt you have, you'd still buy every unit you could.
>>But the whole point of this discussion is trying to answer the question, "How much of Nvidia's revenue is real?"
>This is such a strange question.
Yet it's the actual question being discussed in the article, and not whether the arrangements are GAAP-compliant. Saying the revenue is real, and that there are "gobs" of it, doesn't make it so. For the record, I'm sure much of it is real. But I'm equally sure that much of the demand is artificial, driven by the business practices we're talking about. I think the root of the problem is that you sincerely believe that if revenue is accounted for in a GAAP-compliant way, it must be real, organic growth driven by real demand. By that standard, you can say that Beanie Baby demand was "real". But it wasn't tied into any underlying economic utility, it was a speculative bubble. So is the AI infrastructure market, just embedded in a much more complex system of deals, as we've been discussing.
>The question you seem to really be asking is: is the demand for chips that is driving this revenue sustainable, or will it collapse, leading to a massive rapid drop in revenue?
Yes, this also is the question really being asked in the article. Among the factors to take into account when judging sustainability are whether the sales are for cash or credit, and whether those sales are being incentivized extrinsically (like, say, guaranteeing ROI).
> Nvidia is essentially guaranteeing CoreWeaves return on the chip investment.
You cannot say that Nvidia is guaranteeing a return on investment for CoreWeave without showing the actual math for a positive ROI.
CoreWeave has over $35 billion in debt, most of which has been used to build out datacenters. Nvidia's backstop is capped at $6.3 billion, and only through 2032.
You need a new theory here.
> Saying the revenue is real, and that there are "gobs" of it, doesn't make it so.
I'm really sorry, but with all due respect, this is getting into sillyland. Q1 2027 revenue of $81.6 billion is not a made up number. The $58.2 billion in net earnings is not a made up number. The $20 billion Nvidia returned to shareholders via share repurchases and dividends is not a made up number. The $80 billion the company added to its share repurchase program is not a made up number. The $50+ billion in cash and short-term investments on the balance sheet is not a made up number.
If every dollar in was being used to drive a dollar of new demand, Nvidia's financials could not look like this.
Now is there circularity in the AI market? Yes. Are there legitimate reasons to pay attention to it and be concerned? I think "yes" is a reasonable answer. But you cannot claim this is all "fake" and expect people to take you seriously because none of the financials supports it.
You keep saying this. That is not what the article says. It says the backstop is currently valued at $6.3B. That is not the cap.
>Q1 2027 revenue of $81.6 billion is not a made up number.
No one is saying it's a made up number. If you want to argue, at least read the comment and address the actual argument being made, and not a straw man.
>If every dollar in was being used to drive a dollar of new demand, Nvidia's financials could not look like this.
No one is saying Nvidia was spending a dollar for every dollar. It doesn't have to in order for the demand to be manufactured.
>But you cannot claim this is all "fake"...
Nowhere in this thread did I say it was all fake. Quite the opposite, every bubble has to have a core of reality to be sustainable.
You keep insisting that any revenue booked and reported must be real. So answer me this: if I take a GPU order from an insolvent individual for $100M, can I book that $100M as revenue and be GAAP-compliant? Is it real? What if I don't know that they're insolvent? What if I guarantee them ROI of $10M/year on their purchase so they can get a loan and pay me the cash? Still all GAAP-compliant? Still all real? And keep in mind this isn't a binary question---some of the demand can be real and some manufactured. Maybe my customer had $20M, and ordered $100M since I was guaranteeing the ROI. These are the reasons I am "concerned", as you put it. And as you say, those concerns are reasonable.
> You keep saying this. That is not what the article says. It says the backstop is currently valued at $6.3B. That is not the cap.
We don't know what the maximum is because some of the terms are confidential. But if you're going to talk about this agreement so confidently, you should read the actual MSA:
The irony of this is that there are a number of scenarios under which Nvidia can legally terminate the agreement and most of them are precisely the kind of scenarios that would exist if CoreWeave came under significant financial distress. So contractually, the backstop isn't ironclad at all and worth far less to CoreWeave as you seem to believe.
Please, if you're going to make hyberbolic claims about what's going on, at least take the time to read what has been filed with the SEC. The picture is not as black and white as you make it.
> No one is saying Nvidia was spending a dollar for every dollar. It doesn't have to in order for the demand to be manufactured.
So have you quantified how much money Nvidia has to spend to generate a dollar of artificial demand?
If you're going to imply that CoreWeave was induced to take on tens of billions of dollars of debt to buy chips so it could have more compute capacity than it actually believes it needs with a $6.3 billion backstop that could disappear if it came under financial distress, please walk through the math.
> So answer me this: if I take a GPU order from an insolvent individual for $100M, can I book that $100M as revenue and be GAAP-compliant? Is it real?
No, under ASC 606 you cannot recognize revenue unless it's "probable" that you'll collect substantially all of the consideration due. And revenue would not be recognized until title to the assets being sold was actually transferred; a purchase order with nothing shipped is a backlog.
Before you ask these questions, why don't you do some research? These are not complicated accounting questions.
> What if I don't know that they're insolvent?
CoreWeave is publicly traded. Feel free to evaluate the financials and explain how the company is insolvent.
Stockholder equity is positive to the tune of nearly $5 billion, so its assets exceed its liabilities. And it generated nearly $3 billion in net cash in Q1 and is currently servicing its debt obligations. It even added an $8.5 billion non-recourse debt facility and was upgraded to positive from stable by S&P.
You don't get those types of debt facilities and S&P upgrades if you're insolvent. And ironically, insolvency is one of the potential triggers for the Nvidia backstop to go away.
A smarter discussion would be around CoreWeave's leverage and what happens if the AI demand dries up. That would be infinitely more useful than trying to make arguments that you haven't even researched.
> What if I guarantee them ROI of $10M/year on their purchase so they can get a loan and pay me the cash?
You haven't actually demonstrated that Nvidia has guaranteed CoreWeave a positive return on its investment. Once again, the $6.3 billion backstop is nowhere near the amount of money CoreWeave has spent building out capacity, which is tens of billions of dollars.
>We don't know what the maximum is because some of the terms are confidential.
So we agree on something: the $6.3B is not the cap: we don't actually know what the cap is. (And for the record, I never spoke confidently about the agreement, I spoke confidently about what was stated in the article.) The SLA you linked is the operating agreement that is effective once the revenue guarantee comes into effect, it is not the agreement that guarantees the revenue. Insolvency is a canard: the whole point, from CoreWeave's perspective, is that the guarantee helps insure that it won't go insolvent if demand doesn't materialize. It can essentially pass the loss back to Nvidia. That's also why you indicated that it makes the deal more attractive to third-party lenders, correct?
>So have you quantified how much money Nvidia has to spend to generate a dollar of artificial demand?
No, because the precise amount is irrelevant. What matters is that the chip demand is induced, not organic.
>CoreWeave is publicly traded. Feel free to evaluate the financials and explain how the company is insolvent.
I did not state that CoreWeave was insolvent. I was pointing out a flaw in your apparent belief that reported, GAAP-compliant revenue is unimpeachable. At least now, you're asking the right questions.
>A smarter discussion would be around CoreWeave's leverage and what happens if the AI demand dries up.
That is the discussion I have been trying to have. Nvidia's demand-goosing is only one piece of a much larger circular system. One way that demand would appear to dry up is that it was never as rapidly growing as these deals assume it is. Once everyone in the chain buys into an assumption of growth, it can become a self-fulfulling prophecy, at least until reality becomes unavoidable. Right now, the assumptions are not really about AI demand, they are about data center demand. That's where the money is being spent, and that's where the circularity is appearing. We are in a regime where companies are richly rewarded whenever they participate in a data center deal, because "everyone knows, AI is going to be huge." There is a good discussion of how this is happening in https://www.groundbrkr.com/p/the-second-derivative-why-no-on... (see section III, "the AI Boom is a Credit-Driven Real Estate Cycle"). The gist of the article is that actual AI demand doesn't have to "dry up" for the system to collapse, it only has to accelerate at a slower rate than the assumptions made to support the data center deals.
>You haven't actually demonstrated that Nvidia has guaranteed CoreWeave a positive return on its investment.
You're misreading the meaning of the term "investment" here; only the ROI on the GPU purchase is relevant with respect to Nvidia, not any pre-existing 3rd party debt. I'm not saying Nvidia is guaranteeing CoreWeave's net profit as a business, only the marginal ROI it expects from the chips. Which still means that it is incentivized to buy to whatever the limit is of the guarantee, independent of organic demand. And remember, real demand doesn't have to decline, it only has to slow its acceleration. Goosing demand as Nvidia has done is clearly risky in that environment. And it also raises the question, why they had to do it at all if demand is so robust?
Great explanation. Maybe another metaphor, it’s like a builder/developer buying land from someone. They own the land, they get the title, it’s theirs.
The land owner saying “hey if you can’t sell all the apartments we’ll buy what’s left” doesn’t in any way negate the sale or revenue accounting as per GAAP etc.
I mean, okay sure, but modify the counterexample they suggested slightly and then it's the same thing.
If GM promised to "rent out" (instead of buy back) the cars it sold to Hertz as a backstop (if not enough customers are renting), then the comparison is apt.
If GM sold cars to Hertz and then agreed to rent them from Hertz if Hertz was unable to rent them, it would not be consignment. It would be a sale and then purchase commitment, with the cost of the rentals taken as an operating expense.
Is the CoreWeave-Nvidia arrangement "good"? Time will tell. But there's no accounting issue here and even non-accountants can educate themselves on the subject because the least effective way to criticize these deals is to make accounting arguments that don't align to actual accounting principles.
This seems like a really narrow interpretation of what's going on. Is there any room to doubt/discuss whether GAAP rules could be improved? Or why the deal has been structured this way?
Why shouldn't we look through this arrangement? NVIDIA isn't in the business of purchasing outsourced GPU time. They could make better use of unused GPUs by repurchasing them for resale to another customer. If they're not doing that, it already seems likely that they specifically did this to guarantee that the revenue could be recognised.
Sure, NVIDIA's risk exposure could (legally) sit on their books without being recognised until it's already too late. That doesn't mean we shouldn't scrutinize them.
> Is there any room to doubt/discuss whether GAAP rules could be improved?
Looking at the structure of the deal and analyzing the risks is perfectly valid. Screaming "accounting subterfuge!" when this is simple GAAP accounting is a different matter.
CoreWeave is buying chips from Nvidia, paying Nvidia full price, and taking title to them. Nvidia has no right to take them back. It instead has a potential obligation, subject to various conditions, to purchase a separate service (compute) from CoreWeave.
GAAP rules are updated on a regular basis. If you want different GAAP rules for this type of deal, you at least need enough knowledge about accounting to make a sensible suggestion.
> NVIDIA isn't in the business of purchasing outsourced GPU time.
This simply isn't true. Google "DGX Cloud". Nvidia has a real business selling cloud-based compute for training foundation models and running heavy AI workloads, and leasing compute from its hyperscaler chip customers instead of competing with them was a strategic decision Nvidia made.
So yes, these types of arrangements should be scrutinized. But to do so intelligently requires a basic grasp of accounting rules and the business models.
I'm not the commenter claiming that this currently violates GAAP - that's someone else.
To summarise my opinion, subjectively it seems like a better distinction could be made in GAAP to look through this agreement and others like it. (Hypothetically if Hertz agreed to rent back rather than repurchase, like mentioned in a previous comment, that would also be suspect). But I'm not the one to propose what the preconditions would be.
> Looking at the structure of the deal and analyzing the risks is perfectly valid. Screaming "accounting subterfuge!" when this is simple GAAP accounting is a different matter.
It can be legal and still be subterfuge. Everyone involved in the deal has a clear incentive to ensure Coreweave gets to recognise revenue, and gets to show growth on paper. It's the same reason why SoftBank paying OpenAI $800mln for services in 2025 stinks a bit - they don't need the services but the deal goes ahead anyway.
> This simply isn't true. Google "DGX Cloud". Nvidia has a real business selling cloud-based compute for training foundation models and running heavy AI workloads, and leasing compute from its hyperscaler chip customers instead of competing with them was a strategic decision Nvidia made.
Sorry - you're entirely correct here. Though remember we're talking about a scenario where Coreweave aren't able to sell their capacity. If there's such a dramatic hole in demand, who are NVIDIA selling their compute to? This repo agreement won't give NVIDIA capacity that they need in the 90% of cases but will force them to purchase capacity they won't need in the 10%.
> To summarise my opinion, subjectively it seems like a better distinction could be made in GAAP to look through this agreement and others like it.
There's two things here: accounting and disclosure.
The accounting, which is what GAAP deals with, really doesn't seem problematic. CoreWeave is giving Nvidia cash for the chips and taking title to them. There's no associated repurchase right or obligation. So treating this as a sale and booking the revenue is the most sensible accounting approach. Trying to make it into something it's clearly not because it makes some people feel better isn't sensible.
I think the more important discussion is around disclosure: how much information Nvidia should be required to provide about its relationships with companies like CoreWeave, and where and when. Right now, we have to paint the picture based on multiple disclosures. We know about the equity stake through a 13F. The backstop was in an 8-K that was filed two years after the agreement was signed. The equity stake is not high enough that most of the rules around "related party" disclosures come into play.
I suppose you could make the argument that the market obviously sees the circularity here despite the patchwork disclosures that apply, so the circularity is ostensibly being priced in to the stock prices, debt, etc. But there's a legitimate argument that the market would be better served if disclosure was earlier and cleaner.
Even so, none of this would prevent Nvidia from engaging in these types of transactions because there's nothing inherently illegal about them.
> If there's such a dramatic hole in demand, who are NVIDIA selling their compute to?
NVIDIA itself is also training foundation models (and open-sourcing them). If there is excess compute available, NVIDIA can increase the scale of such models.
>Screaming "accounting subterfuge!" when this is simple GAAP accounting is a different matter.
Looking back up the thread, I don't see anybody screaming about anything. And I think "accounting subterfuge" is a broad concept that could certainly include GAAP-compliant but nonetheless suspect business practices.
>CoreWeave is buying chips from Nvidia, paying Nvidia full price
I'm not sure this is the case. They are agreeing to pay them some price, it's not clear whether they are getting them for cash or credit but I strongly suspect it's on credit. That doesn't change the GAAP compliance, does it? As I said before, I think they are exploiting an accounting loophole, regardless of whether it is strictly compliant.
> I think they are exploiting an accounting loophole...
With all due respect, you haven't articulated what that accounting loophole is. I've explained why the examples/comparisons you've made aren't equivalent according to GAAP.
From everything I've read and seen disclosed, CoreWeave pays full price for its Nvidia chips. Nvidia is not financing the sale. CoreWeave has taken on large amounts of debt financing from unrelated third parties. It's highly like that the Nvidia backstop helped CoreWeave get better financing terms, but Nvidia isn't actually providing the financing.
If CoreWeave is paying cash and taking title to the asset, and Nvidia has no obligation or right to take the asset back, it is GAAP 101 that the transaction would be booked as a sale because...that's what it is.
I think the definition of an accounting loophole is something that is technically legal but nonetheless suspect because it lets you appear to get something for nothing. According to your best-case scenario, Nvidia helps CoreWeave get a loan it would otherwise not get by guaranteeing revenue for CoreWeave; this allows CoreWeave to borrow money and give it to Nvidia as revenue; bulls point to Nvidia's revenue as a reason to ignore obligations like the CoreWeave guarantee, because after all, just look at all that revenue! And of course it makes GPUs look scarce and valuable, which helps CoreWeave get the next round of debt financing, since presumably the GPUs are the collateral. And the cycle starts again.
I'm not saying these deals are crooked, but the incentives are aligned so that everyone involved is biased toward over-estimating real demand. They are systematically prone to spinning out of control.
> I think the definition of an accounting loophole is something that is technically legal but nonetheless suspect because it lets you appear to get something for nothing.
But you haven't even articulated what the loophole here is.
> According to your best-case scenario, Nvidia helps CoreWeave get a loan it would otherwise not get by guaranteeing revenue for CoreWeave; this allows CoreWeave to borrow money and give it to Nvidia as revenue; bulls point to Nvidia's revenue as a reason to ignore obligations like the CoreWeave guarantee, because after all, just look at all that revenue!
Nvidia's guarantee is almost certainly a consideration for CoreWeave lenders that could lead them to provide financing on more favorable terms than they would if there was no guarantee. However:
1. Building out capacity for CoreWeave isn't just about buying chips. It has to build datacenters, pay for electricity, etc. The amount of debt raised ($35 billion+) far exceeds what it has paid Nvidia so lenders are nowhere close to having a make-whole guarantee from Nvidia here.
2. The backstop is subject to termination if certain events occur, and these events are far more likely to be triggered if CoreWeave comes under financial distress, which is when it would need the backstop the most.
It's not that there are no risks here; it's that you haven't actually articulated in legitimate terms what they are and you haven't quantified anything.
>But you haven't even articulated what the loophole here is.
I did, but here it is again: if you book revenue for sale of an asset where you guarantee the ROI on that asset (not on the entire business, you keep confusing those two very separate concepts), that revenue is suspect. You can stamp your feet and turn blue in the face claiming GAAP-compliance all you'd like, but that revenue should be regarded skeptically, just as revenue from an insolvent customer should be.
>Nvidia's guarantee is almost certainly a consideration for CoreWeave lenders that could lead them to provide financing on more favorable terms than they would if there was no guarantee.
Ha, nice side-step. Certainly CoreWeave isn't benefitting here, it's just those poor lenders. C'mon, man. You're right that the lenders will ultimately be left holding the bag, but that doesn't change the fact that CoreWeave is being induced to buy chips to the maximum limit of the ROI guarantee, independent of underlying demand. I've repeatedly said that, and you keep completely ignoring it and complaining that I'm not describing the problem.
>Building out capacity for CoreWeave isn't just about buying chips.
You're assuming that CoreWeave has to build out marginal capacity for those chips. We don't know, because the agreement is not public. But all CoreWeave has to do is have the capacity, which could easily---even probably---come from capacity already built but unsold, by the time the guarantee comes into play.
>The amount of debt raised ($35 billion+) far exceeds what it has paid Nvidia so lenders are nowhere close to having a make-whole guarantee from Nvidia here.
Again, you keep attacking a straw man. Not only have I never said Nvidia was guaranteeing CoreWeave's entire debt, I've explicitly said they were not, and did not need to in order to make this deal suspect.
>It's not that there are no risks here; it's that you haven't actually articulated in legitimate terms what they are and you haven't quantified anything.
I have in fact articulated them multiple times, you simply either haven't read them or for some reason lack the capacity to understand what I'm saying. And as I have also noted multiple times, the exact amounts don't matter. If you guarantee ROI on a piece of equipment in order to sell more of it, that is a red flag when your official narrative is that you can't make enough of these things to satisfy demand. And that is just as true on a $1M deal as it is on a $10B deal.
This is a specific example of a more general problem: the ability of capital to manufacture demand and not merely satisfy it. An implicit assumption about supply and demand is that they are largely independent market forces, that demand is an organic and emergent phenomenon arising from the desires and ambitions of free individuals, and that supply reacts to demand. But first mass marketing and now hyper-targeted marketing turn that assumption on its head, and given sufficient capital, you can manipulate consumers into buying something that, left to their own devices, they otherwise would not. The harm that this can cause is most easily seen in the addictive context, but it appears throughout the market as a lost opportunity cost: to what more beneficial use would they have put that money had they not been subject to the manipulation?
Given that the President is a convicted felon who maintains that what he did was fine, and that he has pardoned thousands of unrepentant criminals, and that the vast majority of his party enthusiastically endorsed all this, I would say "pro-crime" is an understatement.
on a daily basis the current US president commits treason against the people of the united states, which im pretty sure even presidential immunity doesnt protect against.
Just one of the shady dealings with foreign monarchies, laudering their bribes directly to the president thru billion dollar purchases of worthless crypto "assets" ala world liberty financial; should land the president and his entire family in capital punishment
I believe treason is literally defined in the constitution itself as taking up arms against the country. By that definition, I think there have been few to almost no treasonous people in recent decades.
From Article III: "Treason against the United States, shall consist only in levying War against them, or in adhering to their Enemies, giving them Aid and Comfort."
For example, if Russia met the definition of an enemy, and if the President compromised the interests on the country in favor of Russian interests (presumably for his own personal benefit), that would probably count. The compromise would have to be knowing, though, mere incompetence is not enough. So if the issue ever arose, Trump has a pretty good defense.
I don't understand why people keep using things like this as their examples.
JFK met with Khrushchev. Bush Senior and Reagan had regular contact with Gorbachev. It's not weird for heads of state to talk to each other.
What's weird is for Trump to be paying tax dollars to cancel energy projects because he doesn't like windmills or using the FCC to investigate media outlets who criticize him.
There are plenty of good examples to use that it makes no sense to resort to evidence-free innuendo.
> President Donald Trump’s special envoy broke with long-standing protocol by not employing his own interpreter during three high-level meetings with Russia’s Vladimir Putin, opting instead to rely on translators from the Kremlin, a U.S. official and two Western officials with knowledge of the talks told NBC News.
This seems like the same class of complaining about something ridiculous. If Putin wants to lie to you about something, do you expect him to speak the truth in Russian and then have his translator tell you something else? What advantage would that have compared with lying to you in Russian and having it translated accurately?
You expect Putin of the KGB to presume the person he's speaking to doesn't have a recording device or the ability to remember the words he used and have them translated by someone else later? And if he did think that, wouldn't that be an advantage to the US of doing it that way, to capitalize on the chance he lets his guard down? And if he did let his guard down and Trump failed to capitalize on it, what disadvantage would the US suffer from that compared with the situation where he knows there is a US translator in the room and doesn't say it to begin with?
You expect the Russians to strip search the President of the United States to check for a listening device when they know that person will be in the room to hear whatever is said anyway, and the US to agree to that?
At which point they can tell that he wears a watch but not whether the watch is a recording device, and can't detect shielded devices that can record audio onto internal storage with no RF emissions, and still can't do anything if he remembers a phrase and asks for it to be translated later.
I think we're worried about the Trump administration lying about what they talked to the Russians about. A translator with loyalty to the USA might request what they heard.
Which is an even more fanciful conspiracy theory since the objection was Trump not bringing his own translator, a person he could otherwise personally select for their loyalty to him.
> Only Mr. Trump, who has alternately contradicted his own narrative of what was said and complained about a lack of fair coverage from a meeting only four people witnessed, could permit Ms. Gross to tell anyone about what she heard. The White House has not said whether Mr. Trump has asked her to do that.
Translators are civil servants with security clearances, not reporters. When the leaders of Germany and Brazil have a meeting, they may say things they don't intend for China and everyone else to know about, and it isn't the translator's job to report what was said to the media. Unless they've witnessed something on the scale of criminality that they need to act as a whistleblower (in which case they wouldn't need anyone's permission), they should never be telling anyone what happened in the room since their job is to keep those secrets.
The media are fully aware of that, but they're being disingenuous and pretending that allowing a diplomatic translator to act as a media witness is a request anyone is likely to grant. Then Trump predictably responds to that by removing the translator from the room, so now they're complaining about that instead.
There what is? That Trump contradicts himself, a thing that was never in dispute?
This is kind of the point. Trump is such a polarizing jackass that people default to attacking him even in cases when there is nothing there, and then end up eroding their own credibility by making too many accusations they can't actually substantiate instead of sticking to the ones they can.
You don't have to reflexively defend claims so silly that you can only try to escape by moving the goalposts.
That was literally the point. People should stop making farcical complaints that discredit the critics. When you cry wolf every day, eventually the wolf comes.
When the farcical comments are from major institutions writing under their own name, that's not a false flag, that's just what their own flag stands for now.
Wait, are you saying leadership at those institutions (say the Washington post) secretly want to help Trump while appearing to be aligned against him, because of say the Zionism angle or like maybe Bezos wants a president that will lower his taxes? If so, that’s not a thing many ears would be able to hear here. More parsimoniously, perhaps the journos and pundits crying wolf (to discredit Trump’s critics) are acting alone. It only takes a few, because real idiots on the internet will latch on to the farcical nonsense and amplify it further. It doesn’t mean NYT/WAPO stand for false flags, it just means they have some political actors within them using plausibly deniable attacks on Trump to help him because they are for lower taxes/zionism/etc…
>I still haven't figured out how he's profiting from Trump Accounts yet
I would guess that he could have been paid in many various ways (TrumpCoins anyone) by financial institution(s) that were set to benefit from the accounts. Have the trump crypto companies followed all of the KYC laws?
Two issues with this. One, it's profitable assuming you just keep serving the same model forever, which is not realistic in this market. A given model has a shelf-life, which these days is measured in months, not years. Which means that trying to separate the cost of training the model from the cost of serving it doesn't make much business sense. And two, for providers that provide inference only via open weight models, the margins quickly move to commoditization. The "someday" when frontier model providers can enjoy their current high inference margins without the burden of significant training costs is never going to arrive.
Commoditization means there's price competition. From a consumer perspective, that's good. You want it to be a low-margin, high volume, competitive business.
Although from a business perspective, it can end up being ruinous competition like solar panels or airlines. A stable equilibrium with prices neither too low or too high isn't guaranteed; it depends on market structure.
It's anyone's guess whether this reaches an equilibrium or not, but I still expect that there will be companies like OpenRouter and Fireworks that offer inference at reasonable prices.
>Commoditization means there's price competition. From a consumer perspective, that's good. You want it to be a low-margin, high volume, competitive business.
You and I may want it to be a low-margin, high-volume business. But the valuations of OpenAI, Anthropic, and much of the rest of the AI industry are not based on that assumption. They are based on the assumption that there will be a couple of winners, like in the smartphone wars, and that those winners will be able to maintain good margins.
There are low-end and premium smartphones, with different profit margins. Similarly, it seems like open weights models and high-end models could co-exist?
I don’t think there’s any way of knowing what the market structure will be in the end.
>Similarly, it seems like open weights models and high-end models could co-exist?
Low-end and premium smartphones are still overwhelmingly either Android or Apple. And I'm sure open weights and premium models will co-exist. But that's not the future the current valuations are predicated on. I agree there's no way to say where the market structure will end, but I think we know enough to say where it won't end up. And some of the supporting players are starting to see the handwriting on the wall:
> A given model has a shelf-life, which these days is measured in months, not years.
Not all new models are trained from scratch. ChatGPT 5.3 to 5.4 (and likely 5.5) was basically the same model, but probably trained a bit more, not a new model from scratch.
> The "someday" when frontier model providers can enjoy their current high inference margins without the burden of significant training costs is never going to arrive.
That is debatable. I believe the moat for the frontier model providers is the compute. At the level of 10 trillion parameters (that Fable/Mythos are rumored to have), you need serious compute to serve inference, and you also need serious compute to train. Will DeepSeek, Qwen, Kimi, GLM come up with a 10T new model anytime soon? I doubt that. People keep saying that the Chinese labs are catching up to the US big 3, and measured in months the gap is now only 4-6 months. I doubt a Chinese version of Fable/Mythos will be released in the next 12 months.
>ChatGPT 5.3 to 5.4 (and likely 5.5) was basically the same model, but probably trained a bit more, not a new model from scratch.
Then those models have an eroding moat and will be quickly driven down to commodity pricing. The only thing propping up inference margins are the cap-ex costs of training. That's the moat. That's why there's no way to win this game. You cannot have low training/infrastructure cost and high margins (such as would justify today's valuations).
>I doubt a Chinese version of Fable/Mythos will be released in the next 12 months.
reply