Openai, Nvidia Fuel $1t AI Market with Web of Circular Deals
Key topics
The article discusses how OpenAI and Nvidia are fueling a $1T AI market with circular deals, raising concerns about the industry's financial practices and potential bubble. The discussion revolves around the legitimacy and sustainability of these deals.
Snapshot generated from the HN discussion
Discussion Activity
Very active discussionFirst comment
-83938s
Peak period
111
0-12h
Avg / period
26.7
Based on 160 loaded comments
Key moments
- 01Story posted
Oct 8, 2025 at 7:04 PM EDT
3 months ago
Step 01 - 02First comment
Oct 7, 2025 at 7:45 PM EDT
-83938s after posting
Step 02 - 03Peak activity
111 comments in 0-12h
Hottest window of the conversation
Step 03 - 04Latest activity
Oct 15, 2025 at 3:36 PM EDT
3 months ago
Step 04
Generating AI Summary...
Analyzing up to 500 comments to identify key contributors and discussion patterns
Want the full context?
Jump to the original sources
Read the primary article or dive into the live Hacker News thread when you're ready.
It’s all about ‘accelerating’ forward at all costs.
Bubble's only bad if you get out at the wrong time.
Is it worse than the dot-com bubble? I remember everyone and their mom who knew HTML could get hired, and there were way more companies that went IPO during the dot-com bubble, like theglobe.com.
AI is a bubble, but is it worse than the dot-com bubble or the real estate bubble in 2008?
IDK, the number of companies and employees might be smaller, but the valuations and comp packages are so much bigger that I suspect there's more money in the bubble this time around, just a bit less spread out.
Now once these folks don't get 800B per year in revenue and the money runs out, all of the banks go as well. But don't worry - they'll get bailed out with our money...
The money between OpenAI, Nvidia, Oracle, AMD is not circulating. There is no cashflow, only future commitments that may (and quite likely will) collapse. Yet the stock market & media react as if it's a sure thing. Even in the criticisms of these deals, the hype is affirmed.
This is the same problem as Enron's accounting, minus the fraud. (No need for fraudulent accounting when people simply don't read the fine print.)
I deeply feel buybacks shouldn’t be illegal but treated shamefully.
Instead of using profits to build up long term savings or fund R&D, they basically choose to do as little as possible.
There is no vision.
Isn't buying stock a form of long term savings? After all, they can always sell that stock when they want to "withdraw".
Sure, they may not get the same return as simply stashing it into a bank account, but it's also a statement of confidence in their future: "We are sure that our stock will outperform every other option there is for storing our money, because our long-term plans include extracting more profit from the market."
Not just different, it specifically a lower tax rate assuming that the stockholder has held long enough to use the long term capital gains tax rate (which lower than the dividend tax rate).
When the fed investigates this does it matter if one of the 3 companies is not a publicly traded company?
The issue is that it's an industry of investment which exists solely to power more investment in AI - the entire chain is still assuming that someone will eventually pay for this.
At the end of the day all that money leaks out to employees and suppliers...but no one those people transact with may have any interest in buying what was produced.
It's entirely based on the perception that LLM training & inference is here to stay at ever growing scales when the shortcomings of Artificial Dreaming are increasingly scrutinized. Not all businesses want to end up paying refunds to their clients like Deloitte [1] because the LLM hallucinated crap into their reports (and they failed to correct it).
[1] https://www.theguardian.com/australia-news/2025/oct/06/deloi...
This assumes Deloitte didn't make more $$$ from the deal by "outsourcing" it to AI than not. It was a partial refund.
And you haven't considered that Deloitte might get this money some other way anyway. It's been budgeted and needs to be spent by the government department for this financial year. They'll find another project to hand out.
I used it to learn terraform and have a kubernetes cluster running.
I know it's always trendy to bash oracle and simp for AWS, but it is THE best option for learning and hobbyists.
Learning terraform isn't hard, it's what you stitch together with terraform that is hard.
Can you get that without adding a payment mechanism? For some reason, I'd be more cautious about putting my card into Oracle's offering than anyone else.
No!
There are actually two free tiers.
The limited free tier you get without adding a CC and then this advanced "always free" tier I am using, where you have to use and verify a CC and have to pay big bucks if you mess up.
limited free tier: 300$ to be spent in 30 days
advanced always free tier: https://docs.oracle.com/en-us/iaas/Content/FreeTier/freetier...
The big limitation is the 10mbit egress bandwidth, but you can work around that with (free) Cloudflare.
What's he going to do with all that money, and what does he care for the risk it's bad or shady?
Worst case, he got to be #1 for a bit for a few dozen billion, best case he's hoping AGI will extend his life before he croaks.
The implied equity risky premium is getting pretty low - a sign of greed.
As a more serious point related to this though, I was under the impression that organ replacement didn't address issues with telomere length?
The thing that unites Ellison, Trump, Musk, Thiel and a fair few of the other politically active billionaires is the obsession with "legacy" - they want to leave their mark in the history books, figures which will likely be remembered and taught in schools in thousands of years similar to Roman emperors.
Musk is the most obvious with his obsession of settling (and eventually dying) on Mars, Trump is dreaming of getting a Nobel Peace Prize (if only to not let Obama be the only US President who got one), and the rest is hunting for the "invented AGI" crown.
We’re already seeing revolutions elsewhere — and it’s likely that trying to loot them further by generations who sold out the nation will simply lead to social collapse.
A pack of 3 oldies burst through our perimeter one winter night... the screaming woke me up. Outside my tent the forest was lit up red by our laser blasts, trying desparately to take them out.
We thought that a revolution would be a good idea, but an upside-down population pyramid is a hell of a thing when you're on the bottom.
Just using Google / Gold as a comparison [1].
Assume you have 100 units of each.
In late 2021, Googs gone up ~100% so you have to rebalance because you have $200 in Goog and $92 in Gold. So lets say you rebalance to 80 Goog (160$) and 144 Gold ($130).
In late 2022, Googs gone down ~40% so you have to rebalance because you have $96 in Goog and $141 in Gold. So lets say you rebalance to 100 Goog ($120) and 118 Gold ($112).
So over the course of 2 years Goog has gone up 20% and Golds gown down 5% but your investments are overall up 16%. Obviously a 100% Goog investment is higher but with more risk.
If you didn't do any rebalancing then you have a gain of 7.5% (100*1.2 + 100*0.95 = 215)
[1]: https://www.google.com/finance/beta/quote/PHYS:TSE?compariso...
Very simple example, and not the only way to do it - but people need to remember net worth being 500B is not 500B in the bank, and at some point the number doesnt matter
Inflation on the other hand definitely does reallocate real value to whoever already owns a lot of assets, and I believe inflation often goes along with bubbles, since credit expansion is what usually kicks off the bubble in the first place.
Only if those are the only actual real customers. It's not. The pie is a lot bigger. And with the current hype some other AI company will just take over and the bubble will continue.
In terms of actual earnings not sure Tesla is doing better.
>By 1998, Yahoo was the beneficiary of a de facto Ponzi scheme. Investors were excited about the Internet. One reason they were excited was Yahoo's revenue growth. So they invested in new Internet startups. The startups then used the money to buy ads on Yahoo to get traffic. Which caused yet more revenue growth for Yahoo, and further convinced investors the Internet was worth investing in. When I realized this one day, sitting in my cubicle, I jumped up like Archimedes in his bathtub, except instead of "Eureka!" I was shouting "Sell!"
The real problem seemed to be that you can only put so much money into pets.com before it becomes stupid. You had more short term investment capital than could be _effectively_ spent at the time. The long term players, as usual, avoided the Archimedian idealism, and were heavily rewarded anyways.
pg has startup brains.
Words have meaning in context, and calling something a "Ponzi scheme" these days means any pyramid-like system that only works as long as new investors come to the table, because that new investor money is the sole source of gains for earlier investors. But once you run out of new investors, which is inevitable, the whole thing collapses. What pg described was exactly a Ponzi scheme in that sense, even if it wasn't a deliberate scam. And that's very different from other types of business ventures that eventually fail because of the more "normal" reason that they just don't gain traction.
Language is fluid, and many of words got their very known & popular meanings later, when people used for things that it didn't exactly mean what it was initially intended to, and it got popular.
People are ever less empathetic and sociable, this being the reason for such comments.
In this case dotcom investing was so stupid that it should have been recognized as stupid at the time, without hindsight. For someone who lost money, gave bad advice etc. insisting that they didn't fall for a Ponzi scheme means representing themselves as less terrible investors, regardless of facts.
And you do recognize the context of those words was 15 years ago at this point?
> was exactly a Ponzi scheme in that sense, even if it wasn't a deliberate scam
There are better terms of art to describe the scenario. If we lacked for those words I might grant you this point, but since we don't, I find the description lacking in precision if not completely faulty.
> the more "normal" reason that they just don't gain traction.
Is that actually the "normal reason" most businesses fail? I'm not sure it is.
I don't get this. Of course there was. Specifically, March 10, 2000.
To declare that something is not worth investing in doesn't imply that it's useless. (and actually, swing investors will argue something can be worth investing in even though it is in fact useless, e.g. web3 and most crypto).
That means: 1) Late investors end up taking a bath because it will actually take 10x longer to get a return than they thought. 2) Investment becomes inefficient - there are lot of GPUs being bought in 2025 that should have been bought in 2030. By the time 2030 comes along, those 2025 GPUs will be outdated, so the value they will provide will be less than if they had been purchased at the correct time
Though I guess with high velocity of money circulating in tight loops, everyone involved can feel rich. 1 million dollars changing hands once per day can potentially mint 365 'millionaires' during the course of a year. If they just pass it back and forth among themselves to buy each other's stocks and products and don't let that money escape their network to actually pay for something useful, they can all be millionaires... On paper.
Not to mention how relatively small amounts of money, moving at high velocity, can inflate asset prices... On Alice's birthday, Bob can buy 100 shares of Alice's shell company at a price of $10 per share. If Alice owns 1 million shares of the company (which she founded), it means that Alice's net worth is at least $10 million... It cost Bob only $1000 to give Alice a net worth of $10 million. Now imagine that same $1000 moving in and out of that company's stock, traded 1000 times per day between various people. With just $1000 circulating back-and-forth at high frequency (A.K.A. high-frequency trading), you can generate a trade volume of $1 million per day... So it all seems reasonable; a company with a $10 million valuation with $1 million daily trade volume... Nothing suspicious. So you can basically start a shell company and fake everything; from the market cap, to the trade volume; using only a relatively small amount of money. This is what they were doing in the early days of crypto.
The same money can hop around in circles between an insane number of people when it's just 'revenue' because revenue is not taxed (after expenses). Only profits are taxed... But the same money, as profit, can barely hop between 6 people before it's taxed down to just 10% of its original number. High-velocity revenue can severely distort people's perceptions of the market, especially in the era of media filter bubbles.
Under GAAP (normal accounting) A has a profit.
In reality are they profiting? Who knows - it depends how A does as a business.
It's crazy how all these small subtleties in definitions of various terms can create something so abominable in the aggregate. Makes one want to reach for a tinfoil hat. Not sure I can even trust the company which makes the tinfoil at this point.
You see companies hand waving this sort of thing away on earnings calls when they have a really bad quarter but the CEO is blowing smoke at shareholders saying “On non-GAAP this was an amazing corner.” Which is code speak for “that terrible decision I made a few quarters ago is coming home to roost and our lawyers have said I need to reflect that on our books. If you ignore where we screwed up badly, our numbers look good!”
Some people might interpret or represent the aforementioned transaction as justification for thinking Alice's net worth is at least $10M, but it is not an objective measure of Alice's purchasing power (as it would be in the case that Alice had $10M cash).
>The same money can hop around in circles between an insane number of people when it's just 'revenue' because revenue is not taxed (after expenses). Only profits are taxed
On the US federal level. Not sure about other countries, but quite a few other governments in the US do tax revenue.
https://taxfoundation.org/data/all/state/state-corporate-inc...
>Nevada, Ohio, Texas, and Washington impose gross receipts taxes instead of corporate income taxes. Delaware, Oregon, and Tennessee impose gross receipts taxes in addition to their corporate income taxes. Some localities in Pennsylvania, Virginia, and West Virginia likewise impose gross receipts taxes, which are generally understood to be more economically harmful than corporate income taxes.
See also:
https://en.wikipedia.org/wiki/Gross_receipts_tax
> OpenAI: We would like six gigawatts worth of your chips to do inference.
> AMD: Terrific. That will be $78 billion. How would you like to pay?
> OpenAI: Well, we were thinking that we would announce the deal, and that would add $78 billion to the value of your company, which should cover it.
> AMD: …
> OpenAI: …
> AMD: No I’m pretty sure you have to pay for the chips.
> OpenAI: Why?
> AMD: I dunno, just seems wrong not to.
> OpenAI: Okay. Why don’t we pay you cash for the value of the chips, and you give us back stock, and when we announce the deal the stock will go up and we’ll get our $78 billion back.
> AMD: Yeah I guess that works though I feel like we should get some of the value?
> OpenAI: Okay you can have half. You give us stock worth like $35 billion and you keep the rest.
https://www.bloomberg.com/opinion/newsletters/2025-10-06/ope...
https://archive.is/tS5sy
> AMD: Oh, since you put it like that, why don't you pay us $x for the chips and give us $ 0.5 x worth of stock? When your market cap rises by $ 0.5 x you'll get the money for those stocks. And hell, it might even rise further, giving you back some of the money for those chips. You're really robbing me blind with this deal, you'd be a fool not to take it!
Matt's telling is a little facetious, so the exact negotiation probably didn't go like that. More importantly, OpenAI has better options. If they want to pay AMD cash for chips, they could do that without also giving up their equity. The same can't be said for AMD. They're the second choice when it comes to GPUs, and they're desperate for market share, so they're willing to cut deals like this. Without this deal OpenAI probably would have bought Nvidia chips with cash.
> NVIDIA intends to invest up to $100 billion in OpenAI progressively as each gigawatt is deployed.
Replace AMD & OpenAI with some crypto company and Trump, hypothetically the latter doesn't really obviously get any significant benefit from the former's endorsement.
From the OpenAI side, it’s an amazing deal. $35B profit in one day… at a pe of 30, it’s already worth a trillion if it can repeat similar deals every year.
As to the stock going up - we have a bit of a stock bubble thing going on.
0 - https://en.wikipedia.org/wiki/AOL#2000%E2%80%932008:_As_AOL_...
OpenAI is good at deals - https://news.ycombinator.com/item?id=45493815 (40 comments)
Now we are just transferring the power from bankers -> tech/AI.
But "circular deals" has such a nice ring to it, that you hear it everywhere nowadays. People are just hungry for negative soundbites.
But this is not circular. Circular would be if I sell you an apple worth $0.25 for $2.00, and then you sell it back to me for $2.00, or other similar amount, and I get to mark all the apples in my inventory at $2.00 and show a huge profit (on paper). One can create variations of this blatant deal. Like I sell you some rubber for 10 times the market price, you make a balloon and then I buy the balloon for 10 times the market price. I may not have other balloons in my inventory, but plenty of rubber, and I show some nice profit. One can imagine other, fancier deals.
But in the case of AMD and NVidia, and OpenAI and Oracle, the direction is clear. OpenAI has a clear need for compute. They can buy it directly from NVidia and AMD, or indirectly from Oracle. They can buy it with hard cash (of which they don't have that much), or with their own equity, or some form of deal that offers the seller an upside in OpenAI's equity.
But there is not back and forth buying of the same item, or of rubber/balloons. All the deals seem legit. Is it possible that all the future compute will not be needed, because the AI craze will fizzle. It is, lots of things are possible. But that's general business risk.
When you draw a circle on a piece of paper do you put your pencil to the paper, start drawing a curve, stop and write the word FRAUD, and then complete the curve? Or do you just draw a circle?
I’m assuming that people are saying “circular” in that it looks like the money goes in a circle. (For one example) Nvidia invests in Lambda, Lambda buys GPUs from Nvidia, Nvidia leases GPUs back from Lambda, Lambda uses the revenue from leasing the GPUs to Nvidia to raise debt to buy more GPUs from Nvidia…
When you replace people with companies, the financing can become much more complex. The example you provided with Nvidia and Lambda seems quite reasonable to me. Here's an example that happens every day in the world of housing: banks lend money to house buyers. Then they package the mortgages and sell the resulting mortgage back securities. Then they take the money from the proceeds, and give more loans, and package them and create more mortgage back securities. Seems circular, right? But that's just how business is done. There is no Ponzi aspect to this, or fraud, or smoke and mirrors. It's just every day business. Nobody labels that as being circular.
> Is that circular?
Doesn’t really seem so because at the end of the day the money goes from me to them. I don’t get my money back, I get a car in exchange for my money.
Also this deal didn’t begin with the manufacturer purchasing shares of me before offering me debt to buy a car from them.
>But that is exactly what happens most of the time when people buy a car.
Your car manufacturer leases your car back from you? And you use that revenue to raise debt to buy more cars from them? What manufacturer are you doing this with? What do you end up driving?
Honestly, if this was possible people would be doing it (not that they are not - fleet services and rental fleet services do some pretty funky accounting sometimes, so I wouldn't be surprised at all if this sort of thing happened).
If this was possible, I'd be doing it.
> What manufacturer are you doing this with?
Good question. The minute an answer gets posted I'm going to have a really good side-hustle.
> What do you end up driving?
I presume, at the point that this is being done, there is no actual car involved, so nobody involved will be driving because there is no car to drive.
Yeah “everybody buys cars through the infinite car glitch” sounds like the sort of thing that would be part of an enormously long answer to “should I try WoW for the first time in 2025”
You sound like you know what you’re talking about. I only think I know what I’m talking about. Help me understand: What am I missing in the OpenAI / AMD deal that makes it non circular?
Honestly, I'm not an expert either, but I've run a company, and I can all but guarantee that credit_guy above really does not know what he is talking about.
I've replied else-thread, in detail, on exactly why his analogy to mortgage and car loans are incorrect.
His main point that "these deals cannot be circular because mortgage and car finance are not circular" is incorrect. The mortgage and car finance deals are not analogous to the Nvidia/OpenAI or AMD/OpenAI deals.
AMD: We need cash too, you know? We'd love to sell you those thousands and thousands of high end GPUs, that would cover some of our R&D, and we could one day match NVidia. But we don't swim in cash either. We can't give you the discount.
OpenAI: What can you give us? You must be able to throw in something there. Otherwise, honestly, we can't make the deal.
AMD: What if we give you some equity? And if our deal goes well, and our GPUs start being viable alternatives to Nvidia's, maybe we'll be able to get close to Nvidia's market cap and even surpass them, just like we did with Intel.
OpenAI: Brilliant. We love it.
AMD: Yes, but that equity will be contingent on how the deal goes.
OpenAI: Sure thing, we'll take that.
This process is described in the Bible for example!
The AI startup valuation largely is. I feel it quickly becomes circular because people make projections purely on other projections since the world is too impatient to wait and find out.
A single hamburger store is never going to be projected to have a billion dollars of revenue because people understand the total addressable market. Doesn’t matter how good the burger is.
The AI stuff is too new that people don’t have a firm grasp on the costs and profit opportunities. They don’t really even know how to define the TAM. Too many unknowns. Entire classes of labor could be replaced by AI —- or perhaps not.
With little grounding, it quickly becomes a circle of hype.
OpenAI => nVidia => OpenAI. That's pretty circular.
145 more comments available on Hacker News