For the AI bubble too many people assume that large companies having a stake in it will of course know what they are doing, be careful and not expose themselves too much or do wild bets that don’t pay off. But looking at the level of capex from hyperscalers, the amount of circular financing by NVIDIA/google/microsoft, the level of debt raised for datacenters (and its associated raising interest rates), the lack of moat for AI labs, the absurd AI labs valuations, OpenAI ever increasing infra expenditure commitments (we are at more than $750B for 2030), Oracle dire situation (to say the least), the mounting pressure from China/open models, and the fact that 2 companies represent the vast, vast majority of the compute demand. None of that looks like a healthy, sustainable industry. In fact it looks like the most obvious financial engineering ever, where the only ones benefitting are NVIDIA, memory manufacturers, and hyperscalers. And they are doing what is necessary to keep the game going. If the demand for AI vendors isn’t increasing massively in the coming years the whole thing will go down. And the level of demand required need to be pretty much the AI booster dreams where everything becomes agentic everywhere. Short of that we are very likely to see things go downhill","offTopic":true},{"id":"38d39810-b9f5-4eb8-8bd0-b0c92a126423","excerpt":" — There also seems to have been a major shift in strategy from AMD / Intel / Apple / NVidia due to this, as I was used to CPU / GPU / chipset upgrades coming out more regularly with double-digit percentage improvements each year, but that seems to be basically over for the consumer space now t","url":"https://news.ycombinator.com/item?id=49335699","role":"pain","weight":0.8455833,"occurredAt":"2026-08-17T18:41:08.000Z","sourceKey":"hackernews","sourceName":"Hacker News","credibility":0.7,"venue":"news","intent":"feature_request","painScore":0.46,"sentiment":-0.25,"confidence":0.57916665,"matchedPatterns":["missing_feature"],"statement":"I'm fully anticipating AI hardware demand to dry up once Anthropic and OpenAI start tanking financially in a few years after missing earnings and investors and banks wanting their loans paid back combined with more folks realizing tha…","title":null,"body":"There also seems to have been a major shift in strategy from AMD / Intel / Apple / NVidia due to this, as I was used to CPU / GPU / chipset upgrades coming out more regularly with double-digit percentage improvements each year, but that seems to be basically over for the consumer space now that everyone is just trying to build as much AI hardware as fast as possible for enterprise markets.
I also used to rebuild my desktop every 1-2 years and upgrade laptops every 2-3 years, but I'm currently a bit past that and have no plans to refresh anything in the near future. I'll probably wait another CPU and GPU generation or two to even consider it, especially with prices being this ridiculous.
I'm fully anticipating AI hardware demand to dry up once Anthropic and OpenAI start tanking financially in a few years after missing earnings and investors and banks wanting their loans paid back combined with more folks realizing that LLMs cannot achieve AGI. After that, I would hope that the demand that's driving this pricing spike will fall along with it, but I also fully expect hardware companies to be as greedy as possible, so maybe a 32TB HDD will just cost $1000 going forward and the era of "a decent gaming PC made up of the tier of components one step below TOTL costs ~$2000" we've been in since the 1990s is fully over, and it's just going to be more like $5k going forward.","offTopic":false},{"id":"748bac00-27e3-4e5b-995f-fe64cecb6f01","excerpt":"Nvidia’s top 2 mystery customers made 39% of Q2 revenue, up from 25% last year, raising concentration risk concerns — No paywall: [https://www.cnbc.com/2025/08/28/nvidias-top-two-mystery-customers-made-up-39percent-of-its-q2-revenue-.html](https://www.cnbc.com/2025/08/28/nvidias-top-two-mystery-customers-made-up-39perc","url":"https://www.reddit.com/r/wallstreetbets/comments/1n4wvy8/nvidias_top_2_mystery_customers_made_39_of_q2/","role":"pain","weight":0.77583337,"occurredAt":"2025-08-31T14:58:16.000Z","sourceKey":"reddit","sourceName":"Reddit","credibility":0.62,"venue":"wallstreetbets","intent":"other","painScore":0.26666668,"sentiment":-0.6666667,"confidence":0.6125,"matchedPatterns":[],"statement":"Nvidia’s top 2 mystery customers made 39% of Q2 revenue, up from 25% last year, raising concentration risk concerns.","title":"Nvidia’s top 2 mystery customers made 39% of Q2 revenue, up from 25% last year, raising concentration risk concerns","body":"No paywall: [https://www.cnbc.com/2025/08/28/nvidias-top-two-mystery-customers-made-up-39percent-of-its-q2-revenue-.html](https://www.cnbc.com/2025/08/28/nvidias-top-two-mystery-customers-made-up-39percent-of-its-q2-revenue-.html)\n\nTwo Nvidia customers made up 39% of Nvidia’s revenue in its July quarter, the company revealed in a financial filing on Wednesday, raising concerns about the concentration of the chipmaker’s clientele.\n\n“Customer A” made up 23% of total revenue, and “Customer B” comprised 16% of total revenue, according to the company’s second-quarter filing with the Securities and Exchange Commission.\n\nThat’s higher than the same quarter a year ago when Nvidia’s top two customers made up 14% and 11% of sales, according to the filing.\n\nThe company regularly publishes information on a quarterly basis about its top customers, but the disclosure this week is fueling a renewed debate about whether Nvidia’s explosive growth is being driven by a handful of large cloud providers such as Microsoft, Amazon, Google and Oracle.\n\nNvidia finance chief Colette Kress said in a Wednesday statement that “large cloud service providers” made up about 50% of the company’s data center revenue. That’s important as the data center business made up 88% of Nvidia’s overall revenue in the second quarter.\n\n“We have experienced periods where we receive a significant amount of our revenue from a limited number of customers, and this trend may continue,” Nvidia wrote in the filing.\n\nIncreasingly, analysts are looking to those cloud capital expenditure spending commitments to model the future growth of Nvidia.\n\n“We see limited room for further earnings upside revision or share price catalyst in the near-term unless we have increasing clarity over upside in 2026 \\[cloud service provider\\] capex expectations,” wrote HSBC analyst Frank Lee in a note on Thursday. He has a hold rating on the stock.\n\nBut Nvidia’s Customer A and Customer B are not necessarily cloud providers. It’s a bit of a mystery, and an Nvidia representative declined to share the identities of Customer A and Customer B.\n\nIn its filing, Nvidia says it has both “direct customers” and “indirect customers.” Customer A and Customer B are listed as “direct customers.”\n\nDirect customers are not the end users of Nvidia’s chips. They’re companies that buy the chips to build into complete systems or circuit boards that they then sell to data centers, cloud providers and end-users. Some of these direct customers are original design manufacturers or original equipment manufacturers like Foxconn or Quanta. Others are distributors or system integrators like Dell.\n\nIndirect customers, meanwhile, include cloud service providers, internet companies and enterprises, which typically buy systems from Nvidia’s direct customers. Nvidia says it can only estimate revenue to indirect customers based on purchase orders and internal sales data.\n\nDeciphering if any of those cloud providers are Nvidia’s mystery customers is difficult, in part because the chipmaker has wiggle room in the definitions of its direct and indirect customers.\n\nNvidia, for example, wrote in the filing that some direct customers buy chips to build systems for their own use.\n\nAdditionally, Nvidia noted that two of its indirect customers each accounted for over 10% of its total revenue, primarily buying systems through Customers A and B.\n\nContributing further to the mystery of it all, Nvidia said that an “AI research and development company” contributed a “meaningful” amount of revenue through both direct and indirect customers.\n\nNvidia told investors on Wednesday that demand for the company’s AI systems remains high, not just among cloud providers, but among other kinds of customers, including enterprises buying systems for AI and “neoclouds,” which are companies that are taking on the biggest providers with services more tuned for AI. Nvidia also listed foreign governments, saying it would record $20 billion in revenue this year for “sovereign AI.” All of these product categories are contributing to Nvidia’s revenue growth, Kress told analysts on an earnings call.\n\nNvidia CEO Jensen Huang also said that the company has a new forecast of $3 to $4 trillion in AI infrastructure by the end of the decade. It said that it could take about 70% of the total cost of a $50 billion AI-focused data center, not just for its graphics processing units but for other chips it sells, too.\n\nHuang told investors it was a sensible target for the next five years because of how much hyperscalers were spending and committing to spend — $600 billion this year, according to Huang. He also said new kinds of customers, such as enterprises or overseas cloud providers, were joining the build-out.\n\n“As you know, the capex of just the top four hyperscalers has doubled in two years as the AI revolution went into full steam,” Huang said.","offTopic":true},{"id":"ab37cd1f-c113-48b2-a71b-753647c7401d","excerpt":" — Its bad if the expected demand is an illusion. For example, when a company builds out a data center they don't build it for demand today, they build it for the demand they expect when the data center is running and for how much they expect demand to grow over the lifetime of the data center (this is a simplific","url":"https://news.ycombinator.com/item?id=49075661","role":"pain","weight":0.64750004,"occurredAt":"2026-07-27T21:19:44.000Z","sourceKey":"hackernews","sourceName":"Hacker News","credibility":0.7,"venue":"news","intent":"other","painScore":0.4,"sentiment":-1,"confidence":0.4625,"matchedPatterns":[],"statement":"Its bad if the expected demand is an illusion.","title":null,"body":"Its bad if the expected demand is an illusion. For example, when a company builds out a data center they don't build it for demand today, they build it for the demand they expect when the data center is running and for how much they expect demand to grow over the lifetime of the data center (this is a simplification, they build a financial model of how they can grow capacity as demand increases over the lifetime of the data center). If the demand is lower than expected then the counterparty cannot spend that Y amount back. In other words, Nvidia now holds bad debt (really worthless equity since these aren't loans on paper). Furthermore, Nvidia has been making the same bet with multiple companies. That Y amount the counterparty can't pay back is probably correlated with all of the counterparties Nvidia lent X amount to. Suddenly this circular flywheel begins operating in reverse. Now Nvidia has no X amounts to lend to AI companies which makes their ability to pay back Nvidia worse which means Nvidia has less money to lend out and on and on.
There's other problems too, why do we think AI demand is ferocious right now? Nvidia's revenue is one of the biggest signals we use to determine that. Why is Nvidia's revenue so large? They're spending their revenue on more revenue. This process overinflates what AI demand might actually be.
The issue really boils down to that this is a risk that gets reported in a way that makes it look less risky than it really is and therefore actors make investment decisions that they might not otherwise make. Sure, it might work out. But if it doesn't, the pain could be way more painful than it looks on paper.","offTopic":false},{"id":"f612215f-55d3-483f-b485-020b817797b3","excerpt":" — It's possible to be economically feasible. It requires the ability to pay off the capex. Not only do they have to pay off their loans in record time (prob 2-3 years), they also have to keep spending every 3 years because the failure rate of GPUs is something like 20%. In addition they need to complete building ","url":"https://news.ycombinator.com/item?id=49352375","role":"pain","weight":0.616,"occurredAt":"2026-08-18T20:40:21.000Z","sourceKey":"hackernews","sourceName":"Hacker News","credibility":0.7,"venue":"news","intent":"other","painScore":0.4,"sentiment":-1,"confidence":0.44,"matchedPatterns":[],"statement":"It's possible to be economically feasible.","title":null,"body":"It's possible to be economically feasible. It requires the ability to pay off the capex. Not only do they have to pay off their loans in record time (prob 2-3 years), they also have to keep spending every 3 years because the failure rate of GPUs is something like 20%. In addition they need to complete building out the stuff they've started, which means not failing to acquire land, energy, and water (the most dangerously rare and absolutely necessary resource for AI), not dealing with collective bargaining, nor any increased shortages or price hikes in materials. So there is a lot of risk.
Ballpark that they need to make around 110 billion a year, each, to break even on these investments. Let's estimate 550 billion a year in necessary profit required for the major frontier companies. That means there needs to be 550 billion of money, available to customers today, that isn't being spent on anything else, that they will now spend on AI. Maybe some of that comes from increased value, efficiency, or layoffs. But 550 billion is not a small amount of money.
Spread over the whole globe, the cash is there. But it is a hunt for cash, combined with a battle to successfully complete their buildouts, keep them running, and make bank, before the bookie comes knocking.
The railroad panics of the 19th century (and subsequent depressions) happened because they over-leveraged private capital without the ability to profit from it quick enough. So it really is a question of 1) can they really build it all, and 2) will people really pay for it all. If either answer is No, we are looking at economic catastrophe.","offTopic":false},{"id":"4faab8e5-f62d-4643-8c1a-85a37dc85bc5","excerpt":" — > 1. The whole Burry argument is that AI hardware becomes obsolete faster. If aux hardware can be reused, that works against the argument.
Burry’s main argument is depreciation is being understated and the capex vintages will not be paid off before they are essentially useless. This can happen whether or not au","url":"https://news.ycombinator.com/item?id=49365990","role":"pain","weight":0.5914394,"occurredAt":"2026-08-19T19:19:15.000Z","sourceKey":"hackernews","sourceName":"Hacker News","credibility":0.7,"venue":"news","intent":"other","painScore":0.27878788,"sentiment":-0.6969697,"confidence":0.4625,"matchedPatterns":[],"statement":"The whole Burry argument is that AI hardware becomes obsolete faster.","title":null,"body":"> 1. The whole Burry argument is that AI hardware becomes obsolete faster. If aux hardware can be reused, that works against the argument.
Burry’s main argument is depreciation is being understated and the capex vintages will not be paid off before they are essentially useless. This can happen whether or not aux is reused.
> Total consumption drives more demand for the already supply constrained hardware.
Demand is the wrong metric.
Only number that matters is whether AI attributable revenue will be sufficient to pay back enough of each successive capex vintage (e.g. 750B this year, 1T next year, 1.2T in 2028) so that hyperscalers and neoclouds can either self-fund or continue to issue debt as bond markets are already straining and tax-payer backed sovereign debt is providing a high baseline. Otherwise they downgrade capex projections and the bubble pops.
Expensive compute needs expensive inference to justify 30-40B/year/GW of compute. There are many reasons why frontier API pricing which is what the industry is based on may not persist. It is also almost certainly the case that 2026 is the worst year of supply and demand mismatch to allow for 80%+ margins. HBF next year has the potential to single handedly pop the DRAM spot bubble.
> Can AI hardware market go bust? Sure it can.
This is the bear thesis. It is not that AI will crash or be useless.
> But being early is the same as being wrong in the investment market. When do you predict the bust to be?
Q4 27-Q2 28 is when the bill becomes due at the latest. There are sufficient financial levers left to buy time without returns until then.
> Google, Amazon, Microsoft, Meta are all buying as many Nvidia GPUs as they possibly can.
All of these companies have rock solid revenue streams and can easily swallow 500B of capex devaluation over time. Their buying of Nvidia today is not necessarily the indicator you are implying as there are strong competitive reasons to make the game more expensive for everyone else.","offTopic":false}],"breakdown":[{"sourceKey":"hackernews","sourceName":"Hacker News","count":5},{"sourceKey":"reddit","sourceName":"Reddit","count":1}],"total":6}}