Ouch, investment in GPU isn't a one off, you also need to invest newer generations (that naturally comes with a bigger price tag) to attract customers long term. How will this business ever become profitable before the bubble bursts?
We have some idea about the demand. As a data point, the cloud hyperscalers have in aggregate 2.3T in backlogs or “Remaining Performance Obligations” — i.e. payments committed by customers for remaining services provided over the next few years — of which about ~1T is from OpenAI and Anthropic. That still leaves 1.3T from other customers.
As another data point, these backlogs were at ~0.5T in 2022, before the AI boom started. Google for instance registered a 390% growth YoY in backlogs.
That’s not too far from the trillions being invested in the buildout.
I would say the irrationality of this buildout is exaggerated. It makes more sense to look at it as these companies making an existential bet on their futures.
That's a fair consideration. I look at it two ways, rationalizing and data:
Firstly, I would just ask, from a game theoretical perspective, why would these companies, all bitter competitors, most with very high fiscal discipline and absolute firehoses of double-digit billions of free cash flow per quarter pre-AI, actively do something that would be so destructive to themselves and the world economy?
Secondly, there are a ton of sources indicating significant growth and revenue coming from outside the AI industry. A few references:
1. The State of the AI Economy is the most thorough evaluation of deduplicated AI revenues I've found: https://www.exponentialview.co/p/the-state-of-the-ai-economy -- see especially the linked PDF report. I'm not a financial analyst, but they seem to cover all the bases. If nothing else, you should read this, it paints a very nuanced picture.
- https://www.genaiadoptiontracker.com/?tab=genAIwork&view=ove... - Toggle "Overall" and check the drop down for the variety of industries using GenAI; apparently 40% of "Agriculture, Forestry, Fishing, and Hunting" workers use GenAI. Also relevant: one tab over shows how much work hours they use AI for -- still only at ~5%.
2) In the slide on data center economics, the pricing calculations are based on an assumption of 50-75% gross margins. But token prices are collapsing faster than serving costs are dropping (https://www.accountingtoday.com/news/tokenomics-firms-see-su...). So margins are trending toward commodity levels. This is consistent with Anthropic's S1, which showed revenue growth slowing faster than cost growth.
>[W]hy would these companies, all bitter competitors, most with very high fiscal discipline and absolute firehoses of double-digit billions of free cash flow per quarter pre-AI, actively do something that would be so destructive to themselves and the world economy?
One might wonder the same thing about the 2008 GFC, the 2000 Internet bubble, the Great Depression, etc. Because their individual incentives are at odds with the global good. Let's just look at the role of private credit. Private credit made overzealous acquisitions during and after the pandemic, to the point that they are all sitting on depreciating assets while owing their investors outsized returns. It sort of made sense when prevailing interest rates were low, but it's untenable now. Investors were starting to ask for their money, and funds were having to shut their window and enforce their contractual withdrawal limits. Enter AI and (more importantly) data centers. This was a deus ex machina that allowed private credit to draw much-needed new capital from insurance trusts and pension funds. They've now made their overall financial position much worse, but have put off the day of reckoning into the future. It's the same reason people get in trouble with loan sharks, they just don't want to go through the pain of acknowledging the institutional loss, which could also mean realizing significant personal losses. (This also goes for some of the major AI players.) As long as they can use OPM to put off the pain, they will.
Good questions. Note that it's hard to tell what's happening from a single reference, and any single reference could be wrong, which is why I look at multiple sources including those I linked above to piece together the bigger picture. Here's what I see.
For 1) the report cannot determine an exact split between VC-backed startup vs other revenue because the startups are private. What it focuses on is deduping the revenues, so the headline number of $110B - 175B annually is the outside money flowing into the system, and not circular flows.
However that outside money includes revenues, VC-backed cash, and training costs, which cannot be disentangled by this methodology. But note:
a) That does not imply circularity, as the apollo.com link suggests.
b) We also can't assume this is mostly from burning investor cash rather than actual going concerns. Large VCs like A16Z, YC and Menlo Ventures, without disclosing margins, report seeing dramatic revenue growth in their startups, e.g.
So while often VCs are also investors in the frontier labs, the flow of money here is explicitly not "hyperscalers -> startups -> frontier labs -> hyperscalers," but rather "VCs + revenue -> startups -> frontier labs -> hyperscalers." And VCs report that revenue is the fastest growing in history, which is corroborated by all the other links measure AI adoption growth.
Furthermore, this $110 - 175B number also gels with the $177B number that the a16z link estimates.
For 2) the slides are much better for this (and the caveats are important) but the simplest I can put it is, at this point the marginal cost to serve tokens is far below and falling faster than token prices. Each new generation of hardware drops the cost to serve even further. As above, this rhymes with other sources including Anthropic's S-1.
Regarding your point about the 2008 GFC, the difference here is that their individual incentives are not at odds with the global good, because at this point it appears to be bad for everybody! In the GFC there was no real pain until the crash; here the pain is all around from the start. The hyperscalers are actively torching their extremely profitable cash flows (along with a lot of goodwill) and taking on heavy debt, despite being punished by the stock market, to fund this buildout that apparently everybody hates but can't stop using.
So we are not allowed to guess anything because not every facts are known ? Not nice try.
Fact: they are burning someone money. Or rather: they already burned a lot of money. For sure that was VC's and investors cash. Maybe some loans.
And now they trying to offload debt via IPO's.
Now: hardware side. Micron stopped producing what peoples wanted and still want to buy. So memory prices skyrocketed - llm industry own way of crafting inflation.
Nvidia sells and also buys and lots of promises circling around.
Oracle and Amazon buys at least from Nvidia and sells and take lots of promises too.
That influences markets and now IPO tries to cash on hype wave.
That's what is known so:
a) That does suggest circularity as some suspects
b) We know VC's cash is used for burning since few decades - they trow cash and dream about swans and unicorns :>
So who from "outside" throw money "into the system" ?? Certainly it not coming from selling inference.
All that is enough to start investigations.
Or you are trying to push: we don't know let's they do whatever they want with all that pensions invested ? Then I say: who pays you ?
> apparently everybody hates but can't stop using
Yep. llms momorization of knowledge is helpful for fast retreaving. But making it artificially scarcity and cashing on it is not good heart doing and nowhere close to goodwill. Especially in US making money is not a crime. But some schemes are.
There is a big question on the ROI of those customers. It's clear that the frontier labs are delivering something people care about, but how much do the tier 2 labs deliver that people are paying for now?
Of course there's still a big demand as there are massive plans on building AI DC all over the planet, but at the same time this are supposedly not an anomaly - wouldn't most customers eventually want newer gear (or reduction in cost)?
Before Tuesday the plan was to get a bailout from their buddies in the White House on "national security" grounds. That became much less viable now that Trump is a lame duck and anti-datacenter ballot measures passed everywhere, which is probably why e.g. CoreWeave is down 20% since the election.
Ouch, investment in GPU isn't a one off, you also need to invest newer generations (that naturally comes with a bigger price tag) to attract customers long term. How will this business ever become profitable before the bubble bursts?
You're somewhat assuming there aren't customers lined up wanting capacity.
To what extent is that demand from frontier labs? Purportedly, Anthropic is a significant chunk of GCP rev.
How much runway will the creditors give those labs until they expect a ROI?
The financial underpinning the AI build out make zero sense, but we've seen the market remain irrational for long periods of time.
We have some idea about the demand. As a data point, the cloud hyperscalers have in aggregate 2.3T in backlogs or “Remaining Performance Obligations” — i.e. payments committed by customers for remaining services provided over the next few years — of which about ~1T is from OpenAI and Anthropic. That still leaves 1.3T from other customers.
As another data point, these backlogs were at ~0.5T in 2022, before the AI boom started. Google for instance registered a 390% growth YoY in backlogs.
That’s not too far from the trillions being invested in the buildout.
I would say the irrationality of this buildout is exaggerated. It makes more sense to look at it as these companies making an existential bet on their futures.
You don't address circular financing at all.
That's a fair consideration. I look at it two ways, rationalizing and data:
Firstly, I would just ask, from a game theoretical perspective, why would these companies, all bitter competitors, most with very high fiscal discipline and absolute firehoses of double-digit billions of free cash flow per quarter pre-AI, actively do something that would be so destructive to themselves and the world economy?
Secondly, there are a ton of sources indicating significant growth and revenue coming from outside the AI industry. A few references:
1. The State of the AI Economy is the most thorough evaluation of deduplicated AI revenues I've found: https://www.exponentialview.co/p/the-state-of-the-ai-economy -- see especially the linked PDF report. I'm not a financial analyst, but they seem to cover all the bases. If nothing else, you should read this, it paints a very nuanced picture.
2. Enterprise surveys:
- https://knowledge.wharton.upenn.edu/special-report/2026-ai-a... (Has links to reports from prior years)
- https://menlovc.com/perspective/2026-the-state-of-enterprise... (Enterprise AI deployment spend up to 177B, 5x YoY)
- https://a16z.com/state-of-markets-ii/
- https://www.deloitte.com/us/en/what-we-do/capabilities/appli...
3. Government surveys:
- https://www.federalreserve.gov/econres/notes/feds-notes/moni...
- https://www.genaiadoptiontracker.com/?tab=genAIwork&view=ove... - Toggle "Overall" and check the drop down for the variety of industries using GenAI; apparently 40% of "Agriculture, Forestry, Fishing, and Hunting" workers use GenAI. Also relevant: one tab over shows how much work hours they use AI for -- still only at ~5%.
- https://ec.europa.eu/eurostat/statistics-explained/index.php...
- https://www.oecd.org/en/about/news/announcements/2026/01/ai-...
4. Payment processor data:
- https://stripe.com/guides/indexing-the-ai-economy
- https://ramp.com/data/ai-index (also here: https://fred.stlouisfed.org/blog/2026/10/which-businesses-ar...)
This link mentions a bunch of the above and more: https://www.ryzlabs.com/guides/enterprise-ai-adoption-statis...
There are many more of such sources out there, Google AI overview is pretty good at surfacing them.
A couple of questions regarding the "State of the AI economy"Report:
1) How much of that revenue growth is driven by VC-backed startups burning through investor cash rather than going concerns paying out of cash flow? Some reports suggest quite a bit (https://www.apollo.com/wealth/insights-news/insights/daily-s...)
2) In the slide on data center economics, the pricing calculations are based on an assumption of 50-75% gross margins. But token prices are collapsing faster than serving costs are dropping (https://www.accountingtoday.com/news/tokenomics-firms-see-su...). So margins are trending toward commodity levels. This is consistent with Anthropic's S1, which showed revenue growth slowing faster than cost growth.
>[W]hy would these companies, all bitter competitors, most with very high fiscal discipline and absolute firehoses of double-digit billions of free cash flow per quarter pre-AI, actively do something that would be so destructive to themselves and the world economy?
One might wonder the same thing about the 2008 GFC, the 2000 Internet bubble, the Great Depression, etc. Because their individual incentives are at odds with the global good. Let's just look at the role of private credit. Private credit made overzealous acquisitions during and after the pandemic, to the point that they are all sitting on depreciating assets while owing their investors outsized returns. It sort of made sense when prevailing interest rates were low, but it's untenable now. Investors were starting to ask for their money, and funds were having to shut their window and enforce their contractual withdrawal limits. Enter AI and (more importantly) data centers. This was a deus ex machina that allowed private credit to draw much-needed new capital from insurance trusts and pension funds. They've now made their overall financial position much worse, but have put off the day of reckoning into the future. It's the same reason people get in trouble with loan sharks, they just don't want to go through the pain of acknowledging the institutional loss, which could also mean realizing significant personal losses. (This also goes for some of the major AI players.) As long as they can use OPM to put off the pain, they will.
Good questions. Note that it's hard to tell what's happening from a single reference, and any single reference could be wrong, which is why I look at multiple sources including those I linked above to piece together the bigger picture. Here's what I see.
For 1) the report cannot determine an exact split between VC-backed startup vs other revenue because the startups are private. What it focuses on is deduping the revenues, so the headline number of $110B - 175B annually is the outside money flowing into the system, and not circular flows.
However that outside money includes revenues, VC-backed cash, and training costs, which cannot be disentangled by this methodology. But note:
a) That does not imply circularity, as the apollo.com link suggests.
b) We also can't assume this is mostly from burning investor cash rather than actual going concerns. Large VCs like A16Z, YC and Menlo Ventures, without disclosing margins, report seeing dramatic revenue growth in their startups, e.g.
https://a16z.com/revenue-benchmarks-ai-apps/
https://www.cnbc.com/2025/03/15/y-combinator-startups-are-fa...
So while often VCs are also investors in the frontier labs, the flow of money here is explicitly not "hyperscalers -> startups -> frontier labs -> hyperscalers," but rather "VCs + revenue -> startups -> frontier labs -> hyperscalers." And VCs report that revenue is the fastest growing in history, which is corroborated by all the other links measure AI adoption growth.
Furthermore, this $110 - 175B number also gels with the $177B number that the a16z link estimates.
For 2) the slides are much better for this (and the caveats are important) but the simplest I can put it is, at this point the marginal cost to serve tokens is far below and falling faster than token prices. Each new generation of hardware drops the cost to serve even further. As above, this rhymes with other sources including Anthropic's S-1.
Regarding your point about the 2008 GFC, the difference here is that their individual incentives are not at odds with the global good, because at this point it appears to be bad for everybody! In the GFC there was no real pain until the crash; here the pain is all around from the start. The hyperscalers are actively torching their extremely profitable cash flows (along with a lot of goodwill) and taking on heavy debt, despite being punished by the stock market, to fund this buildout that apparently everybody hates but can't stop using.
Niiiiice joke :)
> along with a lot of goodwill
One word: Oracle :))
Or two more: Google Microsoft :))
So we are not allowed to guess anything because not every facts are known ? Not nice try.
Fact: they are burning someone money. Or rather: they already burned a lot of money. For sure that was VC's and investors cash. Maybe some loans.
And now they trying to offload debt via IPO's.
Now: hardware side. Micron stopped producing what peoples wanted and still want to buy. So memory prices skyrocketed - llm industry own way of crafting inflation.
Nvidia sells and also buys and lots of promises circling around.
Oracle and Amazon buys at least from Nvidia and sells and take lots of promises too.
That influences markets and now IPO tries to cash on hype wave.
That's what is known so:
a) That does suggest circularity as some suspects
b) We know VC's cash is used for burning since few decades - they trow cash and dream about swans and unicorns :>
So who from "outside" throw money "into the system" ?? Certainly it not coming from selling inference.
All that is enough to start investigations.
Or you are trying to push: we don't know let's they do whatever they want with all that pensions invested ? Then I say: who pays you ?
> apparently everybody hates but can't stop using
Yep. llms momorization of knowledge is helpful for fast retreaving. But making it artificially scarcity and cashing on it is not good heart doing and nowhere close to goodwill. Especially in US making money is not a crime. But some schemes are.
Nah
There is a big question on the ROI of those customers. It's clear that the frontier labs are delivering something people care about, but how much do the tier 2 labs deliver that people are paying for now?
Of course there's still a big demand as there are massive plans on building AI DC all over the planet, but at the same time this are supposedly not an anomaly - wouldn't most customers eventually want newer gear (or reduction in cost)?
At what price? I would have unlimited customers lined up selling dollars for 50ct.
but..... there aren't
Before Tuesday the plan was to get a bailout from their buddies in the White House on "national security" grounds. That became much less viable now that Trump is a lame duck and anti-datacenter ballot measures passed everywhere, which is probably why e.g. CoreWeave is down 20% since the election.
Given the average bubble-prediction accuracy, there might as well be a few centuries before it bursts. Which would give some time.
"UUUAAAAHHHH!!!"
That was the sound of Ed Zitron ghost....
This is a nothingburger. Microsoft declining to renew a deal for 2023 A100s that power Bing because they want newer chips in house.
Oracle can re-rent them at spot at a higher price anyway.
Apparently a lot of market participants are too scared to be able to read today.