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?
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.
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)?
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.
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?
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?
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.
You're somewhat assuming there aren't customers lined up wanting capacity.
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)?
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.
Nah
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.
At what price? I would have unlimited customers lined up selling dollars for 50ct.
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?
"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.