Key Takeaways
- Gartner forecast that more than 80% of neoclouds will stop operating independently by 2030, and CoreWeave fell 25% last week.
- Hyperscalers that pay for AI capacity from their own cash flow, led by Amazon and Alphabet, stand to win as credit gets tighter for indebted neoclouds.
- Amazon’s $140 billion of operating cash flow covered its $132 billion of capex in fiscal 2025, while CoreWeave’s total debt reached $29.8 billion.
- Hyperscaler earnings in the coming weeks will test whether sold-out capacity is showing up in results.
Gartner just put a number on the AI infrastructure shakeout, and CoreWeave (CRWV) fell 25% last week.
On Oct. 14, Gartner [forecast](LINK NEEDED) that more than 80% of neoclouds will cease independent operations by 2030. Its analyst Will Sommer said spending growth on AI-optimized infrastructure “will be too low for most AI compute providers to recover their AI investments.”
Paul Kedrosky, a partner at SK Ventures, summed it up on Prof G Markets: “Gartner came out with an analysis. They estimated that 80% of neoclouds will be out of business by 2030.”
Someone still has to own the compute the world is short of. And the market has already picked who.
Daniel Niles, founder of Niles Investment Management, pointed out on X that Microsoft (MSFT), Alphabet (GOOG) and Amazon (AMZN) are up 15.6% from their July 29 closing low. Over the same stretch, IREN (IREN), CoreWeave and Nebius (NBIS) are up 1.6%.
That’s a 14-point gap in less than three months.
Debt is the dividing line
Kedrosky’s case against the neoclouds comes down to their debt. He pointed to debt trading at “say 80 cents on the dollar,” with maturities in 2032, and said what he thinks the bond market is telling you: “this stuff gets restructured. You’re not getting a dollar.”
CoreWeave’s total debt went from $2 billion in fiscal 2023 to $29.8 billion in fiscal 2025, nearly 15x in two years.
Nebius’s balance sheet is in better shape, as TIKR covered when it fell 12% in a day on CoreWeave’s debt scare. But investors are selling it along with CoreWeave anyway.
Ted Mortonson, technology desk strategist at Baird, told Bloomberg Surveillance that his clients are moving “to anything with strong free cash flow,” and that “The names that don’t have free cash flow like Oracle or NBIS or CoreWeave in this area is just being thrown out at this point.”
Amazon is sold out, and paying its own way
Meanwhile, the hyperscalers can’t build fast enough. On All-In, David Sacks described a conversation with Amazon CEO Andy Jassy: “I said, yeah, we’re out of everything until 2028. He’s like, oh, it’s beyond 2028.”
Even Elon Musk can’t get capacity. He shared a post on what it’s like trying to rent AI compute, then added that it “describes the conversation SpaceX has had when we have tried to rent inference compute from AWS, GCP or anyone.”
And Amazon pays for its buildout out of its own pocket. Its cash from operations hit $140 billion in fiscal 2025, more than covering $132 billion of capex. That capex has more than doubled since fiscal 2021, when it was $61.1 billion.
(In fiscal 2021 and 2022, Amazon spent more on capex than its operations brought in.)
Alphabet’s margins hold up
Alphabet pays for its compute the same way, and its profitability shows it. Its operating margin rose from 26.5% in fiscal 2022 to 32.0% in fiscal 2025.
Scarce compute also helps on pricing. Mortonson put it bluntly: “The hyperscalers can charge anything they want.”
Niles expects the squeeze on “the more marginal neo-clouds” to “benefit the hyper-scalers who do not have the same funding issues and help with the rate of decline in token prices.” Put differently, AI prices should fall more slowly.
Stranded contracts cut both ways
The hyperscalers aren’t fully protected. They rent neocloud capacity themselves, so a neocloud that goes under could strand a contract they’re counting on.
Gartner’s forecast is about neoclouds ceasing independent operations, though. That covers restructurings and sales as well as shutdowns, and Kedrosky expects restructuring. The chips and data centers don’t disappear when the equity does. This is pure speculation, but with Amazon sold out past 2028, the hyperscalers look like the buyers with cash for those assets.
That’s also why Meta Platforms (META) belongs on the winners’ list, even though it doesn’t rent compute to anyone. Its ad business brings in the cash to keep building its own data centers, so tight credit doesn’t slow it down.
Amazon and Alphabet come out ahead
If Gartner is even half right, the winners are the companies that pay for compute out of cash flow. For me, Amazon and Alphabet top that list: they have sold-out capacity and profitable core businesses to fund more of it. Microsoft makes Niles’s list for the same reason, and Meta gets there as a buyer.
Of course, the hyperscalers report earnings in the coming weeks, and their results have to show that sold-out capacity. Niles is watching total token spend and token volumes, which both hit new highs last week on the Ramp Index.
So what is Amazon stock actually worth?
TIKR lets you forecast the future price of any stock in less than a minute. Just enter a few assumptions into TIKR’s valuation model and see what Amazon could be worth. Start from Wall Street consensus estimates, or adjust the inputs to reflect your own view of the business. It’s free to use.
Disclaimer:
Please note that the articles on TIKR are not intended to serve as investment or financial advice from TIKR or our content team, nor are they recommendations to buy or sell any stocks. We create our content based on TIKR Terminal’s investment data and analysts’ estimates. Our analysis might not include recent company news or important updates. TIKR has no position in any of the stocks mentioned. Thank you for reading, and happy investing!
