Thesis · Hyperscalers, Cloud & Compute Services
Will aggregate 2027 capex guidance from Alphabet, Amazon, Microsoft and Meta imply year-over-year growth of less than 15%?
Cash constraints force capex growth to decelerate toward the pace of operating cash flow, with 2027 guidance implying under 15% aggregate growth.
Who is on each side
The two sides.
Each side in DeepStack’s words, then the voices who argue it, attributed and paraphrased, dated where the date is known. A voice links to its page on DeepStack where one exists.
The case for yes
Capex growth plateaus under cash constraints
Free cash flow approaching zero removes the capacity for continued aggressive capex escalation. Speculative optionality in AI stocks rests on capex continuing to rise, so when hyperscalers hit physical cash constraints, capex growth must plateau or reverse, with aggregate 2027 guidance implying under 15% growth over 2026 actuals.
The case for no
Capex growth stays at or above 15% as funding keeps pace
Usage deepens the ecosystem and makes the leader harder to displace, and because marginal serving cost is small once models exist, none of the majors can afford to be the one that waits, so spending stays aggressive. Executives who understand that failing to cannibalize their own business invites disruption keep spending heavily on AI infrastructure regardless of near-term earnings optics. Firms that can fund this growth internally rather than depending on external capital markets gain an extended runway that keeps the buildout pace intact. On this view aggregate 2027 capex guidance from Alphabet, Amazon, Microsoft and Meta implies year-over-year growth of 15% or more.
1 voice
Capex growth plateaus under cash constraints
Speculative optionality in AI stocks rests on the assumption that capex will continue rising, allowing allocation to peripheral technologies. When hyperscalers hit physical cash constraints, capex growth must plateau or reverse. This removes the…
4 voices
Capex growth stays at or above 15% as funding keeps pace
Usage feeds the ecosystem and makes the leader harder to displace, while marginal serving cost is small once models exist, so spending early is how a durable position is bought and none of the majors can afford to be the one that waits.
Because today's tech leaders understand that failing to cannibalize their own business invites disruption by new entrants, they are willing to spend heavily on AI regardless of when returns show up in earnings, unlike incumbents 30 years ago who…
Hf Best Ideas
The compiled letter argues internal cash flows eventually cannot fund the compute buildout, so firms turn to private credit; Brookfield's $200 billion five-year commitment covering only about 6 gigawatts is read as a sign the cycle is early, with…
When companies can fund growth internally rather than depending on external capital markets, they survive downturns that would have killed earlier-stage companies. This extended runway gives the technology time to develop applications, build…
What would settle it
The dated tests.
Each test names the public source that settles it, the rule and the day. The Call’s number, the readers’ median and the machines’ numbers stand beside it, as on the Docket.
CoreWeave, purchases of property and equipment as a share of revenue, quarter ended September 30, 2026: whether it comes in at or above 200%.
On the Docket Pendingth027-t01Open question
Specification pending
The machinesClaude 60% · DeepSeek 87%
Your numberSign in by email to forecast
CoreWeave, purchases of property and equipment as a share of revenue, quarter ended September 30, 2026: whether it comes in at or above 230%.
On the Docket Pendingth027-t02Open question
Specification pending
The machinesClaude 40% · DeepSeek 60%
Your numberSign in by email to forecast
CoreWeave, purchases of property and equipment as a share of revenue, quarter ended September 30, 2026: whether it comes in at or above 260%.
On the Docket Pendingth027-t03Open question
Specification pending
The machinesClaude 30% · DeepSeek 35%
Your numberSign in by email to forecast
Oracle, purchases of property and equipment as a share of revenue, quarter ended November 30, 2026: whether it comes in at or above 120%.
On the Docket Pendingth027-t04Open question
Specification pending
The machinesClaude 33% · DeepSeek 45%
Your numberSign in by email to forecast
Oracle, purchases of property and equipment as a share of revenue, quarter ended November 30, 2026: whether it comes in at or above 145%.
On the Docket Pendingth027-t05Open question
Specification pending
The machinesClaude 6% · DeepSeek 17%
Your numberSign in by email to forecast
Oracle, purchases of property and equipment as a share of revenue, quarter ended November 30, 2026: whether it comes in at or above 165%.
On the Docket Pendingth027-t06Open question
Specification pending
The machinesClaude 7% · DeepSeek 5%
Your numberSign in by email to forecast
In The Buildout
The Parts on the same question.
The stories that take up this thesis, with their own Calls and tests.
Part 04: AI Spending Has Met Its First Real Bill
AI’s buildout is entering its first financing test: spending outruns cash flow, investors have separated the winners from the believers, and the next budget will expose who is right.
Part 14: The Cash Arrives Unevenly
Microsoft pays for its data centers from its own cash with room to spare. Alphabet and Meta now spend most of theirs and Amazon more than all of it; Alphabet and Amazon have covered the gap with stock and bonds. DeepStack reads the filings company by company and names the October 28 report that tests the view.
Part 11: The Backlog Is Not the Quarter
AI suppliers report record order books. DeepStack reads the near book as revenue, the far book as a promise and the margin as the number that arrives last.
A thesis page states one question and both sides of it; DeepStack takes no side in words, and its view is the number it gives each test before the date. Analysis and opinion, not investment advice. Voices are paraphrased from public writing and attributed by name; DeepStack shows no score, ranking or accuracy for any person. Pre-specified means the words were approved on or before the test’s date. Specified after the date means they were written afterwards: such a test is counted and shown, and never rated. An open question has no words yet. The Call’s number is the chance, from 0 to 100, that the Call gives the yes answer to a test’s question, recorded before the test’s date; it is DeepStack’s own reading, not a market price and not advice. The readers’ mark on the same line is the readers’ median, the middle of the numbers signed-in readers gave the test, shown from 25 readers. The machines: Each model read this page and nothing else. The number that counts is the latest one each model gave before the test closed.