Key Takeaways
- • Bill Ackman's Pershing Square Q3 letter, released Nov. 2, argues that permitting, power, chip supply, capital and political considerations will slow new AI data center builds and favor Amazon, Microsoft, Alphabet and Meta Platforms, according to Motley Fool reporting on Yahoo Finance.
- • Ackman wrote that the current compute shortage "will persist for longer" if these hurdles slow rivals, and he placed neoclouds and other secondary participants on the weaker side of the trade.
- • The source establishes Ackman's thesis only; it does not show that restrictions have already produced measurable financial gains for the four hyperscalers.
- • Motley Fool reports the four hyperscalers trade at forward P/E ratios between 21 and 28, while neoclouds trade at enterprise-value-to-sales multiples only slightly below the hyperscalers, an opinion the author argues should be wider given regulatory risk.
- • IDC reports worldwide AI infrastructure spending reached $82.0 billion in Q2 2025, up 166% year over year, with hyperscalers, cloud providers and digital service providers representing 86.7% of spending.
Summary
In the letter, Ackman wrote that "standing up new data center capacity requires permitting, power, chip supply, and the navigation of political considerations" in addition to capital. He argued that if these challenges slow other companies' capacity additions, "the current shortage of compute will persist for longer," with benefits accruing to companies that already have capacity and the means to expand it. He also said that "the neoclouds and other secondary participants face meaningful challenges with each of these requirements." Motley Fool notes that Pershing Square owns stock in the hyperscalers, and that Elon Musk expects SpaceX to add 10 gigawatts of data center capacity next year versus 1.7 gigawatts at the end of Q3, a plan that would run into each constraint Ackman lists.
The political backdrop has intensified. Motley Fool reports that many observers read the midterm results as a referendum on data center buildouts, with local ballot measures aimed at blocking or limiting development. Power and chip constraints are longstanding: operators are pursuing their own electricity because grid connections take years, and SK hynix has reportedly committed more than half of its 2027 and 2028 HBM4 capacity, leaving smaller buyers to wait or pay more.
Demand data supports the thesis that supply will remain tight. McKinsey projects AI capacity will rise 3.5 times from 2025 to 2030 and account for roughly 70% of total data-center capacity demand by 2030, with global data-center demand rising from 82 to 219 gigawatts. IDC expects global AI infrastructure spending to reach $758 billion by 2029.
The market has not clearly rewarded the setup. Motley Fool says the stocks have been held back by concerns that capital spending will not earn strong returns, and an Investor's Business Daily report said Microsoft, Google and Amazon fell after the election. Ackman's thesis also implies a nuanced position for buyers: tight capacity may bring longer waits and firmer pricing, making workload diversification across providers a practical hedge. Analysts should note that the four companies differ in model, since Amazon, Microsoft and Alphabet sell capacity to outside customers while Meta mainly builds for its own services.
Go deeper than the headline
Research every firm in this story, track the ones that matter and get the daily briefing — free for 7 days.