Pershing Square Models 15% to 25% Returns on AI Data Centers

Pershing Square argues that Wall Street’s estimates may understate the potential payoff from AI data-center spending at major technology companies, because forecasts capture infrastructure costs more clearly than possible revenue and earnings.

CEO William A. Ackman and Chief Investment Officer Ryan Israel laid out the case in a Nov. 2 shareholder letter. Pershing Square said its analysis covered Microsoft, Meta, Amazon and Google and used consensus estimates dated Sept. 30, 2026.

The firm estimated the four companies’ combined capital expenditures at about $140 billion in 2023 and about $740 billion in 2026, using company guidance. It put consensus spending for 2027 above $1 trillion. Pershing Square’s argument is that forecasts account for the scale of those outlays while giving less weight to the earnings capacity of the assets they fund.

The firm’s example starts with one gigawatt of AI data-center capacity. It assumes a $40 billion to $50 billion build cost and $15 billion to $20 billion in annual revenue. At a 50% incremental operating margin, those figures imply $7.5 billion to $10 billion in incremental operating income and a 15% to 25% pretax return on invested capital.

If future growth spending earns the modeled return, Pershing Square estimated, annual revenue growth could exceed consensus by five to 10 percentage points through 2029. After three years, revenue could be more than 15% and up to 35% above consensus estimates, the firm projected.

Amazon Web Services (AWS) revenue reached $45.984 billion in the quarter ended Sept. 30, up 39.3% from a year earlier. Amazon CEO Andy Jassy said the company expected AWS growth above 40% in both 2027 and 2028. At the same time, Amazon reported a $42.4 billion free-cash-flow outflow over the trailing 12 months. The company linked the outflow mainly to increased property and equipment purchases and said AI investment was the main reason for those purchases.

Pershing Square cautioned that its projections are estimates that may prove inaccurate and that actual returns may differ.