Bill Ackman filed a 13F showing Pershing Square Holdings began accumulating Microsoft in February 2025, entering the position as AI infrastructure stocks corrected on valuation fears and DeepSeek narrative contagion. The stake, now valued at roughly $2.8 billion, represents Ackman's first significant tech exposure since his 2020 Netflix exit. He bought while the Nasdaq corrected 7.4% and Microsoft traded below $390, a discount the firm classified as temporary mispricing.
The February window was deliberate. AI stocks had rotated hard on DeepSeek's December inference cost claims, which traders misread as demand destruction for hyperscale compute. Microsoft shed $210 billion in market cap across January and February before stabilizing. Ackman's entry came as Azure growth deceleration fears peaked, despite the unit posting 31% revenue growth in its January earnings print. The 13F shows Pershing built the position across three tranches, the largest coming in mid-February when the stock briefly touched $385. No options overlay disclosed.
The timing matters because Ackman rarely touches software infrastructure. His portfolio historically skews toward consumer brands with pricing power and regulated monopolies with visible cash flows. Microsoft breaks that mold but fits a secondary thesis Pershing has telegraphed since Q4 2024: AI capex winners will be the hyperscalers controlling inference deployment, not the chipmakers or model labs. The firm's Q1 letter flagged Azure's 53% gross margin on AI workloads as structural, not cyclical. That margin profile, combined with enterprise AI seat expansion through Copilot, gave Ackman a rare software bet with consumer-like unit economics.
The position also syncs with power infrastructure pressure points surfacing this week. NextEra's $67 billion Dominion Energy acquisition, announced Monday, underscores the utility-scale power bottleneck facing AI data center expansion. Microsoft has 37 gigawatts of power commitments locked through 2030, double Amazon's disclosed capacity. Ackman's bet is less about model innovation and more about Azure becoming the only hyperscaler with sufficient power headroom to absorb enterprise inference demand through 2027. The February entry captured that optionality before the NextEra deal repriced power scarcity into the stock.
Allocators should track Microsoft's April earnings for Azure consumption growth and any updates to its $80 billion fiscal 2025 capex guidance. If Azure posts above 32% growth and maintains AI workload margins, the position likely expands. Watch for Pershing's Q2 letter in early May for language around hyperscaler power dynamics—Ackman rarely discloses without a follow-on thesis. The NextEra-Dominion merger, expected to close Q3 2025, will clarify whether Microsoft's power commitments translate to pricing power in inference hosting contracts.
Microsoft's Q1 capex was $22.6 billion, up 79% year-over-year, all directed toward AI infrastructure with locked power. The market now prices that spend as strategic moat, not speculative deployment.