Microsoft's reported 38 GW fleet implies thirteen Pecos-scale campuses by 2032

The additions come to about 26 gigawatts, and the campus at Pecos, Texas is the only piece Microsoft has described in detail ^[6]. Announced on June 22nd, it is expected to provide approximately 2 gigawatts ^[9]. Microsoft called Pecos one of the largest single capacity additions in its history and said the multibillion-dollar investment would span five to seven years ^[10]. Divide the additions by that unit and you get thirteen Pecos-scale campuses ^[1]. Reuters puts it less precisely: several additions on that scale, plus leased sites and smaller campuses ^[11].

The mix moves faster than the total. Overall capacity grows about 3.2 times ^[7]. AI-specific capacity goes from about 2 gigawatts to about 12.7 ^[4][5], which is 10.7 gigawatts to add ^[2] at a multiple of roughly 6.4 ^[4]. Everything else in the fleet grows from about 10 gigawatts to about 25.3, or about 2.5 times ^[3].

Satya Nadella gave a different unit on the July 29th earnings call: 88 data centers added during fiscal 2026, 31 of them in the final quarter ^[14]. That leaves 57 across the first three quarters, about 19 a quarter ^[5]. It is a facility count, and the reported target is denominated in gigawatts, so neither figure converts into the other.

Plans at this scale have been unwound before. That cancelled capacity, about 2 gigawatts, is the same size as the Pecos campus ^[7]. Reuters cited TD Cowen analysts, who attributed the pullback to oversupply relative to Microsoft's own demand forecast ^[13].

The demand evidence is present-tense. Commercial remaining performance obligations reached $678 billion in the fiscal 2026 fourth quarter, up 84% year over year, with roughly 30% expected to be recognized as revenue within 12 months ^[15]. That is about $203 billion inside the year and about $475 billion booked beyond it ^[6]. Microsoft 365 Copilot passed 30 million paid seats, and net paid-seat additions more than doubled from the preceding quarter ^[16]. None of it establishes how much capacity customers will consume in 2032, or how much of the reported fleet Microsoft will build, lease, equip and place into service ^[17].

For the 38 gigawatts to be usable in a multi-year capacity plan, three things would have to be true that the report does not establish: a per-year addition curve, a region breakdown, and a split between owned and leased sites. Reuters notes that a site under consideration, a signed lease, an equipped building and a data center serving paying customers are different levels of commitment, and the reported target does not disclose how much capacity has reached any of them ^[18].