Microsoft CEO Satya Nadella Says He Would Lease SpaceX Compute Capacity If It Were Available
Key Takeaways
- •Microsoft CEO Satya Nadella stated he would lease SpaceX's compute capacity for data centers if it became available, but the remark was conditional, with no contract, price, or start date announced.
- •Exploratory talks between SpaceX and Microsoft in the summer of 2026 reportedly covered potential leases of up to 3 gigawatts of compute, an amount exceeding SpaceX's 1.7 GW installed base as of September 2026.
- •SpaceX's AI and compute leasing business reached an annualized revenue run rate of approximately $41.1 billion by September 2026, driven largely by long-term contracts with customers including Anthropic and Google.
- •On November 12, 2026, SpaceX announced plans for more than $74 billion in new AI supercomputing campuses, with $24 billion in Texas and $50 billion in Arizona, and anticipates installed capacity could reach potentially 10 GW by the end of 2027.
- •Microsoft reported on its October 28 earnings call that it added 24 new data centers across five continents in the third quarter of 2026, reflecting its continued investment alongside openness to leasing external AI capacity.

Microsoft runs one of the largest cloud businesses in the world, and its chief executive now says he would be prepared to rent computing power from a rocket company. In remarks underscoring how intense competition for artificial intelligence infrastructure has become, Microsoft CEO Satya Nadella said he would lease SpaceX's compute capacity for data centers if it were available.
The comment arrives as demand for AI compute continues to outrun what even the largest builders can bring online, and as SpaceX scales its AI unit into a major provider of leased computing capacity. In that environment, capacity that already exists—rather than capacity still under construction—carries particular weight for cloud providers trying to serve AI workloads today.
What Nadella said—and the talks behind it
The remark was conditional. Nadella did not announce a contract, a price, or a start date; he signaled only that if SpaceX had spare capacity to offer, Microsoft would be a willing tenant.
That willingness has some history behind it. Exploratory talks between SpaceX and Microsoft reportedly took place in the summer of 2026, covering potential leases of up to 3 gigawatts (GW) of compute. A gigawatt—one billion watts, or 1,000 megawatts—has become the industry's yardstick for sizing AI data centers, because the power available to a campus determines how much computing it can run. No formal agreement has been signed, however, leaving the situation one of strong mutual interest rather than a done deal.
Nadella's openness to outside capacity fits a broader pattern. He has previously expressed a willingness to lease external AI computing capacity as demand for those resources spikes, rather than insisting that Microsoft build every server it needs. At the same time, the company has stayed cautious about certain rental models in the market. The distinction matters: with AI demand outpacing construction timelines across the industry, the option to rent capacity that already exists gives cloud operators a second lever alongside their own buildouts.
Nor is Microsoft standing still. On its October 28 earnings call, the company said it added 24 new data centers across five continents in the third quarter of 2026.
SpaceX, the unexpected compute landlord
If the idea of SpaceX acting as a data center provider sounds unusual, the company's financials make it less so. SpaceX's AI unit has been expanding rapidly since integrating xAI—the artificial intelligence company founded by Elon Musk, who also founded SpaceX—and its AI and compute leasing business reached an annualized revenue run rate of approximately $41.1 billion by September 2026. A run rate projects current revenue over a full year, so the figure reflects momentum rather than a booked annual total.
Much of that growth comes from long-term contracts with major customers, including Anthropic and Google, which are paying substantial monthly fees.
Capacity, rather than demand, is the constraint. SpaceX had 1.7 GW of installed compute at the end of September 2026 and targets exceeding 2 GW by the end of the year. The 3 GW reportedly discussed with Microsoft would exceed SpaceX's entire installed base as of September—a gap that helps explain the "if it were available" framing of Nadella's comment.
SpaceX is working to close that gap quickly. On November 12, 2026, the company announced plans for more than $74 billion in new AI supercomputing campuses—$24 billion in Texas and $50 billion in Arizona—and anticipates that installed capacity could jump from 1.7 GW to potentially 10 GW by the end of 2027. The talks with Microsoft remain exploratory for now, and the near-term markers include whether SpaceX hits its year-end capacity target and how quickly the new Texas and Arizona campuses come online.