SpaceX is transforming into a hyperscale computing provider at a pace that has surprised the market.
Research firm SemiAnalysis highlights in a recent report that SpaceX could add more than 10GW of computing capacity by the end of 2027, potentially generating up to $300 billion in annual recurring revenue (ARR), with Microsoft emerging as its largest client.
Previously, during SpaceX's first earnings call, Elon Musk stated a "conservative" target of building and delivering 6-8GW of incremental computing capacity in a single year by 2027, with an upside potential exceeding 10GW. Based on $50 billion in capital expenditure per GW, this implies a 2027 capital expenditure range of $300 billion to $500 billion, rivaling the expected investments of AWS and Google—a staggering figure for a company with profitability far below that of mainstream hyperscale cloud providers.
SemiAnalysis notes that it has evaluated all suitable sites for SpaceX and tracked available gas turbine equipment on a quarterly basis, concluding that the targets are realistic. The report also points out that Microsoft has signed over 10GW in data center contracts year-to-date, with total contract value exceeding $300 billion. The potential demand for SpaceX's computing power could be a key variable in boosting Microsoft's Azure revenue growth from roughly 42% to over 100%.
The high profit margins of inference computing are driving a surge in demand.
SemiAnalysis's core argument is based on the high profitability of AI inference. According to its Tokenomics model and inference simulator, when OpenAI and Anthropic offer API inference services on GB300 clusters, each GW can generate over $100 billion in revenue annually. At a conservative lease price of $3 per GPU per hour, the cost per GW per year is about $12 billion, resulting in gross margins exceeding 60% for inference, with some flagship models reaching over 85%.
This economic model also applies to Microsoft. SemiAnalysis points out that the renegotiated OpenAI agreement in April 2026 removed the original 20% revenue-sharing clause, meaning Microsoft can achieve similar per-MW revenue and margins to OpenAI and Anthropic when offering inference services using OpenAI models, without bearing any training costs. The report characterizes this situation as a "once-in-a-century opportunity."
Microsoft's computing shortfall opens a window for SpaceX.
Microsoft had slowed down data center leasing activities significantly in late 2024, but SemiAnalysis states this trend reversed completely between 2025 and 2026. In October 2025, Microsoft signed a $250 billion infrastructure-as-a-service agreement with OpenAI, which SemiAnalysis estimates corresponds to about 7GW of computing power, leaving Microsoft severely constrained in its Foundry API business and applications like Copilot. To fill this gap, Microsoft has signed over 10GW in binding contracts year-to-date across leasing, self-building, and long-term power purchase agreements. SemiAnalysis believes that a $1.5 trillion contract between Microsoft and SpaceX for 3GW is "not impossible," citing two reasons: first, Microsoft is already in a massive computing expansion cycle; second, SpaceX offers a 90-day cancellation clause, consistent with its contracts with Anthropic and Google, making it a low financial risk and easy to secure internal approval.
SpaceX's construction speed is disrupting industry norms.
SemiAnalysis attributes SpaceX's competitive advantage to a systematic disruption of traditional data center construction logic. The report lists several specific examples: Colossus 1's 300MW facility was completed in 122 days; the Southaven power plant expanded from 27 gas turbines (about 495MW) in February 2026 to 69 turbines (1.7GW) by July 2026; the "MiniHard" project began vertical construction in March 2026 and is expected to reach 450-500MW in about five months.
In addressing supply chain bottlenecks, SpaceX has adopted unconventional strategies: replacing large power transformers with a two-year backlog using Chinese-made power modules; purchasing gas turbines on the secondary market, including those originally destined for an Oracle site in New Mexico; and compressing commissioning cycles through highly parallelized construction and pre-assembly. Colossus 2 peak daily labor was about 3,000 people, significantly lower than the industry standard for a data center of similar size. SemiAnalysis argues that SpaceX's core strength lies in prioritizing "speed over efficiency," a common industry logic. In an era of extreme computing scarcity and high AI inference margins, a 500MW cluster that can be delivered in three months with a 90-day cancellation clause is one of the most scarce assets in the market. Google's decision to ultimately sign a contract with SpaceX is a testament to this logic.
Financing paths: Nvidia supplier financing and cash flow rollover.
Addressing concerns about SpaceX's capital strength, SemiAnalysis proposes two financing paths. First, Nvidia could reduce SpaceX's upfront cash outlay through supplier financing—a key reason why Musk announced exclusive use of Nvidia chips during the earnings call, despite SpaceX having previously evaluated alternatives like TPUs and AMD. Second, thanks to the industry's fastest delivery cycle, SpaceX can charge a premium of $30 million to $50 million per MW per year, allowing capital expenditure to be recovered through operating cash flow in less than a year. Combining these factors, SemiAnalysis predicts that, assuming only 50% of SpaceX's 2027 incremental computing power is used for commercial inference monetization (with the rest reserved for training Grok and Cursor teams), SpaceX's ARR could reach $300 billion by the end of 2027.
Comments