Elon Musk’s latest AI pitch is giving SpaceX and Microsoft a jolt in premarket trading, as investors weigh whether a sprawling new compute buildout could turn into one of the market’s most lucrative infrastructure deals. SpaceX rose 2.4% to $136.48 in Monday premarket trading, while Microsoft was fractionally lower at $498.47, after a fresh SemiAnalysis note argued Musk’s ambitions deserve to be taken seriously. The core bet is simple and enormous: if SpaceX can add compute fast enough, Microsoft could be in position to monetize the capacity and do it at scale.
The setup traces back to Musk’s first earnings call as a public company, where he outlined a plan to expand AI compute capacity from about 1.4 GW to 6–8 GW next year, with the possibility of topping 10 GW by the end of 2027. That kind of jump would put SpaceX in the rarefied territory of giant industrial buildouts, not just another AI infrastructure story. It also explains why the stock reaction is showing up now, even before any formal deal announcement. Traders are not waiting for a press release to start pricing the possibility.
SemiAnalysis, the San Francisco-based research boutique, published a Substack post on August 7, 2026, saying the targets look plausible. The report, written by Jeremie Eliahou Ontiveros, Rey Knuhtsen and Jordan Nanos, did not frame Musk’s plan as fantasy. Instead, it treated the numbers as something SpaceX could actually attempt if the company keeps moving at the speed Musk has promised. The firm’s view matters because it translates a headline-grabbing vision into a financing and capacity question that Wall Street can model.
The scale is where the story gets serious. SemiAnalysis said a 10 GW buildout would require $300 billion to $500 billion in capital spending by the end of 2027. That is an extraordinary sum by any standard, but the research note argues the economics can work because large-scale near-term compute is priced at a premium of about $50 billion per GW per year, while frontier AI labs can generate about $100 billion per GW in annual recurring revenue. In other words, the buildout is expensive, but the revenue profile may be even bigger.
The Microsoft angle is what gives the thesis its market punch. SemiAnalysis said Microsoft holds a 27% stake in OpenAI and renegotiated its deal in April 2026 to eliminate a 20% revenue share, which could let it monetize SpaceX compute at similar margins without paying training costs. That matters because Microsoft does not need to own the hardware outright to benefit from it. If SpaceX can deliver the infrastructure and Microsoft can sell access to it through Azure, the software giant can capture upside without bearing the full burden of building and training the models itself.
SemiAnalysis went further and described a potential 3 GW, $150 billion contract between Microsoft and SpaceX. It said that deal could push Azure revenue growth from about 42% to more than 100% in 2027. The firm’s research team put the opportunity bluntly: “Put simply, Microsoft has a giant incentive to procure as many megawatts as possible, as fast as possible. The potential impact is Microsoft Azure accelerating revenue growth from around 42% at present to in excess of 100% in 2027.” That is the kind of line that can move a stock, especially when it suggests growth acceleration rather than just steady demand.
The thesis is not only about demand. It is also about who can build the capacity fastest. SemiAnalysis said SpaceX has already shown an ability to move quickly through Colossus 1, a 300 MW Memphis cluster that went from scratch to operation in 122 days. The group also cited the Southaven plant expansion, which it said grew from 27 to 69 gas turbines and expanded from about 495 MW to more than 1.2 GW between February and July 2026. For investors, those details are the practical backbone of the story: speed, not just vision, is what would make a deal like this possible.
SemiAnalysis also argued the financing is manageable. The firm said a combination of Nvidia vendor financing and premium compute pricing of $30 million to $50 million per MW per year could allow the capex to pay back in under a year. That is an aggressive claim, and it should be read as the research group’s own estimate, not as a settled market fact. But it is also the kind of framework that makes a giant capital plan sound less like a moonshot and more like a high-stakes industrial arbitrage.
That explains why the rhetoric around the deal has turned so forceful. In the report, SemiAnalysis said, “SpaceX will develop anything they can and bring it online as fast as possible.” Jeremie Eliahou Ontiveros also described the opportunity as “a once-in-a-generation opportunity that SpaceX (SPCX) is well-positioned to serve.” Those are not casual lines. They signal that the researchers see a real window for Musk’s company to become a compute supplier rather than just a rocket and satellite operator.
There is still a lot that has not happened. No formal Microsoft-SpaceX contract has been announced, and the story remains a thesis until there is something concrete to trade on. The next real milestone would be a deal announcement or SpaceX’s next quarterly earnings report, which could offer clues on capex and compute-build progress. Until then, the market is left with a high-conviction research note, a huge capacity target and a pair of stocks reacting as if the idea may be more than hype.
For Microsoft, the upside is obvious if the plan works. A massive external compute pipeline could help Azure grow much faster without Microsoft taking on training costs directly. For SpaceX, the opportunity is different but just as striking: Musk would be trying to turn execution speed into a new revenue engine, one that could sit alongside his other businesses rather than replace them. For now, the market is signaling that investors are willing to take the idea seriously, and that alone is enough to keep SPCX and MSFT in focus.
