Vera Rubin image from NVIDIA
Over the past several months, SpaceX CEO Elon Musk has been increasingly vocal about putting AI data centers in space. In February, he predicted that, within five years, “we will launch and be operating every year more AI in space than the cumulative total on Earth.” He said then that SpaceX, then still a private company, was preparing for 10,000 to 30,000 launches annually.
Now, SpaceX has made the biggest stride towards realizing that vision with a $100 billion Louisiana spaceport designed to support thousands of launches per year. Meanwhile, Musk set a fourth-quarter 2027 target for the first Starmind AI1 satellite, which will carry an optimized NVIDIA Vera Rubin NVL72 system rated for 175 kW of average compute and up to 250 kW at peak. NVIDIA announced at its January 2026 CES keynote that Vera Rubin chips had entered full production. NVIDIA confirmed then it expected the Vera Rubin Superchip in volume in the second half of 2026.
“SpaceX, in partnership with Nvidia, has designed a space-optimized Vera Rubin NVL72 system for launch to orbit in Q4 next year, with significant scale in 2028,” Musk wrote on X. “Our design is significantly simpler, lower cost, denser and lighter than a traditional rack,” Musk added.
The Q4 2027 target extends beyond the projections of SpaceX’s earlier securities filings. The company’s May 20 Form S-1 described the orbital data-center program as relying on “unproven technologies, or technologies that do not exist,” and projected orbital deployment as coming “as early as 2028.” The same filing classified SpaceX under SIC code 7370, “Services, Computer Programming, Data Processing,” a computing category rather than an aerospace one.
On the 2026 innovators list
SpaceX and NVIDIA are among R&D World’s 100 most innovative companies of 2026. The list examines what each company built, who is using it and the results it has reported. R&D World will publish the full comparison in the 2026 Global R&D Funding Forecast.
The S-1 also put SpaceX’s thermal approach on the financial record, naming radiators, vapor chambers, active cooling loops, coatings and high-bandwidth intersatellite lasers. While Musk has described the benefits of solar power in space, heat may be a central engineering constraint. An orbital data center has no ambient air or water to carry heat away, so every watt of waste heat must ultimately leave as radiation. Reporting on the uprated AI1 design describes roughly 75-meter solar wings paired with 30-meter radiator panels.
Musk has acknowledged the potential design considerations to deal with the heart. “You want to design it to be more radiation tolerant and run at a higher temperature,” Musk told Dwarkesh Patel in February. “Roughly, if you increase the operating temperature by 20% in degrees Kelvin, you can cut your radiator mass in half.”
Musk recently told The Economist that power and cooling have become tighter constraints on AI growth than chip availability outside China. SpaceX argues that near-continuous solar exposure, reduced battery requirements and freedom from terrestrial grid interconnection bottlenecks could eventually make orbit the cheapest place to run AI.
According to Louisiana Economic Development, construction is expected to begin in 2027, with the first launch targeted for as early as 2029. That would place the Louisiana spaceport’s debut after the first Starmind AI1 mission and Musk’s planned move to “significant scale” in 2028. Early Starmind missions would thus rely on SpaceX’s existing launch infrastructure, with Louisiana providing longer-term capacity.
Wells Fargo’s post-earnings model illustrates the scale of growth already contemplated by Wall Street. Analyst Ken Gawrelski projects SpaceX revenue of $231 billion in 2028, more than seven times the company’s second-quarter annualized run rate. At the current $1.85 trillion market value, that forecast corresponds to roughly eight times 2028 sales. In addition, Gawrelski estimates $316 billion in AI capital spending across 2026 and 2027. Consequently, he lowered his SpaceX price target from $230 to $215 in light of the financing burden attached to the growth forecast. For context, Goldman Sachs and Morgan Stanley, the company’s lead IPO underwriters, initiated coverage with respective price targets of $205 and $300.
The FCC accepted for filing SpaceX’s application for a constellation of up to one million orbital data-center satellites in February. The American Astronomical Society petitioned to deny, citing visible-light, infrared, and radio interference along with atmospheric reentry effects and collision risk, and Amazon’s Leo unit asked the commission to refuse the application as speculative. Amazon’s filing distills the scale problem into one number: sustaining a million-satellite constellation could require roughly 200,000 replacement satellites per year, about 550 per day, each needing radiation, thermal-vacuum, vibration, and functional testing before flight. The docket drew roughly 1,500 comments.
If SpaceX can navigate the challenges involved in orbital data center, the move offers SpaceX a path from launch provider to vertically integrated computing company. Currently, the company’s most profitable division is Starlink, and SpaceX has also signaled its intent to launch a standalone direct-to-consumer Starlink Mobile service to compete against traditional wireless carriers.
