Nvidia ties itself to SpaceX: a space entry ticket to computing power hegemony
Many people treat Nvidia’s stake in SpaceX as a financial investment. That’s a narrow view. With this move, Huang Renxun is laying the groundwork—an entry ticket to future computing power hegemony.
The origin of this investment isn’t that Nvidia directly bought SpaceX shares. In January 2026, Nvidia injected $10 billion into Elon Musk’s xAI. Two months later, SpaceX acquired xAI through a stock-swap arrangement. Nvidia’s xAI equity automatically converted into approximately 122.8 million Class A shares of SpaceX. At one point, the book value was around $21 billion, making Nvidia the sixth-largest shareholder of SpaceX.
A double on half-year floating gains is certainly impressive, but that’s never the point.
From “selling shovels” to “turnkey contractor”
NVIDIA is undergoing an identity transformation. In the past it only sold GPUs; now it wants to bind every link of AI infrastructure with capital—chips, systems, networks, energy, financing, and even space.
At SpaceX’s first earnings call after going public, Musk made Huang Renxun the toughest promise: SpaceX’s AI data centers will be “built entirely on NVIDIA,” and “we believe the Vera Rubin architecture is the best architecture, so we work only with NVIDIA.”
This isn’t small talk. SpaceX has already built two Colossus data centers: in the first facility, more than 220,000 NVIDIA GPUs are deployed; in the second, more than 550,000. Just GPU procurement alone cost about $18 billion. Musk plans to push computing capacity to nearly 10 GW by the end of 2027 and has reserved a “large allocation” of the next-generation Vera Rubin GPU.
What this investment truly buys is having the future orders of this super customer welded in advance.
Space computing: the next strategic high ground
A deeper chess move—above the sky.
SpaceX and NVIDIA have already launched a joint R&D project, Starmind AI1 satellite computing payloads, to send AI computing power into orbit. Musk’s goal is clear: deploy data centers in space, using infinite solar energy and natural radiative cooling to solve the power and land bottlenecks of ground data centers.
The Starlink satellite network itself is a global communications infrastructure, and distributed AI inference, edge computing, and sovereign AI deployments all require a global connectivity layer that doesn’t rely on traditional fiber optics.
Whoever controls space computing nodes controls the throat of global communications and edge intelligence. By deeply binding itself to SpaceX through equity, NVIDIA effectively locks in advance the right to set the chip standards for space infrastructure.
The deep logic of the “Ma-Huang” alliance
Musk’s ultimate goal is colonizing Mars, while Huang Renxun’s ultimate vision is computing power everywhere. Their ambitions converge here: without computing power, Mars is just a wasteland; without space infrastructure, computing power can’t break beyond Earth’s boundaries.
That explains why Musk has shifted from early in-house Dojo chips and complaints that NVIDIA GPUs are expensive to today’s full volte-face. The technical roadmap is just the surface—capital ties are the truth. Over the past two years, NVIDIA has poured more than $100 billion into AI companies, building a “chain” network covering CoreWeave, Anthropic, OpenAI, and SpaceX.
What NVIDIA is buying with this investment isn’t SpaceX’s shares—it’s the topology of the AI computing-power network for the next decade. From ground data centers to orbital computing nodes; from chip suppliers to infrastructure EPC contractors—Huang Renxun is rewriting the power map of AI deployment with capital.
While Wall Street is still calculating the paper-gain profits of doubling this investment in half a year, Huang Renxun and Musk are already planning the next era’s computing-power frontier.