Why Did Musk Say SpaceX Will Use "Only Nvidia"? international
Elon Musk, who leads SpaceX and Tesla, made a surprise declaration on the 4th (local time). The day drew attention as the first earnings announcement after SpaceX's listing, but what the market focused on was Musk's vision for the future. SpaceX said it would secure more than 2 gigawatts (GW) of computing infrastructure by the end of this year and expand it to around 10 gigawatts by the end of next year.
What the semiconductor market focused on was Musk's abrupt declaration that he would use only Nvidia chips in data centers. Through his own social media platform X, he said, "SpaceX has decided to use only Nvidia GPUs. Because they are the best." In the conference call held after the earnings release, he also said, "We think Nvidia's Vera Rubin architecture is the best, so we decided to use only Nvidia products. We think this is the best AI computer, and we value the close collaboration and partnership with Nvidia very much in many respects. We think the NVL72 computer design is a far better design than the standard rack style."
Musk's remarks threw cold water on AMD, which had begun its full-fledged pursuit of Nvidia with the shipment of "Helios" in the second half. Helios is an AI server structure that combines AMD's latest GPU, central processing unit (CPU), networking and software into a single equipment structure, and had been cited as a challenger to block the dominance of Nvidia's Vera Rubin NVL72. The NVL72 integrates seven chips, including the Nvidia GPU Rubin and the CPU Vera, into a single system and connects them with NVLink communication technology.
AMD Chief Executive Officer (CEO) Lisa Su praised SpaceX as a "very important technology company" in a CNBC interview on the 5th, the day after Musk's remarks. "I have deep respect for Elon Musk and everything he has achieved, and I look forward to continuing our partnership over the long term," she said, appearing to make an effort to reassure investors.
As recently as May, Musk had said it was highly likely that Tesla and SpaceX, which he leads, would continue to buy both Nvidia and AMD chips, but he suddenly chose Nvidia. AMD took a direct hit from Musk's unexpected remarks. Even though AMD posted results that exceeded expectations in its earnings announcement held on the same day as SpaceX, its stock price plunged. Why did Musk suddenly say he would use only Nvidia?
◇Judged Optimal for Space Data Centers
First, Musk judged that Nvidia chips are the optimal infrastructure to realize SpaceX's vision of building data centers in space. In this conference call, SpaceX said it would launch Starmind AI satellites, which are space data centers, starting next year and equip them with the Vera Rubin NVL72. Following SpaceX's ground data centers, Nvidia chips will also play a key role in space data centers.
In March this year, Nvidia unveiled its latest space-accelerated computing platform, which implements data-center-class performance in an environment constrained by size, weight and power. It connects artificial intelligence (AI) to operate seamlessly from ground to space and between space. The "Space-1 Vera Rubin Space Module," a component of the platform, is built on the Rubin GPU. It delivers up to 25 times improved AI computing performance in space-based inference compared with the previous H100 GPU. The "Nvidia IGX Thor," which helps process data on spacecraft, and the Nvidia Jetson Orin, an ultra-compact AI inference module, are also included in the platform.
AMD also operates the "AMD Versal AI Edge" series platform targeting the aerospace market, but analysts say SpaceX preferred Nvidia, which uses the latest GPU modules. Nvidia's strengths also include the "CUDA" software optimized to run Nvidia chips, as well as networking equipment.
In November last year, AMD said it had advanced packaging technology capable of supporting geostationary satellites, lunar exploration and deep-space probe missions in space for up to 15 years, but it is evaluated as lagging behind Nvidia in terms of the infrastructure full stack (package). AMD is collaborating with Blue Origin, which is regarded as a SpaceX rival.
◇Nvidia's Solid Supply Chain
Nvidia's chip supply chain is also cited as a reason Musk chose Nvidia. Nvidia is expected to overtake Apple this year to become the largest customer of Taiwan's TSMC, the world's largest foundry (contract chipmaker). This means that as the industrial landscape shifts from mobile to AI data centers, TSMC is prioritizing the production of Nvidia chips.
Although Musk is pushing to build Terafab to directly produce his own AI chips for use at Tesla and SpaceX, analysts say it will be difficult to acquire foundry capabilities like those of TSMC, Samsung Electronics and Intel in a short period. After Musk said he would build a factory producing 1 million chips a month, Nvidia CEO Jensen Huang took a negative stance at a TSMC event in November last year, saying, "Building an advanced chip manufacturing facility is very difficult. Not only building the factory itself, but the engineering, science and craftsmanship are all very difficult." Analysts say that because Terafab is a formidable undertaking, from Musk's standpoint he would have needed Nvidia, which controls a powerful chip supply chain.
◇Reliant on Nvidia From the Start... A Foregone Conclusion
There is also an analysis that, because SpaceX has relied on Nvidia from the start, continuing to use Nvidia chips was the natural course. xAI, SpaceX's AI development subsidiary, has mainly used previous GPUs to Rubin, such as Hopper and Blackwell, to train its AI model Grok. IT media outlet Tom's Hardware noted, "There is no public announcement or evidence that xAI developed Grok with chips from AMD or any other company," adding, "Because xAI's supercomputers have already used Nvidia accelerators for years, the exclusive deal officially announced this time appears to be closer to a formality than a decision reached after much deliberation."...
Published 2026.08.06. 06:02:23