The company you know as xAI doesn't exist anymore. It's now SpaceXAI, and its next move is sending Nvidia's newest chip architecture into space.
Nvidia announced Monday that SpaceXAI will deploy its new Vera CPU to accelerate agentic AI workloads. The chip is designed specifically for AI agents that take action, not just generate answers.
Wait, Who Is SpaceXAI?
SpaceX acquired xAI in an all-stock deal in February 2026. In May, Elon Musk announced xAI would stop existing as a separate company entirely.
Grok and X now operate under SpaceXAI, a division of SpaceX itself. Michael Nicolls, a former Starlink vice president, runs it as president.
That corporate restructuring matters for understanding this announcement. SpaceXAI isn't a startup buying chips. It's SpaceX's AI arm, with SpaceX's capital and infrastructure behind it.
Nvidia's own IPO plans and revenue depend heavily on exactly this kind of large, well-capitalized customer. SpaceX has already signaled ambitions for a Nasdaq listing that could value the combined SpaceX-xAI entity in the trillions.
A CPU Built Specifically for AI Agents
Agentic AI systems need more than raw model power. They need CPUs to orchestrate tools, execute code, process data and run simulations between each model call.
Nvidia's Vera is built for exactly that gap. It packs 88 Nvidia-designed cores and up to 1.2 terabytes per second of memory bandwidth.
Nvidia says Vera completes agentic AI tasks up to 1.8 times faster than standard x86 CPUs. That figure comes from Nvidia's own testing, not an independent benchmark.
"Agentic AI requires a new kind of computing system," said Ian Buck, Nvidia's vice president of hyperscale and high-performance computing. "One built not only to generate answers, but to take action."
SpaceXAI president Mike Nicolls framed the appeal in practical terms.
"Vera gives us the CPU performance and memory bandwidth to run enormous amounts of orchestration, code and data processing," he said. "That means higher-performance AI agents."
The Real Story: An AI Satellite
The more striking detail is where SpaceXAI plans to take this architecture next. Not a bigger data center. Orbit.
SpaceXAI's planned first-generation Starmind satellite will run on an optimized version of Nvidia's Vera Rubin NVL72 system. That's the same rack-scale architecture powering Nvidia's largest terrestrial AI factories.
Space imposes constraints no data center faces. Power, cooling, bandwidth and physical durability all work differently once you leave Earth's atmosphere.
Nvidia and SpaceXAI say they're adapting that architecture for orbital use. They plan to keep the same underlying software ecosystem intact throughout.
If it works, the same chip stack could eventually run AI training on Earth and in orbit. One shared foundation, two completely different environments.
What This Means for Miami
This is a genuinely different kind of AI infrastructure story. It's not the data center fights MAIN has covered across Florida this year.
Orbital computing sidesteps the exact fights over land, water and grid capacity driving local moratoriums.
Whether that becomes commercially real or stays aspirational is still an open question. Musk's companies have a track record of announcing ambitious infrastructure years before it actually ships.
Starlink itself took years to move from announcement to a working commercial network. Starmind may follow a similar timeline before any of this reaches orbit.
For Miami's aerospace and AI investors, this is worth tracking regardless. It's one more sign the AI infrastructure race is expanding.
Not every fight is about which state gets the next data center anymore.