😺 NVIDIA built a CPU. Musk shot it into space.
The Neuron Eric Gerard Ruiz ● Covered by 3 sources
NVIDIA’s new CPU is going with SpaceX into orbit. It’s a weird sign that AI agents need more than just flashy GPUs.
Based on reporting by The Neuron, Eric Gerard Ruiz — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
NVIDIA has landed a very public customer for a very unglamorous piece of silicon. SpaceXAI, Elon Musk’s AI arm behind Grok, is using NVIDIA’s new Vera CPU to help run the next wave of AI agents, and it’s taking that same chip architecture into orbit on its first AI satellite, Starmind.
That matters because agents are not just crunching numbers. They browse, run code, call tools and shuffle work around in the background. NVIDIA says Vera is built for that busywork, with 88 Olympus cores, up to 1.2 terabytes per second of memory bandwidth, and as much as 1.8x faster agentic task completion than the x86 chips that Intel and AMD have sold for decades.
The broader move is even more interesting. SpaceXAI is also expanding around NVIDIA’s Vera Rubin platform as it scales toward gigawatt-sized data centers. So this is not just a satellite story. It is an attempt to own the whole stack, from the model to the chip to the infrastructure that keeps the models moving.
NVIDIA says this is its first time selling standalone CPUs, and its own CFO has pegged the market it’s chasing at $200 billion. That helps explain why the company is suddenly acting less like a GPU vendor and more like a full-stack infrastructure shop. GPUs still get the headlines, but if the CPU can’t coordinate the mess around them, the expensive parts sit there waiting.
If SpaceXAI’s space plan works, “AI infrastructure” stops meaning only racks on Earth. If it doesn’t, it will still make for a very expensive demo reel.
My take — AI-written commentary, not fact-checked reporting
This is Musk logic in its purest form: if a datacenter is good, put it in orbit and call it strategy. NVIDIA is smart to chase the boring part of AI, because the boring part is where the bottlenecks hide and where real money gets spent. The rest of the industry keeps worshipping shiny model launches while the plumbing quietly eats the budget, which is usually how tech actually works.
Read more about this at: The Neuron