NVIDIA said SpaceXAI will deploy its Vera Rubin platform to accelerate agentic AI workloads, bringing the first CPU designed specifically for AI agents to the company's expanding infrastructure behind Grok. The deployment spans terrestrial data centers and a planned first-generation Starmind AI satellite, which will carry an optimized Vera Rubin NVL72 system into orbit.
Agentic AI applications depend on CPUs to orchestrate tools, execute code, process data, and run simulations between model calls. SpaceXAI will use Vera for these workloads, which helps AI agents act faster while keeping GPUs fully utilized.
Why CPUs matter for agentic AI
Agentic AI systems don't just generate responses - they take actions. That requires coordinating multiple steps, managing large token histories, and running precise reasoning between GPU operations. NVIDIA's Vera architecture is built for those demands, pairing CPU performance with memory bandwidth to handle complex reasoning workloads.
"Agentic AI requires a new kind of computing system - one built not only to generate answers, but to take action," said Ian Buck, vice president of hyperscale and high-performance computing at NVIDIA. "Vera gives AI agents the CPU performance to act in real time - executing code, processing data and coordinating complex tasks. SpaceXAI is taking this architecture from earth to orbit."
SpaceXAI plans to scale toward gigawatts of computing capacity as it expands its AI infrastructure for Grok (from xAi). The company will integrate the Vera platform into its data centers and extend the same architecture to its Starmind satellites, creating a computing backbone that spans both terrestrial and orbital systems.
"Vera gives us the CPU performance and memory bandwidth to execute complex reasoning while keeping GPUs doing high-volume computation," said Mike Nicolls, president of SpaceXAI. "That means higher-quality AI agents and more efficient use of compute."
From data centers to orbit
The Starmind satellite deployment is notable because it extends the same accelerated computing architecture used in Earth-based data centers into space. That approach lets SpaceXAI use one platform for training frontier models on the ground and running specialized edge computing applications in orbit.
The choice also ties SpaceXAI into NVIDIA's broader accelerated computing ecosystem, which covers model training, frontier model development, and edge applications. For developers and IT professionals working on AI systems, the practical takeaway is that CPU performance is becoming a bottleneck for agentic workloads - and infrastructure decisions increasingly reflect that.
Why this matters for AI for IT & Development
For developers building agentic AI systems, this announcement signals that CPU orchestration is now a first-class design consideration, not an afterthought. The pattern here - pairing high-performance CPUs with GPUs to handle the "thinking" and "acting" phases of agentic workloads - is one that applies well beyond SpaceXAI's scale. Teams designing their own agent pipelines should evaluate whether their CPU infrastructure can keep up with GPU utilization, especially as token histories grow and multi-step reasoning becomes standard in production systems. For those tracking the broader trend, the AI for IT & Development space is seeing a shift toward specialized hardware that reflects how AI work has changed - from generating text to executing complex tasks.
Your membership also unlocks:
