NVIDIA Says SpaceXAI Will Deploy Its New Vera Chip for AI Agents
NVIDIA says the chip, built to handle the coding and data work behind AI agents, will also anchor a satellite SpaceXAI plans to put into orbit.
NVIDIA said Tuesday that SpaceXAI will deploy its new Vera CPU to run the AI agent workloads behind Grok, the chatbot NVIDIA's announcement ties to the company. NVIDIA described Vera as the first CPU designed specifically for AI agents, the software systems that use tool use and code execution instead of just answering questions.
According to NVIDIA, agentic AI applications lean heavily on CPUs to direct tasks between calls to the larger AI models: running code, processing data and simulations, and making sure the more powerful graphics processors, or GPUs, stay busy. NVIDIA said Vera is meant to speed up that in-between work so agents can act closer to real time.
A chip and a satellite
NVIDIA said SpaceXAI plans to expand the AI infrastructure behind Grok using NVIDIA's Vera Rubin platform, which pairs the Vera CPU with NVIDIA's networking and software, as SpaceXAI scales toward what NVIDIA described as gigawatts of computing capacity.
The companies also said SpaceXAI intends to take that same computing architecture into orbit. NVIDIA said SpaceXAI's planned first-generation satellite, called Starmind, would run on an adapted version of the Vera Rubin NVL72 system, the same rack-scale hardware NVIDIA sells for data centers on the ground. NVIDIA said it is working with SpaceXAI to adjust that system for the power, heat and reliability demands of space, while keeping the same underlying architecture and software used on Earth. Neither company gave a timeline for when Starmind might launch.
"Agentic AI requires a new kind of computing system, one built not only to generate answers, but to take action," said Ian Buck, NVIDIA's vice president of hyperscale and high-performance computing, in the announcement. "Vera gives AI agents the CPU performance to act in real time."
Mike Nicolls, president of SpaceXAI, said in the same announcement that Vera gives the company "the CPU performance and memory bandwidth to run enormous amounts of orchestration, code and data processing while keeping GPUs doing what they do best."
What NVIDIA says the chip can do
NVIDIA said Vera includes 88 of its own Olympus processor cores, a technology it calls Spatial Multithreading, and memory that can move data at up to 1.2 terabytes per second. NVIDIA said internal testing found Vera completed tasks up to 1.8 times faster than x86 chips, the industry-standard processor architecture used in most existing servers, across agentic AI, reinforcement learning and data processing workloads. NVIDIA did not release independent benchmark data to support that comparison, and neither company disclosed the cost of the deployment or how many Vera chips SpaceXAI plans to buy.
Sources
- Press release SpaceXAI Adopts NVIDIA Vera CPU to Accelerate Agentic AI at Massive Scale NVIDIA Newsroom
Related coverage
xAI's Grok 4.6 Model Now Available on Amazon Bedrock
xAI said its newest artificial intelligence model is now generally available to developers through Amazon's cloud computing platform.
Nvidia to Provide $105 Billion in Financing for OpenAI's Ohio Data Center
The chipmaker will also invest $1.5 billion directly in SB Energy, the SoftBank-backed developer building the site on former nuclear land in Pike County, Ohio.
Asia's AI Boom Is Real, but Economists Warn Southeast Asia May Get Only a 'Short-Term Blip'
Economists say the region's low-cost labor and back-end manufacturing role could leave it stuck at the bottom of the AI supply chain, even as Washington and Beijing pressure it to pick a side.
Apple Cuts More Than 200 Jobs on Vision Pro and Siri Teams
The cuts hit gaming and immersive video staff on the Vision Pro headset team and older Siri engineers, as Apple shifts resources toward smart glasses and its new AI assistant.