SpaceXAI Will Deploy Standalone Nvidia Vera CPUs For Grok’s Agentic Workloads — Will Use Optimized Vera Rubin NVL72 In Space With Starmind Satellite – Tom’s Hardware

TL;DR

Buying for a business?Offer from Amazon

Get business pricing on tech for your team

• Business-only prices and quantity discounts
• Tax-exempt purchasing
• Multiple users, one account, clear invoices

Create a free Business account →Free to register

As an affiliate, we earn on qualifying purchases.Hide

SpaceXAI says it plans to deploy standalone Nvidia Vera CPUs for Grok’s agentic workloads. It also plans to use an optimized Vera Rubin NVL72 system in space with a Starmind satellite, but no schedule, technical configuration or performance data has been disclosed.

SpaceXAI plans to deploy standalone Nvidia Vera CPUs for Grok’s agentic workloads, while an optimized Vera Rubin NVL72 system is slated for use in space with a Starmind satellite. The plan links xAI’s artificial-intelligence computing work with space-based infrastructure, although the company has not disclosed a deployment date, system specifications or evidence showing how the hardware will perform in orbit.

The stated plan separates two computing roles. SpaceXAI will use standalone Vera central processing units for workloads associated with Grok agents, rather than limiting Vera to its expected role alongside Nvidia accelerators. In common AI usage, agentic workloads involve models carrying out multi-step tasks, calling software tools and maintaining working state. SpaceXAI has not detailed which Grok services would run on the CPUs or whether they would handle inference, orchestration, data preparation or other supporting work.

A second part of the announcement concerns an optimized Vera Rubin NVL72 configuration intended to operate in space with a Starmind satellite. The wording indicates that the system would be adapted for orbital use, but no design changes were identified. Space hardware commonly faces constraints involving power, cooling, radiation, communications and maintenance. SpaceXAI has not said how the NVL72 configuration would address those conditions, how much of the system would be placed aboard the satellite or whether the mission has reached hardware testing.

Grok Computing Moves Beyond GPUs

The standalone CPU deployment matters because modern AI services depend on more than accelerator performance. Agentic systems can generate heavy CPU demand when they coordinate tools, retrieve information, process requests and manage several actions at once. Assigning Vera CPUs to Grok workloads could give SpaceXAI a way to tune those supporting tasks around Nvidia’s newer architecture, though the company has not published comparisons against its current infrastructure.

The satellite plan carries a different implication: SpaceXAI appears to be exploring AI computing within orbital infrastructure, not only communications between ground-based data centers and satellites. If the system operates aboard Starmind as described, it could provide a test of how rack-scale AI hardware behaves in space. The practical value will depend on what processing occurs in orbit, how results are transmitted and whether the system offers an advantage over ground computing. None of those outcomes has been established by the announcement.

Vera Rubin Reaches Orbital Plans

Vera Rubin is Nvidia’s planned computing platform built around its Vera CPU and Rubin accelerator architecture, while NVL72 denotes a large integrated system rather than a single processor. SpaceXAI’s plan extends that hardware roadmap into two less conventional uses: Vera operating as a standalone CPU for Grok agents and a modified NVL72 configuration associated with a satellite. The announcement does not establish that either deployment is already operational.

The development also connects Grok’s expanding compute requirements with SpaceXAI’s access to space infrastructure. That combination could support experiments in onboard processing, reduced reliance on constant ground links or distributed computing across future spacecraft. Those possibilities remain interpretations of the stated direction, not confirmed mission capabilities.

Hardware Design and Schedule Undisclosed

Several central details remain unknown. SpaceXAI has not given a launch date or deployment timetable, identified the number of Vera CPUs involved or described the Grok workloads assigned to them. It is also unclear whether the standalone CPUs will operate in existing data centers, new facilities or infrastructure shared across SpaceXAI operations.

For the orbital system, the company has not disclosed power consumption, cooling methods or radiation protections. The announcement also leaves open whether “NVL72” refers to a full rack-scale configuration in orbit or an adapted design retaining the name while changing its physical layout. No benchmark results, test records, satellite specifications or independent verification were provided. The claimed deployment should be treated as a forward-looking company plan until hardware and mission milestones are documented.

Testing Will Define the Deployment

The next measurable steps would include technical specifications for the standalone Vera systems, details about their role in Grok’s software stack and a timetable for installation. For Starmind, mission information identifying the satellite, launch window and orbital computing objective would show how far the project has advanced. Readers should also watch for ground-test or in-orbit performance data, since the announcement alone does not show that the proposed configuration can operate under space power and thermal limits.

Key Questions

What has SpaceXAI announced?

SpaceXAI says it will deploy standalone Nvidia Vera CPUs for Grok agentic workloads and use an optimized Vera Rubin NVL72 configuration in space with a Starmind satellite.

Are the Vera CPUs already running Grok workloads?

No operational start was disclosed. The available information describes a planned deployment, while its current installation status remains unconfirmed.

What will the Starmind satellite system do?

SpaceXAI has not specified the satellite’s computing tasks. It is unclear whether the orbital Vera Rubin system will run AI models, process spacecraft data, manage communications or serve as a hardware demonstration.

Has SpaceXAI released performance data?

No benchmarks or test results were included. Claims about performance, efficiency or orbital reliability cannot be checked until SpaceXAI releases technical evidence or the system begins documented operations.

Source: xAI

Source: xAI