SpaceXAI Deploys Nvidia Vera CPUs for Grok's AI Agents and Orbital Data Centers in Space

Reviewed byNidhi Govil

10 Sources

Share

SpaceXAI has committed to deploying Nvidia's 88-core Vera CPUs to power Grok's agentic AI workloads, becoming the second hyperscaler after Meta to adopt the chip. The company plans to extend this technology into orbit with its Starmind AI satellite launching in Q4 2027, marking a significant step toward space-based AI compute infrastructure.

SpaceXAI Commits to Nvidia Vera for Agentic AI Workloads

SpaceXAI has become the second hyperscaler to deploy Nvidia's standalone Vera CPUs, following Meta's February announcement

1

. The 88-core Vera chip will orchestrate Grok's AI agents in SpaceXAI's data centers and, starting in the fourth quarter of 2027, inside the company's first Starmind AI satellite

4

. Nvidia claims the CPU for AI agents completes agentic AI, reinforcement learning, and data processing tasks up to 1.8 times faster than x86 processors, though this figure hasn't been independently verified

1

.

Why Vera Matters for AI Agent Performance

Agentic AI workloads spend significant runtime off the GPU. When a model writes code, it calls tools, queries databases, and parses results by leaning on the host CPU between every inference pass. Idle GPUs waiting on that work represent wasted capital

1

. Mike Nicolls, president of SpaceXAI, explained that Nvidia Vera provides the CPU performance and memory bandwidth to run enormous amounts of orchestration, code, and data processing while keeping GPUs doing what they do best

3

. This architecture enables higher-performance AI agents and extracts more useful work from every watt of compute

4

.

Source: NVIDIA

Source: NVIDIA

Nvidia Vera pairs 88 custom Olympus cores on a monolithic die with spatial multithreading and LPDDR5X memory delivering up to 1.2 TB/s of bandwidth

1

. The chip accelerates CPU-intensive work surrounding model inference, from tool use and code execution to data processing, orchestration and simulation

3

.

Expanding to Orbital AI Compute Infrastructure

SpaceXAI plans to expand its AI infrastructure for Grok with the Vera Rubin platform as it scales toward gigawatts of computing capacity

3

. The Vera Rubin NVL72 combines 72 Rubin GPUs and 36 Vera CPUs in a fully liquid-cooled rack

1

. For orbital deployment, this space-optimized AI system must be reworked to handle radiation exposure, heat rejection through radiators rather than facility water loops, launch vibration, and the absence of hands-on servicing

1

.

Elon Musk narrowed the launch window to Q4 2027, with significant scale-up following in 2028

1

. SpaceXAI's first-generation AI1 satellite design carries a 120 kW compute payload, peaking at 150 kW, on a craft wider than a Boeing 747

1

. Following a sun-synchronous orbit that keeps it facing the sun 98 percent of the time, AI1's massive solar panel array will reportedly generate 210kW of power

2

. The satellites will manage thermal management with 1,700-square-foot liquid radiators that release heat freely into the vacuum of space

2

.

Source: Engadget

Source: Engadget

Technical Challenges for AI in Space

Putting advanced AI chips in orbit presents unique challenges. Dr. Benjamin Lee, a professor at University of Pennsylvania's Department of Electrical and Systems Engineering, warned that using advanced chip architecture could expose satellites to unnecessary data corruption through bit flips

2

. Radiation from solar weather and other common events in low Earth orbit can force the ones and zeros that make up the chip's binary code to swap values. This susceptibility derives from smaller transistors in newer chips, as the amount of charge required to represent a one goes down as transistors get smaller

2

.

While AI systems are built to detect and correct these data corruption errors, the question isn't only whether Starmind will be able to detect and correct bit flip errors, but whether doing so will make orbital data centers comparatively inefficient

2

. Having to detect and correct errors repeatedly will slow down or add overhead to space-based computation, potentially creating inefficiencies inherently absent from terrestrial rivals

2

.

Industry Context and Competition

Nvidia has disclosed shipping hundreds of thousands of standalone Grace servers and more than 2.5 million Grace CPUs in total, and the SpaceXAI deal extends that dominance directly into territory held by AMD's EPYC and Intel's Xeon lines

1

. AMD has countered that its 256-core Zen 6 Venice processors beat Vera by 3.3 times in rack-level performance

1

.

NASA Administrator Jared Isaacman has advocated for moving data centers in space to address terrestrial energy constraints, arguing that orbital facilities could deliver immense benefits without competing for the same land and grid capacity as ground-based projects

5

. Lawrence Berkeley National Laboratory estimates data centers could consume 11.8 percent of U.S. electricity by 2030

5

. However, Amazon's AWS CEO Matt Garman has called the idea pretty far from reality, citing limited launch capacity and high payload costs

5

. SpaceXAI has acknowledged that orbital compute at the scale it's targeting requires significantly more chips than it currently has access to

1

.

Today's Top Stories

© 2026 TheOutpost.AI All rights reserved