SpaceX Expands AI Infrastructure With Reflection Deal

SpaceX has signed a major compute agreement with Reflection AI, an open-source AI lab backed by Nvidia, giving the startup access to powerful AI infrastructure as it tries to compete with leading frontier model companies.

The deal gives Reflection AI access to SpaceX’s Colossus 2 data center and Nvidia GB300 chips, according to reports. Reflection will reportedly pay $150 million per month starting July 1, 2026, with the agreement running through the end of 2029.

The arrangement shows how compute access is becoming one of the most important weapons in the AI race. Model quality matters, but without enough chips, data centers and power, even ambitious AI labs can fall behind.

Compute Deal Summary

Reflection AI Gets the Compute It Needs

Reflection AI is building open frontier AI models and positioning itself as a serious open-source alternative to closed AI labs.

The company says it is working on “frontier open intelligence” and making it accessible to all. That mission requires huge amounts of compute, especially if Reflection wants to train models that can compete with systems from OpenAI, Anthropic, Google and Meta.

Reuters reported that the SpaceX deal gives Reflection access to Nvidia GB300 chips at SpaceX’s Colossus 2 data center. The agreement starts July 1, 2026, and runs through 2029, with Reflection paying $150 million per month.

SpaceX Becomes an AI Infrastructure Supplier

SpaceX is no longer only being discussed as a rocket, satellite and space-infrastructure company.

Through this compute deal, it is becoming a supplier of high-end AI infrastructure to outside labs. TechCrunch reported that Reflection is now tapping SpaceX for AI chips, following earlier compute relationships involving Anthropic and Google.

That matters because AI infrastructure is becoming a business of its own. Companies that control chips, data centers, power and cooling can sell access to AI labs that need compute but cannot build large-scale infrastructure quickly enough.

The $6.3 Billion Signal

The Wall Street Journal reported that the agreement is valued at about $6.3 billion over 3.5 years. It also said either party can terminate the deal with 90 days’ notice after the first three months.

That size puts the deal in the same conversation as major cloud and AI infrastructure contracts.

For SpaceX, the monthly revenue could make AI compute a meaningful new business line. For Reflection, the cost shows how expensive it has become to compete at the frontier-model level.

Open-Source AI Needs Big Infrastructure Too

Open-source AI is often discussed as a more accessible alternative to closed models. But building powerful open models still requires enormous infrastructure.

Reflection AI’s deal shows that open-source labs need the same core resources as closed AI giants: chips, data centers, engineering talent and long-term funding.

Axios reported that the deal reflects growing interest in open-source AI, especially as companies worry about dependence on proprietary models. A Reflection spokesperson said the market is increasingly recognizing the importance of open-source AI.

Nvidia’s Role in the Background

Nvidia remains central to the story.

Reflection is backed by Nvidia, and the SpaceX deal gives it access to Nvidia GB300 chips. That means Nvidia benefits from both sides of the ecosystem: it supplies the hardware and backs companies trying to build frontier models.

This is how the AI economy is increasingly working. The same companies can be investors, infrastructure suppliers, strategic partners and customers at the same time.

A New Kind of AI Power Structure

The deal also shows that AI competition is not just about model releases.

The real power structure now includes:

• Who has access to the newest chips
• Who can secure long-term data center capacity
• Who can pay for massive training runs
• Who can lower compute costs over time
• Who can distribute models widely after training

Reflection may be open-source in its mission, but it still needs expensive private infrastructure to reach the frontier.

Final Take

SpaceX’s compute deal with Reflection AI is another sign that the AI race is becoming an infrastructure race.

For Reflection, the agreement gives it access to the kind of compute needed to build more competitive open models. For SpaceX, it opens another path to monetize its data center capacity and AI chip infrastructure.

The bigger message is clear: open-source AI may be about access and transparency, but frontier AI still depends on billions of dollars in compute.

If Reflection can turn this infrastructure into strong open models, the deal could become an important moment for the open-source AI movement.