SpaceX and Reflection AI close a billion-dollar deal to use Nvidia chips and strengthen the AI market

SpaceX, Elon Musk’s company, sealed a significant agreement to supply computing power with the startup Reflection AI. The deal positions the open-source artificial intelligence company as the latest to utilize the advanced infrastructure of Musk’s Colossus data center.
Under the terms set, Reflection AI will gain immediate access to Nvidia GB300 chips. These are cutting-edge processors that are crucial for training and running advanced AI models. Reflection AI has committed to paying SpaceX $150 million monthly, starting July 1, 2026, and extending through 2029, according to documents reviewed by CNBC.
If the contract is maintained until the end of its stipulated period, total payments would reach approximately US$6.3 billion.
The contract provides that both parties can terminate the agreement with 90 days’ notice, after the first three months of validity.
This deal underscores SpaceX’s strategy to monetize its vast data center infrastructure, built in part to support Grok, Musk’s AI chatbot and ChatGPT competitor. The company is now expanding this capability to offer computing services to other artificial intelligence companies.
SpaceX had already signed computing capacity partnerships with big names like Anthropic and Google, in addition to acquiring Cursor. The inclusion of Reflection AI in the list of clients is strategic, as it is a laboratory focused on open source models, a segment that is gaining relevance in the current scenario of reevaluation of the dependence on proprietary AI systems by governments and companies.
The current scenario favors open source artificial intelligence. Demand grew after Anthropic restricted access to its Fable and Mythos models, raising questions about the risks of relying exclusively on closed model suppliers for critical operations. This event reinforced the argument that open models offer more control, allowing customers to inspect, customize, and run models with greater autonomy.
Reflection AI has positioned itself directly with this proposal, seeking to develop open source AI models in the United States. Valued at US$25 billion, the startup aims to compete with cutting-edge systems from OpenAI, Anthropic and Google, providing greater flexibility to governments and companies compared to closed systems.
“Recent developments highlight the importance of open source to the AI ecosystem, with more nations and companies recognizing the risks and costs associated with relying exclusively on closed models,” a Reflection spokesperson said in a statement.
The startup stated that this agreement grants it additional processing capacity, which is key to accelerating the development of its “American open intelligence”.
Although it has not yet released a cutting-edge open source model to the public, Reflection AI has gained significant support from government and national security industry customers. The company collaborates with the Department of Energy’s Genesis Mission and participates in broader Pentagon initiatives in the area of artificial intelligence.
For SpaceX, the pact signals that computing power has become a key strategic asset in the global race for AI. Access to advanced Nvidia chips remains one of the biggest challenges for companies developing and deploying cutting-edge models. By opening Colossus to outside users, SpaceX aligns itself with cloud providers and AI infrastructure companies vying to sell graphics processing power, a scarce resource.
This move also strengthens SpaceX’s growing narrative regarding its artificial intelligence infrastructure.
Investors are closely following SpaceX’s strategy, evaluating its ability to expand operations beyond rockets and Starlink, entering markets such as artificial intelligence, data centers and computing services.






