SpaceX is rapidly parlaying its massive investments in data center infrastructure into a lucrative new business line. A leaked agreement reveals that the aerospace giant has signed a monumental $6.3 billion contract with Reflection AI, a leading open-source artificial intelligence lab. Under the terms, Reflection AI will gain immediate access to SpaceX’s "Colossus" infrastructure, specifically utilizing the high-demand NVIDIA GB300 chips to train and deploy its next generation of advanced models.
The financial structure of the deal is aggressive, reflecting the extreme scarcity of top-tier AI compute power. Starting in July 2026, Reflection will pay SpaceX roughly $150 million per month through 2029. While the contract allows for termination after an initial three-month period with 90 days' notice, the potential $6.3 billion windfall represents a significant diversification of SpaceX’s revenue streams beyond satellite launches and Starlink telecommunications.
This partnership signals a profound shift in the AI landscape, where the "landlord" of compute power holds as much influence as the model developers themselves. Originally built to support Elon Musk’s xAI and the Grok chatbot, the Colossus facility is now being marketed to a growing roster of external clients. By hosting firms like Anthropic and Google, and now moving to acquire AI-coding platform Cursor, SpaceX is positioning itself as a formidable challenger to traditional cloud giants like Amazon Web Services and Microsoft.
Reflection AI brings a unique strategic dimension to this deal through its commitment to open-source "American open intelligence." As concerns grow over the monopolistic control of closed-model providers, Reflection’s $25 billion valuation is built on the premise of providing flexible, auditable AI to governments and enterprises. Its deep involvement in US Department of Energy and Department of Defense projects suggests that SpaceX’s infrastructure is becoming central to the national security AI stack.
Ultimately, the deal highlights that in the current AI gold rush, the physical infrastructure of compute has become a strategic resource of the highest order. Access to NVIDIA’s latest silicon remains the primary bottleneck for the industry, and SpaceX’s ability to secure and host this hardware at scale gives it immense leverage. Musk’s empire is evolving into an integrated AI ecosystem where hardware, energy, and intelligence are inextricably linked.