- Lead. Open-source AI startup Reflection AI has agreed to pay SpaceX $150 million per month from July 1 for access to Nvidia GB300 chips at SpaceX’s Colossus 2 data centre near Memphis — a contract running through 2029 for a potential total of $6.3 billion.
- Fact. Colossus 2 houses approximately 555,000 Nvidia GPUs acquired at a cost of around $18 billion; SpaceX’s total committed compute revenues from all Colossus tenants now exceed $80 billion through 2029, with Anthropic (~$45 billion) and Google (~$30 billion) the other major customers.
- Stake. Nvidia invested $800 million in Reflection AI while simultaneously supplying the GB300 chips SpaceX is leasing to it — a circular structure that illustrates how concentrated the capital flows powering the AI infrastructure buildout have become.
Reflection AI will begin paying SpaceX $150 million per month starting July 1, 2026, for dedicated access to Nvidia GB300 GPUs at SpaceX’s Colossus 2 facility near Memphis, Tennessee, according to TechFundingNews. The contract runs through 2029, with a 90-day exit clause available after the first three months, for a potential total commitment of approximately $6.3 billion. The deal adds to SpaceX’s growing position as a compute landlord — a business line that has emerged as one of the defining revenue streams from SpaceX’s AI infrastructure strategy since its Nasdaq IPO in June 2026.
Who Is Reflection AI
Reflection AI was founded in March 2024 by Misha Laskin, who led reward modelling for Google DeepMind’s Gemini family, and Ioannis Antonoglou, DeepMind’s sixth-ever researcher and a co-creator of AlphaGo. The company raised $2 billion in October 2025 at an $8 billion valuation and is now seeking a $25 billion valuation in its next financing round. It has described itself as an open-source frontier AI developer but has not yet released a public model. Nvidia has invested $800 million in the company — the same chipmaker supplying the GPUs that SpaceX will lease to Reflection under the deal.
The Colossus 2 Stack
Colossus 2, the Memphis facility at the centre of the agreement, houses approximately 555,000 Nvidia GPUs acquired at a cost of around $18 billion, with a planned power draw of 2 gigawatts. SpaceX’s total committed compute revenue from the facility now exceeds $80 billion through 2029, with Anthropic reportedly holding a commitment of roughly $45 billion — signed in May 2026 — and Google around $30 billion from early June 2026. The Reflection deal fills the remaining committed slot in what has become one of the largest concentrations of contracted AI compute capacity in a single facility globally.
The Nvidia angle — investing in a customer while supplying the hardware that a third party leases to that customer — illustrates the circular capital structure that has emerged in the AI infrastructure buildout. Nvidia has also recently raised $25 billion in debt, a financing move analysts cited as signalling lender confidence in decade-long AI infrastructure demand rather than a sign of operational strain. The structure of the Reflection-SpaceX-Nvidia relationship compresses supplier, investor, and landlord roles into a loop that raises concentration questions even as it accelerates the pace of compute deployment.