SpaceX compute deal with Reflection AI: $150M/month to 2029

SpaceX will provide Reflection AI with immediate access to Nvidia GB300 AI chips and supporting hardware at its Colossus 2 data center, under a contract in which Reflection will pay $150 million a month beginning July 1, 2026 through 2029. The agreement is worth up to $6.3 billion and includes an option for either company to end the contract with 90 days’ notice after the first three months.

What exactly did the deal commit to?

Reflection AI committed to pay $150 million a month for access to Nvidia’s latest GB300 AI chips and supporting hardware housed at SpaceX’s Colossus 2 data center near Memphis, Tennessee, starting July 1, 2026 and running through 2029, with the full contract valued at up to $6.3 billion. Either company may terminate the contract with 90 days’ notice after the first three months.

The Colossus data center was originally built by xAI for its own AI efforts and is now part of SpaceX’s infrastructure, which SpaceX has been renting out as its internal AI pursuits slowed. Reflection said the deal is one of the largest announced open AI infrastructure commitments to date.

How does this compare to other SpaceX compute deals?

The Reflection agreement is smaller than SpaceX’s earlier deals: Anthropic’s contract cost $1.25 billion per month, and Google’s contract cost $920 million per month, and those contracts also run through July 2029. Elon Musk has publicly downplayed the three-year term of those larger contracts, noting the contracts can be canceled at any time.

Reflection’s $150 million monthly commitment is therefore a fraction of the scale of the Anthropic and Google deals, but the company framed the arrangement as strategic runway to scale open-weight models using the rented GB300 capacity.

Who is Reflection and what is open-weight AI?

Reflection AI, founded in 2024 by two former Google DeepMind researchers, positions itself as an open source alternative to closed frontier labs. Open-weight AI models publicly release their trained parameters, and Reflection has emphasized that approach following recent policy moves, including the U.S. government’s ban of Anthropic’s closed models Fable and Mythos.

A Reflection spokesperson said, "Recent events highlight how important open source is to the AI ecosystem, with more nations and enterprises recognizing the risks and costs associated with exclusively depending on closed models." The company argued the SpaceX compute deal gives it more runway to build open models at scale.

Why it matters

The deal signals continued demand from both closed and open labs for bulk access to advanced AI accelerator hardware. SpaceX is monetizing a substantial installed base of Nvidia hardware initially deployed for xAI, and the pricing spread between Reflection and the larger labs highlights different capital and scale profiles across the sector. For open-weight projects, securing long-term chip access is a core operational constraint; Reflection’s contract ties compute availability directly to its development timeline.

What to watch

Monitor whether Reflection exercises early termination rights or expands the agreement before July 2029, and watch whether other open-weight labs follow with similar multi-year compute commitments. Also watch how SpaceX’s inventory of Nvidia GB300 capacity is allocated among labs as those contracts evolve.