- Reflection AI will pay SpaceX $150M a month from July 1, 2026 through 2029 — roughly $6.3B if the contract runs its full term
- The deal gives Reflection immediate access to Nvidia GB300 chips at SpaceX’s Colossus 2 data centre in Memphis, Tennessee
- SpaceX has now secured more than $80B in committed compute revenue from outside clients, transforming Colossus into one of the most commercially significant AI infrastructure assets in the world
SpaceX has signed a computing deal worth up to $6.3 billion with Reflection AI, the open-source artificial intelligence startup founded by two veterans of Google’s DeepMind lab, according to CNBC, which cited materials viewed by the publication.
Under the agreement, Reflection will pay $150 million a month from July 1, 2026, for access to Nvidia GB300 chips at SpaceX’s Colossus 2 data centre near Memphis, Tennessee. If the contract runs through its full term ending in 2029, payments total approximately $6.3 billion. Either party can walk away with 90 days’ notice after the initial three months.
The deal adds Reflection to a client list at Colossus that already includes Anthropic, Google, and Cursor, the AI coding startup SpaceX is in the process of acquiring for $60 billion. Committed revenues across those customers now exceed $80 billion through 2029, according to regulatory filings and public reporting.
How Colossus became a commercial compute landlord
SpaceX built Colossus to power xAI’s work — the company behind Grok, Elon Musk’s AI assistant — and has since opened it to paying outside customers following its record-breaking IPO. SpaceX priced its IPO at $135 per share on June 12, 2026, at a valuation of $1.77 trillion — the largest in history. Shares surged on debut, briefly pushing the market cap above $2.1 trillion intraday, but the IPO valuation itself was $1.77 trillion. The infrastructure now houses approximately 555,000 Nvidia GPUs acquired at a cost of around $18 billion, spread across multiple buildings in Memphis with a planned total power capacity of two gigawatts.
The commercial ramp has been fast. Anthropic signed on in May 2026 to lease all compute capacity at the original Colossus 1 site in a deal valued at roughly $45 billion through mid-2029. Google followed in early June with a commitment worth approximately $30 billion over the same period.
Reflection’s $6.3 billion agreement is the fourth major deal and, strategically, the most pointed: an AI lab explicitly built around open-source models, arriving days after the event that made open-source AI’s pitch impossible to ignore.
Who is Reflection, and why hasn’t it shipped anything yet
Reflection AI was founded in March 2024 by Misha Laskin, who led reward modelling for Google DeepMind’s Gemini project, and Ioannis Antonoglou, DeepMind’s sixth-ever researcher and a co-creator of AlphaGo, the AI system that defeated the world Go champion in 2016. The company emerged from stealth in March 2025 with $130 million at a roughly $545 million valuation, raised $2 billion at an $8 billion valuation from Nvidia, Sequoia, Lightspeed, and others in October 2025, and was last reported seeking $2.5 billion at a $25 billion valuation. Nvidia alone has invested $800 million.
It has not yet released a public frontier model. That detail sits a little awkwardly alongside a $25 billion valuation and a $6.3 billion compute commitment, and it is worth sitting with rather than explaining away. What Reflection has instead is founders with a track record that commands serious attention, early ties to U.S. government and national security customers including the Department of Energy’s Genesis Mission and Pentagon AI programmes, and a thesis that is gaining ground precisely because of events like the Anthropic Fable shutdown.
Its model weights, once released, will be publicly available. Revenue will come from large enterprises and governments building on top of them: sovereign AI customers who need models they can inspect, modify, and run on their own infrastructure without worrying about a Friday afternoon directive from Washington.
There is also an interesting loop embedded in the deal. Nvidia invested $800 million in Reflection. Reflection will now run on Nvidia GB300 chips that SpaceX purchased. The chipmaker is simultaneously a backer of the startup and, indirectly, a supplier to it through hardware it sold to the landlord. It is the kind of circular capital structure that keeps appearing in AI infrastructure, where the same handful of players keep showing up in each other’s cap tables, customer lists, and supply chains.
The compute race in numbers
Access to top-tier Nvidia chips has become the central constraint for any company trying to build or serve frontier AI. The Reflection deal is one piece of a spending wave that has no clear ceiling. Groq, which is taking a different approach by building its own inference chips rather than renting GPU capacity, raised $650 million to scale its global inference cloud. CoreWeave has expanded through multi-billion-dollar infrastructure commitments. Microsoft, Amazon, and Google continue pouring capital into data centres. Nvidia itself raised $25 billion in debt with a 30-year tranche, signalling that lenders believe AI infrastructure demand will run for decades.
Against that backdrop, SpaceX has moved from rocket company to compute landlord faster than almost anyone anticipated. For Reflection, securing guaranteed GB300 access resolves the single biggest constraint facing any frontier AI lab: chips. Leasing from Colossus rather than building its own data centres lets the company spend its capital on research rather than concrete and cooling systems.
The unanswered question is what Reflection actually ships. Pedigree does not guarantee products, and a $25 billion valuation with no public model is a bet that commands scrutiny as well as attention. The compute deal buys time, credibility, and capacity.
Whether that translates into frontier models capable of competing with OpenAI, Anthropic, and Google is the question the next 12 months will answer, and the one SpaceX has now, in a sense, also staked a claim on.
