Reflection AI is making a large bet on compute, and it is doing it through SpaceX. The open-source AI startup will pay $150 million a month for access to Nvidia’s latest GB300 AI chips and supporting hardware at SpaceX’s Colossus 2 data center near Memphis, Tennessee.
The agreement begins July 1, 2026 and runs through 2029. TechCrunch reported that the deal is worth up to $6.3 billion, with either company able to end the contract with 90 days’ notice after the first three months.
What Reflection AI Is Buying
The core of the agreement is immediate access to AI infrastructure. Reflection AI is tapping SpaceX for Nvidia’s latest GB300 AI chips and the supporting hardware needed to use them at scale.
For frontier AI companies, compute is not a side issue. Large AI models require access to specialized chips, data center capacity, and the surrounding systems that keep the hardware running. In this case, Reflection AI is using its first compute deal to secure a direct path to that infrastructure.
The deal also gives Reflection AI a way to position itself more aggressively in the market for open AI models. The company described the agreement as one of the largest announced open AI infrastructure commitments to date.
How The SpaceX Deal Compares
Reflection AI is not the only AI company renting compute from SpaceX. The source article says SpaceX has also made deals with Anthropic and Google.
Those agreements are larger. Anthropic’s deal costs $1.25 billion per month, while Google’s costs $920 million per month. Both contracts also run through July 2029.
Elon Musk has publicly downplayed the three-year term of those deals, emphasizing that the contracts can be cancelled at any time. Reflection AI’s agreement includes a specific exit structure: either company has the option to end the contract with 90 days’ notice after the first three months.
That structure matters because compute contracts can be large, long-running commitments. Reflection AI’s monthly payment is smaller than the Anthropic and Google deals, but it is still a major infrastructure commitment for a startup founded in 2024.
Why Open-Weight AI Is Central To The Pitch
Reflection AI has framed itself as an open-source alternative to closed frontier labs such as Anthropic and OpenAI. Its strategy centers on open-weight AI models, which publicly release their trained parameters.
The company used the SpaceX compute agreement to argue that open models need serious infrastructure behind them. More access to chips gives the company more room to train and scale its systems.
Reflection AI tied the deal to broader interest in open-weight AI. According to the source article, open-weight AI models have received more attention following the U.S. government’s ban of Anthropic’s closed models, Fable and Mythos.
“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,” a spokesperson said in an emailed statement. “Our deal with SpaceXAI signals Reflection’s strategic importance within the frontier AI ecosystem, and more compute means more runway to build the world’s best open models at scale.”
The statement makes the company’s message clear. Reflection AI is not presenting the SpaceX deal only as a hardware purchase. It is presenting it as a strategic step for open-source AI at the frontier level.
Why SpaceX Has Compute To Rent
The compute is located at the Colossus 2 data center near Memphis, Tennessee. The source article says the Colossus data center was originally built by xAI, a company founded by Elon Musk that is now part of SpaceX, for its own AI efforts.
As those internal efforts faltered, SpaceX began using its AI chip holdings in another way: renting them out to major AI labs. That shift has turned SpaceX’s compute supply into a resource for companies that need large-scale AI infrastructure quickly.
For Reflection AI, that creates an opening. Instead of building every part of the infrastructure stack itself, the company can pay for access to an existing pool of chips and hardware.
The result is a deal that connects several major AI themes at once: the rising cost of compute, the importance of Nvidia AI chips, the competition between open-weight AI and closed models, and the growing role of large data centers in determining which labs can compete at scale.
What The Deal Signals
Reflection AI’s agreement with SpaceX does not make it larger than Anthropic or Google’s compute commitments. The numbers in the source article show the opposite: those companies are paying far more per month.
But the Reflection AI deal is still significant because it gives an open-source AI startup a major compute runway. The startup is using that access to reinforce its argument that open-weight AI can compete closer to the frontier if it has enough infrastructure behind it.
The agreement also shows how valuable unused or available AI chip capacity has become. SpaceX can rent compute to major AI labs, while Reflection AI can use that capacity to support its open model strategy.
Beginning July 1, 2026, that strategy becomes tied to one of the most expensive inputs in AI: sustained access to advanced chips. For Reflection AI, the SpaceX contract is not just a cost. It is the foundation for its next stage of open-source AI development.