Reflection AI Signs Up to
$6.3 Billion Compute Deal With SpaceX for Nvidia GB300 Chips
Reflection AI has signed a major compute agreement with SpaceX, securing access to Nvidia’s latest AI hardware in a deal worth up to $6.3 billion as the startup ramps up its efforts to build open-weight artificial intelligence models.
Under the agreement, Reflection AI will pay $150 million per month starting July 1, 2026, through 2029 for access to Nvidia GB300 AI chips and supporting infrastructure hosted at SpaceX’s Colossus 2 data center near Memphis, Tennessee. Both companies retain the option to terminate the contract with 90 days’ notice after the initial three-month period.
While substantial, the deal remains smaller than SpaceX’s existing AI infrastructure agreements with major industry players. According to the company, contracts with Anthropic and Google are valued at approximately $1.25 billion and $920 million per month respectively, with similar terms extending through mid-2029.
Reflection AI is using the partnership to reinforce its position as a leading advocate of open-weight AI development. Unlike closed AI systems, open-weight models publicly release their trained parameters, allowing researchers, enterprises, and developers to study, modify, and deploy them more freely.
The company said recent developments in the AI industry have highlighted the importance of open alternatives.
“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 company spokesperson said.
Founded in 2024 by former Google DeepMind researchers, Reflection AI has positioned itself as an alternative to frontier AI companies that rely on proprietary models. The startup described the agreement as one of the largest publicly announced infrastructure commitments dedicated to open AI development.
The deal comes at a time when access to advanced AI chips has become one of the industry's most valuable competitive advantages. As demand for computing power continues to surge, companies are increasingly seeking long-term agreements to secure access to next-generation hardware.
The Colossus facility was originally developed by xAI, Elon Musk’s artificial intelligence venture, before becoming part of SpaceX’s broader AI infrastructure strategy. As SpaceX accumulated significant reserves of high-end Nvidia hardware, the company began offering computing capacity to external AI developers, creating a new revenue stream beyond its traditional aerospace and satellite businesses.
For Reflection AI, the agreement provides immediate access to the computing resources needed to train increasingly capable AI models at scale. For SpaceX, it further strengthens its growing position as a major supplier of AI infrastructure to some of the industry's most ambitious model developers.
The partnership also reflects a broader shift within the AI sector, where access to computing power is becoming just as critical as talent and algorithms in determining which companies can compete at the frontier of AI development.