SpaceX has signed a major computing power agreement with Reflection AI, an open-source artificial intelligence startup, according to a CNBC report.
Under the agreement, Reflection will receive immediate access to Nvidia GB300 chips, which are used to train and run advanced AI models.
Reflection has agreed to pay SpaceX $150 million per month beginning July 1, 2026, through 2029, according to materials viewed by CNBC. If the agreement runs through the end of its term, the payments would total about $6.3 billion.
Either company can terminate the contract with 90 days’ notice after the first three months.
The deal highlights how SpaceX is using its Colossus computing infrastructure to sell capacity to outside AI companies. Colossus was built in part to support Grok, Elon Musk’s AI chatbot and rival to ChatGPT.
Reflection is focused on building American open-source AI models that can compete with frontier systems from OpenAI, Anthropic, and Google while giving governments and enterprises more control than closed systems.
The company said the agreement gives it additional computing capacity to accelerate its work on “American open intelligence.”
Reflection has not yet released a public, open-source frontier model, but it has been gaining traction with government and national security customers. The company is reportedly working with the Department of Energy’s Genesis Mission and has been part of broader Pentagon AI efforts.
For SpaceX, the agreement adds another customer for its compute infrastructure. The company has already reportedly struck compute-related deals with Anthropic, Google, and Cursor.
The transaction also reflects the growing strategic importance of access to advanced AI chips. As demand for frontier models grows, compute capacity has become one of the biggest constraints for AI companies training and serving advanced systems.
