Reflection secures SpaceXAI compute deal for open-source AI models
Open-source AI gains momentum as SpaceX expands access to large-scale computing infrastructure.
Open-source AI startup Reflection has signed a major compute agreement with SpaceXAI, giving the company access to Colossus 2 data centre capacity as it works to develop frontier AI models.
According to Axios, Reflection will begin paying $150 million per month from 1 July 2026 for access to the infrastructure through 2029. The deal is intended to give the Nvidia-backed startup the computing power needed to compete with leading AI companies.
Reflection is developing open-source AI models at a time when access to advanced chips and large-scale data centre capacity has become a major barrier to frontier model development.
The agreement highlights the growing importance of specialised AI infrastructure providers. Rather than building all capacity internally, AI developers are increasingly relying on large compute partnerships to secure the resources needed for training and operating advanced models.
It also points to SpaceXAI’s expanding role in the AI infrastructure market. The company has been offering access to Colossus data centre capacity to AI developers, turning large-scale compute into a strategic asset within the AI ecosystem.
The deal reflects a broader shift in the AI race, where access to GPUs, power, data centres and long-term infrastructure contracts can be as important as model design or software talent.
Why does it matter?
The Reflection-SpaceXAI deal shows how compute access is becoming a decisive factor in AI competition. Open-source AI developers may benefit from wider access to large-scale infrastructure, but such deals also concentrate strategic power among companies that control chips, energy, data centres and financing. That makes AI infrastructure a governance issue, not only a business or engineering concern.
Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot!
