Open-source artificial intelligence startup Reflection AI has signed a significant agreement with Elon Musk's SpaceX for access to cutting-edge AI computing power. The deal, which commences on 1st July 2026 and runs through to 2029, will see Reflection AI pay £118 million ($150 million) per month for immediate use of Nvidia's latest GB300 AI chips and supporting infrastructure housed at SpaceX's Colossus 2 data centre near Memphis, Tennessee.
This contract, potentially worth up to £5 billion ($6.3 billion) over its duration, marks Reflection AI's first major compute agreement. It positions the firm alongside other prominent AI developers, such as Anthropic and Google, who have also secured substantial access to SpaceX's valuable AI chip resources. While the total value is considerable, it is smaller than the deals previously struck by Anthropic and Google, which were reported at £985 million ($1.25 billion) and £725 million ($920 million) per month, respectively.
Reflection AI, founded in 2024 by former Google DeepMind researchers, has emphasised the strategic importance of this deal for its open-weight AI strategy. This approach involves publicly releasing trained model parameters, offering an alternative to the proprietary 'closed' models developed by some frontier AI labs. A spokesperson for Reflection AI highlighted that "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."
The Colossus data centre was initially developed by xAI, a company founded by Elon Musk and now integrated into SpaceX, for its own AI initiatives. However, as xAI's internal AI development efforts have reportedly faced challenges, SpaceX has begun leveraging its substantial holdings of high-demand AI chips by leasing them to leading AI research organisations. This move underscores the immense capital expenditure required to acquire and maintain the infrastructure necessary for advanced AI development.
The agreement includes a provision allowing either company to terminate the contract with 90 days' notice after the initial three-month period. This flexibility, despite the multi-year term, has been a feature of SpaceX's other compute deals, with Elon Musk previously downplaying the long-term commitments by emphasising the cancellation clauses.
For the UK, such deals highlight the global scramble for AI infrastructure. Access to powerful chips is fundamental for developing advanced AI, which could drive future economic growth and innovation. The UK government and businesses are increasingly aware of the need to foster domestic AI capabilities and secure access to the necessary computing resources, either through national initiatives or international partnerships. The reliance on a few key hardware providers and data centre operators also raises questions about supply chain resilience and strategic autonomy in the burgeoning AI landscape.
The UK's Information Commissioner's Office (ICO) continues to monitor AI developments closely, focusing on data protection and ethical use of AI. While the EU AI Act aims to regulate AI systems based on risk, the UK is pursuing a more sector-specific, principles-based approach. The escalating demand for compute power, as demonstrated by this deal, could intensify discussions around energy consumption and the environmental impact of large-scale AI operations, which may draw scrutiny from regulators and policymakers across Europe.