Pranav Hotkar
San Francisco – June 23, 2026 – AI startup Reflection AI has signed a long-term computing agreement with SpaceX to access Nvidia's latest GB300 processors at the company's Colossus 2 data center, underscoring the growing strategic value of hyperscale AI infrastructure as demand for advanced compute continues to outpace supply.
Under the agreement, Reflection will gain immediate access to GB300 GPUs hosted at SpaceX's Colossus 2 facility. The startup will pay USD 150 million per month beginning July 1, 2026, with the contract running through the end of 2029. If maintained for its full term, the deal would be worth approximately USD 6.3 billion, although either party may terminate the agreement after the initial three months with 90 days' notice.
The agreement highlights a broader shift in the AI industry, where access to large-scale compute infrastructure has become as important as model development itself. Rather than building its own hyperscale data centers, Nvidia-backed Reflection is leasing capacity from SpaceX, allowing it to accelerate the development of open-source foundation models while avoiding the multibillion-dollar capital costs associated with constructing AI campuses.
For SpaceX, the deal further expands its role as a commercial AI infrastructure provider. Originally developed to support the company's own AI ambitions, the Colossus platform has increasingly evolved into a shared compute service for external customers. Reflection joins a growing list of high-profile tenants that includes Anthropic and Google, both of which recently signed multiyear agreements for GPU capacity at SpaceX-operated facilities. Together, those contracts position the company as one of the largest independent providers of AI compute infrastructure outside the traditional hyperscale cloud market.
The transaction also illustrates how AI infrastructure is emerging as a new asset class within the data center industry. Instead of leasing only colocation space, operators are increasingly commercializing fully integrated AI platforms that bundle GPU clusters, networking, storage, and power into turnkey computing services. That model enables AI developers to scale rapidly without waiting years for new data center construction or long GPU procurement cycles.
Industry analysts expect demand for AI-ready capacity to remain exceptionally strong as enterprises and model developers continue expanding training and inference workloads. Facilities capable of supporting tens of thousands of advanced GPUs have become some of the most sought-after digital infrastructure assets, driving unprecedented investment in power generation, liquid cooling, and high-density data center design.
For Reflection AI, the agreement provides immediate access to frontier computing resources needed to compete in the increasingly crowded open-model market. For SpaceX, it reinforces a strategy of monetizing hyperscale AI infrastructure, demonstrating how data center capacity has become a critical revenue stream alongside the company's traditional aerospace and satellite businesses.
