Open‑Model Startup Reflection AI Lands $1 Billion Compute Deal With Nebius

Reflection AI, the fast‑rising U.S. startup dedicated to building high‑performance open‑weight AI models, has signed a $1 billion multi‑year compute agreement with European AI infrastructure provider Nebius.

Under the terms of the deal, Nebius will provide Reflection with dedicated, priority access to Nvidia’s latest generation of flagship AI accelerators, the massive processing power essential for training state‑of‑the‑art foundation models and running large‑scale inference workloads. The announcement comes just weeks after Reflection secured a separate arrangement to leverage SpaceX’s distributed computing network, and underscores how aggressively open‑model developers are racing to lock in the infrastructure capacity they need to compete with closed‑source giants.

Open Models Gain Ground Amid Policy and Access Uncertainty

Reflection has emerged as one of the most closely watched players in the open‑source space, alongside a wave of increasingly capable releases from Chinese AI labs, including Z.ai, DeepSeek, and MiniMax. Their momentum is accelerating amid growing unease across the industry and governments about relying entirely on closed‑source systems: concerns over data privacy and retention, lack of transparency into how models are trained and moderated, and the rising risk of access being revoked overnight due to policy shifts.

Tensions rose sharply last month when the Trump administration pressured leading closed‑model providers Anthropic and OpenAI to restrict distribution of their most powerful frontier systems, citing national security risks. That move raised widespread fears that access to cutting‑edge AI could become arbitrarily limited and has pushed more developers, enterprises, and policymakers to look for independent, transparent alternatives that remain under their own control.

“Everyone saw how quickly the rules can change,” said one industry analyst familiar with the deal. “Open models aren’t just cheaper or more flexible anymore, they’re the only option that feels stable.”

Founded by DeepMind Alumni, Backed by Global Heavyweights

Founded in early 2024 by two former senior researchers from Google DeepMind, Reflection AI has rapidly become one of the best‑funded startups in the sector. The company is currently valued at $8 billion**, with total funding raised approaching **$2.6 billion. Its investor base reads like a who’s who of global tech finance, including Nvidia itself, Sequoia Capital, Lightspeed Venture Partners, and several major sovereign wealth funds.

For Nebius, the former international division of Russian tech giant Yandex, now registered and operating across Europe, the new agreement is another major milestone in its explosive growth as a neutral global infrastructure provider. Earlier this year, shortly after securing a $2 billion strategic investment from Nvidia**, Nebius announced a **five‑year infrastructure partnership with Meta valued at up to $27 billion, followed last year by a multi‑year deal with Microsoft worth up to $19.4 billion.

The company has positioned itself as a critical alternative to U.S.‑based cloud providers, offering access to top‑tier hardware without the same geopolitical constraints that have shaped allocation policies at Amazon, Google, and Microsoft.

Compute: The New Frontline of the AI Race

The flurry of massive deals highlights a defining reality of today’s AI industry: access to high‑end compute is just as critical as model architecture, research talent, or data. For years, closed‑source leaders like OpenAI and Anthropic held a massive advantage through exclusive hardware allocations and priority supply agreements. Now, open‑model developers are securing their own massive capacity and proving that world‑class AI does not have to be locked behind a small number of corporate walls.

“This deal changes the math,” said a source close to Reflection. “For the first time, an open‑model developer has the same scale of compute as the biggest closed players. That means the performance gap could close faster than anyone expected.”

TechCrunch has reached out to representatives for both Reflection AI and Nebius for additional details on the exact timeline, hardware specifications, and full terms of the agreement.