Reflection AI secures $1B compute partnership with Nebius, expanding GPU access

Reflection AI, an American startup developing open-weight AI models, has signed a $1 billion deal with Nebius, a European AI infrastructure provider, to access Nvidia's computing hardware.

The agreement marks the second major compute partnership Reflection has announced in recent weeks, following a similar arrangement with SpaceX to tap into its computing capacity. Both deals underscore how AI companies are aggressively locking in computational resources to train and operate their models.

The broader shift toward open models#

Reflection sits at the center of a widening debate about the future of AI. As Chinese competitors release increasingly sophisticated open-source models and concerns grow around data privacy and government oversight, more organizations and developers are reconsidering their reliance on proprietary, closed systems.

The shift gained momentum after the Trump administration pressured major AI firms to limit deployments of their most advanced models, fueling fears that access to certain systems could be revoked suddenly or subject to government control. This uncertainty has accelerated interest in open alternatives that users can deploy and control themselves.

Reflection's backing and scale#

Founded in 2024 by two researchers from Google DeepMind, Reflection has assembled an impressive investor base and valuation. The startup is currently valued at $8 billion and has raised approximately $2.6 billion to date, with backing from Nvidia, Sequoia Capital, and Lightspeed Venture Partners.

Nebius's infrastructure expansion#

Nebius, formerly Yandex's international division, has become a significant player in AI infrastructure deals. Beyond its agreement with Reflection, the company recently secured a $27 billion, five-year infrastructure partnership with Meta and maintains a $19.4 billion multi-year arrangement with Microsoft. It also received $2 billion from Nvidia as part of a broader collaboration.