Nvidia backs Reflection AI with $800M to lead in open-source AI, but where’s the model?

Reflection AI has assembled one of the most impressive war chests in the AI industry. The startup, founded in March 2024 by former DeepMind researchers Misha Laskin and Ioannis Antonoglou, raised $2 billion at an $8 billion valuation during its October 2025 funding round. Nvidia led that round with an $800 million contribution. As of mid-2026, Reflection AI has not publicly released a single foundation model.

Big checks, no checkpoints

The investor roster includes Sequoia, Lightspeed, DST Global, Eric Schmidt, and 1789 Capital. Valuation discussions have since pushed the company’s implied worth to somewhere between $20 billion and $25 billion by early-to-mid 2026.

The company also locked in a compute agreement with SpaceX worth up to $6.3 billion, a deal that includes access to Nvidia GB300 chips through 2029.

Reflection AI initially focused on reinforcement learning-based coding agents before pivoting to a broader strategy centered on open-weight models for enterprise and government use. Open-weight means the model’s parameters are publicly accessible, allowing organizations to fine-tune and deploy the technology on their own terms.

The DeepSeek problem

Reflection AI’s positioning as a champion of Western open-source AI came directly in response to China’s rapid advances in the space, particularly DeepSeek’s model releases in early 2025. Reflection AI was explicitly positioned as the American answer to that challenge: build open-weight models that could match or exceed Chinese capabilities while keeping the intellectual property and strategic advantage on US soil.

Domestically, players like Meta with its Llama series have been iterating aggressively in the open-weight space.

What this means for the AI investment landscape

The geopolitical angle adds a layer that pure market analysis can miss. Government interest in domestically developed, open-weight AI models creates a demand floor that doesn’t exist for most consumer tech products. Enterprise and defense contracts could provide Reflection AI with revenue streams that don’t depend on winning a consumer popularity contest.

Nvidia’s $800 million investment provides strong signal, as the chipmaker has better visibility into AI training pipelines than almost anyone. Reflection AI’s $6.3 billion SpaceX deal and its access to next-generation Nvidia chips suggest the company is building capacity for training runs at the absolute frontier.