Reflection AI, a U.S.-based startup specializing in open-weight AI models, has entered into a $1 billion agreement with European AI infrastructure provider Nebius. This partnership grants Reflection AI access to Nvidia’s latest GPUs, enhancing its capacity for training and deploying advanced AI models.
Nebius, formerly the international division of Russian tech giant Yandex, has been expanding its presence in the AI infrastructure sector. Prior to this deal, Nebius secured a $2 billion investment from Nvidia and established a five-year infrastructure agreement with Meta valued at up to $27 billion. Additionally, last year, Nebius entered into a multi-year contract with Microsoft worth up to $19.4 billion.
This collaboration follows Reflection AI’s recent $6.3 billion compute deal with SpaceX, underscoring the company’s aggressive strategy to secure substantial computational resources. These moves are part of a broader trend where AI firms are rapidly forming partnerships to access the necessary compute power for developing and deploying sophisticated models.
Founded in 2024 by two former Google DeepMind researchers, Reflection AI has quickly risen to prominence in the AI community. The company has raised approximately $2.6 billion in funding from notable investors, including Nvidia, Sequoia Capital, and Lightspeed Venture Partners, and is currently valued at $8 billion.
The AI industry is witnessing a significant shift towards open-weight models, especially in light of recent governmental interventions. Last month, the Trump administration pressured leading AI companies to restrict access to their most powerful models, raising concerns about the availability and control of AI technologies. This environment has heightened interest in open-source AI solutions, positioning companies like Reflection AI at the forefront of this movement.
Reflection AI’s strategic partnerships with Nebius and SpaceX not only bolster its computational capabilities but also signal a growing emphasis on open-weight AI models. As the industry grapples with issues of accessibility and control, such collaborations may pave the way for more transparent and widely available AI technologies.
