On October 10, 2026, Investing.com reported, citing the Financial Times, that Nvidia was in preliminary discussions to acquire Reflection AI or increase its investment in the startup. Reuters said it could not independently verify the talks at the time. The account covered discussions, not a completed acquisition.
The report described three possible paths: a full acquisition, additional financing, or an acqui-hire—hiring Reflection AI employees and licensing its technology. It did not identify one as the chosen option.
Reflection AI’s funding history
The reported funding figures belong to separate fundraising periods. Reuters reported on September 10, 2025, that Reflection AI raised $2 billion at an $8 billion valuation. In April 2026, CEO Misha Laskin told CNBC that the company was raising funds at a $25 billion pre-money valuation, according to Investing.com. That was a fundraising valuation, not a reported Nvidia transaction price.
Investing.com also attributed to the Financial Times a prior $800 million Nvidia commitment to Reflection AI. The figure was reported as earlier backing, not as a price for the discussions.
Beam’s reported Nvidia hardware connection
Reflection AI describes Beam as a sparse mixture-of-experts model with 501 billion total parameters and 23 billion active parameters. In this architecture, selected expert components handle each input. The company says Beam is designed for coding, reasoning and agentic workloads. Reflection AI’s announcement also says a four-week high-compute reinforcement-learning run used 10,500 NVIDIA GB300 GPUs and generated more than 100 million rollouts.
Reflection says Beam achieved scores comparable to GLM-5.2 on advanced reasoning benchmarks while using three to four times less inference compute. That is the company’s reported comparison.
