The Financial Times report, published on Saturday, October 10, 2026 and summarised by SiliconANGLE and PYMNTS, said the discussions were at an early stage. People familiar with the matter said an agreement could come within weeks, but the talks could still collapse. Neither Nvidia nor Reflection has commented publicly.
Three ways a deal could work
- An outright purchase. Nvidia agreed on September 3 to pay $12.9 billion for Hugging Face, so it has recently bought a model-ecosystem company outright.
- A bigger stake, possibly alongside a larger supply of chips and computing power.
- An acquihire, in which Nvidia hires Reflection's staff and licenses its technology without taking control. The FT said this structure would likely avoid a lengthy regulatory review. SiliconANGLE notes the Groq precedent: in December, Groq licensed its technology to Nvidia in a deal reported at $20 billion, and its founder moved to Nvidia while Groq kept running its cloud business.
Where Reflection stands
Nvidia led Reflection's $2 billion round in October 2025 and, per the FT, has invested $800 million in total. Chief executive Misha Laskin told CNBC in April that the latest round closed at a $25 billion pre-money valuation. Laskin and Ioannis Antonoglou founded the company in 2024 after working at Google DeepMind.
The report came five days after Reflection unveiled Beam, a 501-billion-parameter mixture-of-experts model pitched against cheaper Chinese open models such as DeepSeek and Kimi on coding and agentic work. As this digest noted on October 5, Beam's weights had not been released at launch; Reflection promised them later in the month. Its training ran on Nvidia GB300 systems rented from SpaceX under a June agreement worth up to $6.3 billion, at $150 million a month through 2029, and Nebius agreed in July to sell Reflection another $1 billion in capacity. PYMNTS adds that the White House has been counting on U.S. open-weight models to rival China's.
Why this is not just a chip story
Nvidia already supplies the hardware, has agreed to buy the largest open-model hub, and ships its own Nemotron models. Taking control of, or the team behind, America's most prominent open-weight challenger would put a single company at the centre of the U.S. open-model supply chain: training compute, model weights and distribution.
What remains uncertain. Which structure, if any, is chosen; whether Beam's weights and licence would stay open under Nvidia; and whether an acquihire would draw antitrust scrutiny anyway.
Analysis: for developers choosing an open model, the useful question is not who owns the lab but what is guaranteed in writing. Licence terms, weight availability and a published release cadence matter more than a funding headline. Teams betting on Beam should treat it as a promising model whose openness has not yet been proven by a weight release, and keep a second open model qualified.
The FT reports Nvidia is in early talks to buy, acquihire or invest more in Reflection AI, which it has already backed with $800 million and whose 501B Beam model launched October 5 without weights; a deal could come within weeks or collapse, and it would concentrate U.S. open-weight compute, models and distribution in one company.