Newsletter Subscribe
Enter your email address below and subscribe to our newsletter
[forminator_form id="25163"]

bloomberg.money.usnews+1.money.usnews+1.Nvidia is in early talks to acquire Reflection AI, a U.S. developer of open-weight AI models, or to put more money into it. The Financial Times reported the talks on Saturday, citing people with direct knowledge of the matter. The talks come days after Reflection released its first model, Beam. They also show how Nvidia, which already owns part of the startup, is moving from selling chips into the business of building the models that run on them.money.usnews+1
According to the FT, the deal could take several forms. One is an "acqui-hire," in which Nvidia would hire Reflection's staff and license its technology instead of buying the whole company. That approach could avoid a lengthy regulatory review. Nvidia could also agree to supply the startup with more chips or raise its equity investment. Several people told the FT a deal could be reached in the coming weeks but warned that the talks could still fall apart. Bloomberg reported that the FT could not establish the terms under discussion.bloomberg+2
Reuters said it could not immediately verify the report. Nvidia did not immediately respond to requests for comment. Reflection declined to comment to the FT.money.usnews+1
Nvidia is already a major financial backer of Reflection, having invested $800 million, according to the FT. Reflection was valued at $8 billion in an October 2025 funding round that included Nvidia. Chief executive Misha Laskin told CNBC in April that the company was raising new capital at a pre-money valuation of $25 billion. The Wall Street Journal News Corp previously reported that Reflection had held talks to raise $2.5 billion at that valuation.runtimewire+2
Former Google DeepMind researchers Laskin and Ioannis Antonoglou founded the company in 2024. Its other investors include Sequoia Capital, DST Global, Lightspeed Venture Partners and 1789 Capital. Reflection has struck billion-dollar computing deals with SpaceX and Nebius Group to get the processing power it needs to train its models.bloomberg+2
On Oct. 5, Reflection unveiled Beam, which it pitches as a Western competitor to lower-cost Chinese open models such as DeepSeek and Kimi on coding and agent tasks. The company said Beam scores about as well as Z.ai's GLM-5.2 on advanced reasoning benchmarks while costing less to run. The figures come from Reflection itself. The model's weights have not been released yet, and the company says it will publish them, along with a technical report, later in October.bloomberg+2
Nvidia sells its own family of models, called Nemotron, but the FT reported that it lacks a frontier-scale model. The FT also reported that Nvidia has called for an ecosystem of U.S.-built open-weight models. Nvidia used an acqui-hire arrangement in its earlier deal with Groq.runtimewire+1