Nvidia is in early discussions to deepen its ties to Reflection AI, with options that include an outright acquisition, a larger equity stake or an acqui-hire arrangement, according to a Financial Times report. The talks affect the startup's staff and technology, and they follow Nvidia's existing position as a major Reflection investor. Neither company has commented publicly on the discussions.

The FT described several possible structures. Nvidia could buy Reflection outright or raise its stake, potentially along with additional chips and computing capacity. A third path, an acqui-hire, would have Nvidia hire Reflection's employees and license its technology without taking control of the company. The FT said that structure would likely sidestep a lengthy regulatory review.

Nvidia already led Reflection's $2 billion round in October 2025, and the FT reported that Nvidia has invested $800 million in the company overall. Reflection Chief Executive Misha Laskin told CNBC in April that the startup's latest round closed at a $25 billion pre-money valuation. Laskin and Ioannis Antonoglou founded Reflection in 2024 after working as researchers at Google DeepMind.

Word of the talks came five days after Reflection introduced Beam, its first open-weight model. Beam is a mixture-of-experts model with 501 billion parameters. Reflection's initial focus was coding agents, and it is positioning Beam against lower-cost Chinese open models including DeepSeek and Kimi for coding and agentic work.

Beam was trained on Nvidia GB300 systems that Reflection rents from Space Exploration Technologies Corp. under a June agreement worth up to $6.3 billion. Payments of $150 million a month cover capacity at SpaceX's Colossus 2 data center in Memphis through 2029. Nebius Group N.V. agreed in July to sell Reflection another $1 billion in AI capacity.

Reflection is backed by venture capital, counting Sequoia Capital Operations, Lightspeed Venture Partners LP and DST Global LP among its investors. For a US technology audience, any deal would touch a company whose computing needs run through domestic data center capacity and whose models compete partly on cost against Chinese alternatives.

More AI news from TechManNews.