The Pulse
Nvidia Weighs Buying Reflection AI or Expanding Its $800M Stake
Nvidia is discussing an acquisition, a larger investment or additional computing support for Reflection AI, the Financial Times reported on October 10. The startup recently introduced Beam, its first open-weight model, built for coding and
AI.info Team · Written with AI ·
Nvidia has already invested $800 million in Reflection AI and is now weighing whether to buy the startup or deepen its backing, the Financial Times reported on October 10. The discussions are at an early stage, and a deal could take several forms—or fall apart, according to the report as summarized by Reuters.
Several routes to a closer Nvidia tie
The options under discussion include a full acquisition, a larger equity investment, or a deal to provide Reflection with more chips and computing power. Another possibility is an “acqui-hire”: Nvidia would hire Reflection staff and license its technology rather than purchase the company outright, Reuters reported, citing the Financial Times.
No terms have been agreed. The talks could produce an arrangement in the coming weeks, but the discussions could still collapse, Reuters reported. The accounts do not identify a proposed purchase price or specify how much additional investment or computing capacity Nvidia might offer.
An acqui-hire would differ substantially from an outright purchase: Nvidia could bring in people and technology without buying all of Reflection. Reuters said the structure could avoid a lengthy regulatory review, but no deal structure has been selected.
Beam gives the talks a model-development backdrop
Reflection introduced Beam, its first open-weight model, on October 5. The company describes it as a sparse mixture-of-experts system with 501 billion total parameters, 23 billion active, aimed at coding, reasoning and AI-agent workloads.
In its technical announcement, Reflection said it trained Beam’s reinforcement-learning run on 10,500 Nvidia GB300 GPUs for four weeks. The company said it plans to release the model weights, technical report and tools for running and adapting the model later in October. Those are Reflection’s own descriptions and plans, not independent assessments of Beam’s performance.
That announcement gives the potential Nvidia relationship a practical dimension: Reflection’s model work depends on large-scale computing, while Nvidia sells the chips used to train such systems. Additional Nvidia capacity could support Reflection’s development; an acquisition or acqui-hire would represent a much closer relationship. The report does not establish that the parties have settled on any of those outcomes.
Investment, capacity or control
The reported $800 million investment is already a substantial tie, but it does not by itself establish Nvidia’s ownership share or the terms of any earlier arrangement. The current talks could extend the relationship through capital or computing, or shift it toward Nvidia owning the company or bringing its staff and technology inside.
For Reflection, the timing comes as it prepares to make Beam available beyond its initial release. For Nvidia, the choices range from financing a model developer to securing a more direct claim on its personnel and technology. The concrete unknowns are whether talks produce a deal, which structure the companies choose, and whether Beam’s promised weight release follows later this month.