Nvidia is talking about buying its way into the open-weight model game. The chipmaker is in early discussions to deepen its investment in Reflection AI — or acquire the startup outright — the Financial Times reported on Saturday, citing people with direct knowledge of the matter.
The talks come just five days after Reflection introduced Beam, its first open-weight model, and underscore how aggressively Nvidia is moving beyond selling GPUs into owning the software stack that runs on them. Nvidia has already poured $800 million into the startup as a major strategic backer, according to the FT.
Nothing is final. The people briefed on the discussions told the FT the talks are at an early stage, a deal could be reached in the coming weeks, and they could just as easily collapse. Reflection declined to comment; Nvidia did not immediately respond.
The talks so far
The shape of any deal is wide open. According to the FT report, carried by Reuters on October 10, the transaction could be a full acquisition, a larger equity investment, or an acqui-hire — an arrangement where Nvidia would hire Reflection’s staff and license its technology instead of buying the company whole.
That last option matters for a reason. An acqui-hire could sidestep the lengthy regulatory review a full acquisition would invite, a pattern Nvidia has seen before with its collapsed $40 billion Arm deal in 2022. With the Trump administration’s Super Intelligence Force actively pressing AI firms over security disclosures, a hire-and-license deal looks like the path of least political friction.
The FT said it could not establish the terms under discussion. That leaves open a wide range, given that CEO Misha Laskin told CNBC in April that Reflection was raising fresh capital at a pre-money valuation of $25 billion. That figure reflects the last funding round, not a proposed price tag.
Why Reflection AI matters
Reflection isn’t a household name, but its pedigree is elite. The startup was founded in 2024 by former Google DeepMind researchers Misha Laskin and Ioannis Antonoglou, and it builds tools that automate software development — one of the fastest-growing real-world uses for AI, and the corner of the market where coding agents are doing actual work rather than demos.
Beam, launched Monday, is designed to compete in coding and agentic tasks against lower-cost Chinese models like DeepSeek and Kimi, which have been winning on price and openness. Reflection says it will release the model’s weights and related materials later this month.
That’s the detail Nvidia is chasing. Beam’s weights aren’t public yet, and the GPUs behind its training run were reportedly Nvidia’s. A deal would tighten that loop: Nvidia hardware, Nvidia-backed weights, Nvidia customers.
The open-weight chessboard
Nvidia’s interest fits a broader land grab. Meta has anchored its strategy to open models with Llama, and Chinese labs keep shipping cheap open-weight alternatives that undercut Western prices. An in-house or closely held open-weight team gives Nvidia a model-side answer to customers who want weights they can actually download and run — ideally on Nvidia hardware.
It’s also a hedge. Nvidia sells shovels, but the companies making the biggest AI revenue gains are the ones with models people pay for directly. Reflection’s coding tools point at enterprise budgets, not just research curiosity. If agentic coding eats developer tooling the way the cloud ate the data center, owning the model team is worth a lot more than $800 million.
And there’s a talent angle nobody says out loud: Laskin and Antonoglou are DeepMind alumni with exactly the kind of systems-and-research combination Nvidia has been hoarding. The FT noted Nvidia could also deepen the relationship through a computing deal, but people don’t talk about acqui-hires when all they want is more H100s.
Why this matters
Watch what happens to Beam’s weights. If they land on schedule later this month under an open license, Reflection keeps its independence play alive and Nvidia’s talks are about influence, not control. If the weights slip or the license tightens, that’s your signal the deal moved toward acquisition.
Either way, the timing is deliberate. Reflection just shipped its first model, and Nvidia just watched its biggest customers — OpenAI, Anthropic, Google — all start designing or diversifying away from pure Nvidia dependence. Paying up for the model layer now is cheaper than competing with it later.
Quick answers
Is Nvidia acquiring Reflection AI? Not yet. The Financial Times reported on October 10, 2026 that Nvidia is in early talks to deepen its investment in Reflection AI or acquire it outright, but the discussions could still fall apart and no terms have been established.
What does Reflection AI make? Reflection builds tools that automate software development. Founded in 2024 by former DeepMind researchers Misha Laskin and Ioannis Antonoglou, the startup launched its first open-weight model, Beam, on October 5, 2026, and plans to release the model weights later in October.
How much is Reflection AI worth? CEO Misha Laskin told CNBC in April 2026 that the startup was raising fresh capital at a pre-money valuation of $25 billion. The FT said it could not establish the terms under discussion, so that valuation is not necessarily the proposed deal price.
Why would Nvidia want to buy an AI model startup? Nvidia is already a major backer with $800 million invested, and Beam targets coding and agentic tasks where lower-cost Chinese models like DeepSeek and Kimi are gaining ground. Owning the model team would put Nvidia on both the hardware and model sides of the open-weight market.
Sources: Financial Times (via Reuters), Reuters, Runtime Wire.