Nvidia Reflection AI talks weigh a compute deal or full buyout

The talks could end in one of several ways. The reported options are:

• A full acquisition, with Nvidia buying Reflection outright.
• A bigger stake, adding to the $800 million Nvidia has already invested.
• More computing power, with Nvidia supplying the startup with additional chips instead of cash.
• An acqui-hire, where Nvidia hires Reflection’s staff and licenses its technology without taking over the company itself.

A separate Bloomberg report said Nvidia was discussing a larger investment or a compute offer, and that the two sides hadn’t agreed terms. The FT stressed the talks were at an early stage and could still fall apart. A later investor note described them as advanced, but as of this update neither Nvidia nor Reflection has announced anything.

The acqui-hire route stands out because of regulators. A normal takeover of a high-profile AI company would likely face a long antitrust review, the process where officials check whether a deal reduces competition. Hiring the team and licensing the code can move faster.

Nvidia has used that playbook before, and it hasn’t gone unnoticed. It struck a licensing deal with chip startup Groq, whose chief executive and chief operating officer then joined Nvidia. The US Justice Department is investigating whether that deal was structured to avoid antitrust review. An Nvidia spokesman defended it as “the American system working as designed.”

So if Nvidia picks the acqui-hire path again, expect the same question to follow it.

Reflection AI was founded in 2024 by Misha Laskin and Ioannis Antonoglou, two former researchers at Google DeepMind. It builds AI that can write software and carry out multi-step tasks on its own, known as agents.

On October 5 it released Beam, its first open-weight model, aimed squarely at coding and agent work. Reflection pitches it as an American answer to low-cost Chinese open models such as DeepSeek and Kimi, and the Trump administration has reportedly been hoping Reflection can fill exactly that role.

The startup has also signed computing deals with Elon Musk’s SpaceX and with cloud provider Nebius. Earlier this year, Laskin said Reflection was raising fresh money at a pre-money valuation (the company’s value before new cash comes in) of $25 billion.

Reflection would be the latest piece of a pattern. In September, Nvidia agreed to buy Hugging Face, the popular hub where developers share open AI models, for $12.93 billion. Add the Groq licensing deal and a picture forms: Nvidia wants open models and the tools around them to run best on its hardware.

That logic is simple. Every company that runs an open model on its own servers needs chips, and Nvidia sells most of the chips used for AI. Owning or closely backing the leading US open-weight developer would give it a say in what those models look like and how well they run on Nvidia systems.

For most people, the effect would be indirect but real. Open-weight models are a big reason AI features are getting cheaper to build. A company can run one in its own data center rather than paying per request to a closed provider, and keep sensitive data in-house while doing it.

If Nvidia ends up controlling Reflection, two things could happen. American businesses might get a stronger homegrown alternative to Chinese open models, which matters to firms and governments wary of foreign software. But one chipmaker would also gain more influence over the open AI tools many developers rely on, which is exactly what antitrust officials will look at.

For you, that could show up as cheaper or more capable coding assistants and AI helpers inside workplace software, built on models like Beam.

The next signal to watch is simple. If an agreement lands in the coming weeks, as the FT suggested it might, the structure will tell you a lot. A clean purchase means a long regulatory review. A hiring-and-licensing deal means a quick close and, most likely, fresh questions from Washington about whether big AI companies are buying talent in a way that sidesteps the rules.