AMD takes aim at Nvidia’s AI hardware dominance with Brium acquisition

AMD’s latest acquisition could help reduce Nvidia’s market dominance when it comes to AI hardware.

Semiconductor giant AMD on Wednesday announced it acquired AI software optimization startup Brium. Terms of the deal were not disclosed.

Brium is a startup that appears to be in stealth mode. The startup builds machine learning applications to enable AI inference, the process a trained AI model uses to draw conclusions from new data, across a variety of different hardware options, according to a blog post on Brium’s bare-bones website.

Cutting through that jargon a bit, Brium can help retrofit AI software to work with different AI hardware than it might have been designed for originally.

In a press release, AMD said its acquisition of Brium will help its commitment to “building a high-performance, open AI software ecosystem that empowers developers and drives innovation.”

While AMD is saying that this acquisition helps create a more open AI ecosystem, which isn’t wrong, it seems clear that it’s also meant to help AMD overcome one of its biggest roadblocks: a large percentage of AI software is being designed for Nvidia hardware and chips.

Brium’s sole blog post, which came out in November 2024, talked about the industry’s reliance on Nvidia and called out AMD specifically.

“In recent years, the hardware industry has made strides towards providing viable alternatives to Nvidia hardware for server-side inference,” the blog post reads. “Solutions such as AMD’s Instinct GPUs offer strong performance characteristics, but it remains a challenge to harness that performance in practice as workloads are typically tuned extensively with Nvidia GPUs in mind. At Brium, we intend to enable efficient [model] inference across a range of hardware architectures.”

This is AMD’s fourth strategic acquisition in the past two years toward the company’s goal of fostering an open-source AI ecosystem, according to the press release. The company previously acquired Silo AI (in July 2024), Nod.AI (October 2023), and Mipsology (August 2023).

TechCrunch has reached out to AMD for more information.