AMD is strengthening its AI chip strategy to compete with industry leader Nvidia by acquiring the AI startup.
AMD said on Wednesday it was acquiring Silo AI, a Finnish startup that calls itself “Europe's largest private AI research lab,” in a cash transaction worth about $665 million that is expected to close later this year.
Related: Nvidia is “slowly becoming the IBM of the AI era,” according to the leader of a $2 billion AI startup.
AMD senior vice president Vamsi Boppana told Reuters the acquisition would not impact AMD's financial results because it “will drive significant growth going forward.”
The $500 million acquisition price eclipses the $125 million AMD invested in AI companies over the past 12 months. AMD also acquired France-based Mipsology and Santa Clara-based Nod.ai for undisclosed amounts last year.
If things had gone differently about 20 years ago, Nvidia could have also been acquired by AMD, which came close to acquiring Nvidia in 2006, according to industry sources. The merger reportedly fell through because Nvidia CEO Jensen Huang wanted to lead the combined company.
AMD CEO Lisa Su. Photographer: Annabel Chee/Bloomberg via Getty Images
Nvidia currently dominates the AI chip market with between 70% and 95% share, but competition is growing from all sides. Major chipmakers like Intel and AMD are stepping up their efforts, and startups like Tenstorrent are joining the fray. At the moment, AMD is seen as a close second to Nvidia.
AMD's newly acquired Silo AI employs more than 300 AI professionals, including more than 125 AI scientists with PhDs. The startup operates across North America and Europe and creates both open source AI models and custom AI solutions for clients such as Unilever and Philips.
Related: According to insiders, Nvidia CEO Jensen Huang turned down a merger offer early on in the company's history, and here's why:
In a press release, Boppana highlighted Silo AI's deep AI talent and the team's experience building AI on AMD platforms. He said the acquisition will “further accelerate” AMD's AI development, a statement echoed by AMD CEO Lisa Su.
AI development can cost millions, or even billions, of dollars. Anthropic CEO Dario Amodei said in June that it currently costs about $100 million to train an AI model, but “there are models we're training today that could cost more than $1 billion.”
Nvidia's biggest customers — Microsoft, Meta, Amazon and Alphabet Inc.'s Google — are also thought to be developing their own AI chips, potentially reducing their reliance on Nvidia over time. The tech giant expects its emissions to rise further as it continues to invest in AI, according to an environmental report released by Google last week.
Related: How much does it cost to develop and train AI? According to the CEO of an $18B AI startup, here's what it currently costs:
