AMD Plans to Take on Nvidia with a $665 Million Acquisition of the Finnish Start-up Silo AI

AMD plans to extend its artificial intelligence services to rival Nvidia, the market leader, by purchasing Finnish start-up Silo AI for $665 million, making this one of the biggest takeovers of its kind in Europe.

The 300-person staff at Silo will employ AMD’s software tools, a California-based company, to create unique large language models (LLMs), the kind of artificial intelligence (AI) technology that powers chatbots like Google’s Gemini and OpenAI’s ChatGPT. With regulatory permission pending, the all-cash acquisition is anticipated to close in the second part of this year.

“This agreement helps us both accelerate our customer engagements and deployments while also helping us accelerate our own AI tech stack,” Vamsi Boppana, senior vice president of AMD’s artificial intelligence group, said.

Dealroom data indicates that this transaction represents the largest buy of a privately held AI start-up in Europe since Google purchased UK-based DeepMind for approximately £400 million in 2014.

The transaction takes place at a time when UK and EU regulators are closely monitoring buyouts by Silicon Valley firms. This year, hundreds of millions of dollars have been raised by European AI start-ups such as Mistral, DeepL, and Helsing, as investors look for a local hero to compete with US-based OpenAI and Anthropic.

One of the biggest private AI labs in Europe, Silo AI is located in Helsinki and provides enterprise clients with customized AI platforms and models. The Finnish business started a project last year to create LLMs in Danish, Swedish, and Icelandic, among other European languages.

AMD’s AI technology is in competition with Nvidia’s, which now holds the majority of the market for high-performance chips. Due to its performance, Nvidia’s valuation surpassed $3 trillion this year as tech companies strive to construct the processing infrastructure required to run the largest AI models. In an attempt to directly compete with Nvidia’s “Hopper” chip series, AMD began to release its MI300 chips late last year.

The acquisition, according to Peter Sarlin, CEO and co-founder of Silo AI, is the “logical next step” in the Finnish company’s plan to establish itself as a “flagship” AI company.

Silo AI is dedicated to providing “open source” AI models that are freely accessible and amenable to customization by anyone. This sets it apart from companies like Google and OpenAI, which support their own exclusive or “closed” models.

In a recent statement, the start-up highlighted the significance of its “Poro” family of open models in “strengthening European digital sovereignty” and “democratizing access to LLMs.”

Meanwhile, antitrust officials in Washington and Brussels are taking notice of the concentration of the most potent LLMs into the hands of a small number of Big Tech companies based in the US.

The acquisition of Silo demonstrates AMD’s goal to swiftly grow its client base and increase engagement with its own product. Silo, which creates customized models for customers, is seen by AMD as a conduit between its “foundational” AI software and the real-world applications of the technology.

As semiconductor makers strive to secure customers’ loyalty to their technology and establish more stable income streams outside the volatile chip sales cycles, software has emerged as a new front in the conflict.

The reason behind Nvidia’s success in the AI space is its multibillion-dollar investment in Cuda, a proprietary program that enables chips meant for computer graphics and video game processing to run a wider range of applications.

Since launching Cuda development in 2006, Nvidia has added several apps and services to its software platform, most of which are targeted at corporate clients who lack the internal resources and expertise that Big Tech companies have to further develop their technology.

More than 600 “pre-trained” models are currently available from Nvidia, making deployment easier for users. The Santa Clara, California-based company launched NIM, a “microservices” platform, this month intending to enable developers to swiftly create chatbots and AI “co-pilot” services.

In the past, Nvidia has given away its software to customers who purchase its chips, but it announced this year that it will start charging for items like NIM.

AMD is one of several companies working on Triton, an OpenAI-led competitor to Cuda that would facilitate easier switching between chip manufacturers for AI developers. On Triton, Meta, Microsoft, and Intel have also collaborated.