Nvidia’s reputed AI moat has been built on an all-encompassing hardware and software vision and proprietary tools like its CUDA programming model for its GPUs. For companies like AMD that are hoping to bridge the moat, an embrace of open source tools could be key, and that appears to have been one of the drivers behind AMD’s acquisition of Finland-based Silo AI.
AMD this week announced a definitive agreement to acquire Silo AI, a Helsinki company described by AMD as “the largest private AI lab in Europe,” in a $665 million all-cash transaction expected to close in the second half of this year. The acquisition comes as AMD has invested more than $125 million in a dozen AI firms in the last year. It also is AMD’s third AI-related outright acquisition in less than a year, following the August 2023 purchase of Mipsology and last October’s deal to buy Nod.AI.
What these acquisitions have in common is that they strengthen AMD’s AI software position and its commitment to open source AI and an open AI ecosystem. Like the acquisition of France-based Mipsology, the Silo AI deal also gives AMD a stronger market foothold in Europe, though it is about more than geographic market reach.
“It’s about more than just having a presence in Europe,” said Jack Gold, president and principal analyst of J. Gold Associates in an email exchange with Fierce Electronics. “AMD is looking to compete more effectively with Nvidia, who has put a lot of emphasis on building out a software and services portfolio to enable AI at scale. AMD wants to offer an open source capability that enables enterprises and others to easily deploy AI solutions. It’s especially important to them since Nvidia is mostly about proprietary (e.g., CUDA) assets while the market is moving towards more open source cross platform capabilities to avoid a lock in.”
Gold noted that Silo AI is working on a number of open AI standards and solutions, “so this gives AMD a major talking point and also provides an enhanced ability to offer broader solutions for creating, deploying and managing AI systems. It’s also a reaction to where Intel has been going lately with its timber AI suites/tools, which are also targeted at Nvidia’s proprietariness.”
Among Silo AI’s achievements, the company has extensive experience creating and training AI models, including the Poro multi-lingual (English and Finnish) large language model, which was trained on Europe’s LUMI supercomputer, which itself was built on AMD processors. The company’s customers include Allianz, Philips, Rolls-Royce, and Unilever, and it counts fellow Finnish firm Nokia as a major partner.
“Silo AI has been a pioneer in scaling large language model training on LUMI, Europe’s fastest supercomputer powered by over 12,000 AMD Instinct MI250X GPUs,” said Dr. Pekka Manninen, Director of Science and Technology at CSC-IT Center for Science, Finland. “Together with university collaborators, they have trained state-of-the-art open-source models for EU languages, such as the Nordic Poro and Viking models. We have collaborated extensively with the team in optimizing the software layer, allowing for efficient training of AI models on LUMI.”
Silo AI CEO and co-founder Peter Sarlin will continue to lead the Silo AI team as part of the AMD Artificial Intelligence Group, reporting to AMD senior vice president Vamsi Boppana, according to AMD.
“Across every industry, enterprises are looking for fast and effective ways to develop and deploy AI solutions for their unique business needs,” Boppana said in a statement. “Silo AI’s team of trusted AI experts and proven experience developing leadership AI models and solutions, including state-of-the-art LLMs built on AMD platforms, will further accelerate our AI strategy and advance the build-out and rapid implementation of AI solutions for our global customers.”
Sarlin added, “We have a well-established history of building successful AI products and delivering value to our customers. We look forward to becoming part of AMD to further scale our impact and develop enterprise solutions and AI models that address the most complex challenges with deploying AI at scale today.”