AMD Shakes Up AI Landscape by Acquiring Untether AI's Talent Trove
AMD Acquires Entire Untether AI Chip Team - Cessation of U Canada AI Inference Product Support Following Takeover
AMD's latest move in the AI sphere is something extraordinary - they've nabbed the entire engineering crew from Untether AI, leaving the AI inference processor speedAI and imAIgine SDK on ice, according to Untether AI's declaration. This unusual arrangement, more akin to a talent poach, will likely disrupt the AI landscape.
"AMD is diving headfirst into a strategic agreement to get a hold of a brilliant gang of AI hardware and software engineers from Untether AI," AMD declared in a statement to CRN. "This transaction tosses a world-class team of engineers into AMD's lap, priming them to level up our AI compiler and kernel development prowess, and boost our digital and SoC design, design verification, and product integration capabilities. We can't wait to open arms and welcome the team's one-of-a-kind expertise to AMD."
Untether AI shines in the realm of AI chips specifically designed for AI inference. While bigwigs like Nvidia's Blackwell Ultra or AMD's Instinct MI350 dominate AI model training, speedAI outperforms them in inference tasks and boasts greater energy efficiency. They achieve this by situating the processors next to the memory, reducing latency and power consumption.
It's puzzling to estimate Untether AI's client base, and how they will cope with this transformation, given that AMD did not set its sights on Untether AI's assets. This outcome leaves companies that purchased Untether AI's products holding the bag.
A massive hiring spree like this suggests that AMD is not stopping at AI GPU raw power. AMD just announced its acquisition of Brium, a firm focused on AI inference optimization, only a day before this bombshell.
Justin Kinsey, President of semiconductor recruiting firm SBT Industries, delivered his thoughts on the matter in a LinkedIn post. "AMD's acquisition of Untether's engineering gang is solid proof that the GPU vendors have figured out that model training is old news, and that a dip in GPU revenue is just around the corner." Although Kinsey's prediction is audacious, he affirms that this trend has been brewing for the past six months.
As the AI power bill continues its upward march, companies are on the lookout for more efficient alternatives. Present AI GPUs, which consume hundreds of watts, are perfect for training workloads, but they're power-hungry when it comes to AI inference. If AMD can deliver a chip tailor-made for this application, it might just topple Nvidia's throne, especially as the AI industry matures.
Join the AI Revolution! Sign up for Tom's Hardware Newsletter
Stay up to date with Tom's Hardware's latest news, in-depth reviews, and more!
Follow Tom's Hardware on Google News to get our cutting-edge news, analysis, and reviews straight in your feeds. Don't forget to click the Follow button.
AMD's strategic acquisition of Untether AI's highly skilled engineers marks a significant shift in their artificial-intelligence strategy, as they aim to strengthen their AI compiler and kernel development prowess and boost digital and SoC design capabilities (technology). This move might disrupt the AI landscape, particularly with a focus on improving efficiency in AI inference, an area where Untether AI's speedAI outperforms current industry leaders (artificial-intelligence).
