AMD continues to grow its artificial intelligence strength by taking over the entire Untether AI team, a Toronto-based company that was focused on energy-efficient AI inference chips. The takeover is AMD’s ongoing attempt to take on Nvidia in the fast-growing AI chip market.
The deal adds a team of seasoned AI hardware and software engineers to AMD, aimed at strengthening the company’s AI compiler and kernel development expertise. The team will also strengthen the company’s digital and system-on-chip design, verification, and product integration, AMD said.
“AMD has entered into a strategic agreement to acquire a talented team of AI hardware and software engineers from Untether AI,” an AMD spokesperson told CRN. “We are excited to welcome the team’s unique expertise to AMD.”
Although AMD declined to provide financial terms of the deal, the acquisition comes on a series of purchases intended to bolster the company’s AI computing heft. A day ago, AMD said it acquired AI compiler company Brium to enhance AI performance on its Instinct datacenter GPUs.
What Made Untether AI Unique
Established in 2018, Untether AI was working on AI inference chips that were designed for edge environments and data centers.
What differentiated the company was its “at-memory” architecture, which significantly improved performance and lowered power consumption in comparison to conventional chips.
The company’s flagship product, the speedAI240 Slim AI inference accelerator card, achieved remarkable boosts in efficiency. In internal testing, the chip was three times more power-efficient in data center workloads and six times more energy-efficient in edge computing, according to peer-reviewed MLPerf benchmarking results.
The speedAI240 also recorded the best performance of any single PCIe card in both data center and edge categories for the ResNet-50 image classification benchmark, a standard for AI chips in the industry.
Emerging Market Demand for AMD
Former Untether AI CEO Chris Walker, who departed in May, had earlier said there was tremendous demand in the market for power-efficient AI chips. The demand has grown more pressing as Nvidia’s high-performance but power-hungry GPUs are driving data center racks to the extent of requiring up to 120 kilowatts of power.
“We’ve demonstrated that the less traffic there is for data within the chip, the more throughput and power efficiency you have,” Walker said in a previous interview. He added that industrial edge and enterprise deployments have great potential for AI deployment where thousands of small data centers are power-constrained.
The speedAI240 75-watt card had previously found customer acceptance from companies like U.S.-based rugged embedded computing solutions company J-Squared Technologies and India-based AI cloud computing solutions company Ola-Krutrim.
Untether AI had also secured partnerships with industry leaders like Ampere Computing, Arm, and NeuReality in semiconductors and solution providers like Boston, Asa Computers, and Vertical Data.
End of One Era, Start of Another
With the acquisition finalized, Untether AI will no longer provide support or maintenance to its speed AI products and software development kit imAIgine. Bob Beachler, an Untether AI management executive, noted the change in a release: “While today marks the conclusion of the Untether AI journey, we are also proud of the innovative research that underpinned our work in advancing state-of-the-art AI chip technology.”
The deal is a sign that AMD is trying to catch up with Nvidia in the AI chip market. With applications of artificial intelligence seeping into industries, there has been an upsurge in demand for power-efficient yet high-performance processors.
Companies are seeking alternatives to Nvidia more and more, especially those that are more energy efficient and cost-effective.
For AMD, the Untether AI team acquisition is not talent acquisition alone—it’s about acquiring tried and tested technology and talent that can potentially speed up next-generation AI processor development.
With the AI chip wars heating up, the acquisition makes AMD even better positioned to empower customers requiring high-performance computing without the enormous power needs of the traditional GPU approaches.
The acquisition is a sign that AMD is determined to make itself heard in the AI chip market, one strategic purchase at a time.
