GoKawiilChallenger AMD’s ability to take data center GPU share from market leader Nvidia will certainly depend on the success or failure of its AI software stack, ROCm. Taking on Nvidia’s CUDA and its enormous installed base, seen by many as the most significant moat of the most valuable company in the world, would seem like an absolutely gargantuan task.
“It’s like climbing a mountain—one step in front of another,” AMD’s VP AI software, Anush Elangovan, told EE Times in an exclusive interview. “Get your direction, lock in, and the rest will follow.”
Anush Elangovan (Source: AMD)
Elangovan joined AMD through the acquisition of his startup Nod.ai two and a half years ago. The startup’s team of 30 had been building AI compilers for five or six years and was well-known as a major contributor to some of the most important AI repositories, including Shark, Torch.MLIR, and IREE. Nod had been working with hyperscalers, enterprises, and startups that were using its compiler-based automation software.
EE Times last spoke to AMD about ROCm just before the Nod acquisition. During that interview, senior VP of AI Vamsi Boppana said that ROCm is AMD’s number one priority, and that the company was aiming to unify AI stacks across AMD’s different hardware types (CPUs, GPUs, and FPGAs).
Partner Content View All
By Lori O'Toole 04.10.2026 By Bluetooth SIG 04.09.2026 By MRPeasy 04.01.2026
Since then, ROCm has had two and a half years of consistent investment, Elangovan said.
“ROCm at that time was a collection of parts,” he said. “It grew up providing [firmware] to ASICs—like, here’s a firmware piece, here’s a firmware piece, let’s tie them together.”
Today’s ROCm team is striving to emulate the Google Chrome team Elangovan was on before founding Nod.
... continue reading