AMD CTO: ‘AI across our entire portfolio’

From PCs to the largest servers, AMD envisions AI integration throughout its products. This move parallels the internet's initial expansion, signaling a transformative era. $400 billion TAM by 2027 The AI market's potential is staggering, with AMD eyeing a Papermaster emphasized the need for a vast AI infrastructure, starting with major hyperscalers and expanding to individual business applications. By considering total ownership costs, AMD's strategy considers total ownership costs. With the MI300 accelerator and a diverse tech arsenal, AMD is poised to challenge the current AI compute monopoly, aiming for efficient power per FLOP. AMD's MI300 targets the computational heft needed for training, while inference demands more economical GPU-CPU pairings. Papermaster predicts a market bifurcation along these lines. Confronting power-limited data centers, AMD commits to a 30x power efficiency improvement by 2025.