$AMD| Inference TAM is 5-10x overnight🧵 Not Financial Advice! DYOR! Generative AI has transitioned from experimental to production at unprecedented speed. Training large models incurs massive one time costs, but inference the ongoing process of running models on new inputs represents the recurring, usage driven expense that grows with user adoption. AMD’s EPYC Venice and Turin generations, with unmatched core density (up to 256 cores/socket), combined with execution advantages in supply and TCO, position the company to capture a leading share of the expanding server CPU opportunity. The directional evidence strongly supports substantial uplift in inference-related CPU demand, with the highest-core-per-socket designs delivering the greatest advantage. As of 2025-2026 data, inference
在 X 查看 ↗