Menlo Park, 6 October 2026 – Meta’s deployment of its Muse personal artificial-intelligence agent on AMD EPYC server processors is strengthening the investment case for central processing units as AI demand expands beyond model training into persistent user-level computing. The development broadens the infrastructure opportunity for AMD while highlighting the substantial operating cost required to provide autonomous agents at consumer scale.
Muse gives each agent a sandbox using two dedicated AMD EPYC Turin cores and 8GB of memory. OpenAI’s competing Dots agents have also been observed running on AMD EPYC hosts, although with different resource allocations. The architectures show that agentic services require general-purpose compute to manage browsers, operating systems, tools and persistent tasks even when specialised accelerators handle parts of model inference.
Unlock the Full Article
This article is exclusive to The Ledger Asia Subsribers / PAID members.
Already have an account? Log in here

