AMD splits Zen 7 server chips into 3 lines for AI agents in 2028
The dream of one big server chip doing everything is officially dead because AMD decided that swarms of autonomous digital workers need their own bespoke silicon babysitters.
At the Advancing AI 2026 event, AMD revealed a massive overhaul to its server roadmap. While the upcoming EPYC generations built on Zen 5 and Zen 6 stick to the traditional one-size-fits-all model, the arrival of Zen 7 in 2028 will fracture the family into three specialized CPU lines.
The new line-up starts with Florence, which handles general-purpose data center tasks like traditional cloud workloads and virtualization. Right next to it sits Ferrara, designed specifically as an AI Host Node to referee traffic in massive GPU clusters that usually melt standard processors.
The third branch, dubbed Fidenza, targets Agentic Sandboxes to run swarms of autonomous software entities simultaneously. To keep these silicon brains fed, AMD is rolling out new AI Compute Extensions alongside support for MRDIMM and LPDDR memory architectures across its SP7 and SP8 server sockets.
This architectural split sets the stage for 2030, when the company plans to launch its follow-up Zen 8 architecture under the codename Ravenna.
Corporate tech has officially abandoned general computing in favor of hyper-specialized silicon built solely to keep autonomous algorithms running on perpetual life support.
Source: AMD
Comments
This is where the magic happens: AI reads your discussion and rewrites the article based on the most interesting comments. Each strong comment adds points to the meter below. Once the meter is full, the article updates live — no page reload needed.