AMD Outlines Multi-Year Data Center and AI Roadmap Extending to 2030
At its Advancing AI 2026 event in San Francisco, AMD unveiled a sweeping, multi-year technological vision encompassing future chip architectures, rack-scale computing platforms, and specialized neural accelerators engineered to power next-generation artificial intelligence workloads.
Addressing attendees during her keynote address, AMD CEO Lisa Su announced that the company’s upcoming Zen 7 architecture is scheduled to debut in 2028. The new microarchitecture will be introduced via the 7th-generation Epyc processor family, codenamed “Florence,” targeting high-performance enterprise systems and AI data centers. To serve distinct computational needs, AMD plans to offer the architecture in both classic Zen 7 and compute-optimized Zen 7c configurations, with production expected on TSMC‘s advanced A16 or A14 process nodes. In semiconductor fabrication, transitioning to sub-2nm angstrom-class nodes allows chip designers to pack significantly more transistors into silicon while reducing overall energy draw.
The Epyc Florence processor lineup will introduce support for next-generation LPDDR and MRDIMM memory standards, as well as ACE AI Compute Extensions designed to accelerate matrix multiplication operations natively across the x86 ecosystem. These processors will deploy across two distinct hardware platforms: the SP7 socket for high-core-count, flagship performance deployments, and the SP8 socket tailored for value-optimized computing. Both platform tiers will serve as host nodes within AMD’s upcoming “Ferrara” next-generation rack solution, built specifically to drive demanding AI infrastructure.
AMD also laid out its multi-generational graphics accelerator roadmap, highlighting the Instinct MI455X built on the CDNA 5 compute architecture. The company confirmed that its CDNA 6-based Instinct MI500 accelerator will launch in 2027, followed by the Instinct MI600 in 2028. These GPUs will directly feed AMD’s integrated “Helios” rack-scale systems. The first Helios models are scheduled to ship this year featuring Epyc Venice processors and MI455X GPUs, evolving into the Helios 500 in 2027 with Epyc Verano CPUs and MI500 graphics, and culminating in the Helios 600 in 2028 powered by Epyc Ferrara and MI600 hardware.
Looking further into the decade, AMD revealed that development is already underway on its Zen 8 architecture. Intended to power the 8th-generation Epyc chip family, codenamed “Ravenna,” Zen 8 is set to arrive in 2030, reinforcing the multi-year engineering pipelines standard within modern semiconductor manufacturing.
The forward-looking announcements follow recent product debuts across AMD’s server and consumer divisions, including the launch of Zen 6-based Epyc Venice processors and details on “Vera,” an Arm-based processor designed to handle autonomous agentic AI workloads. For desktop PC gaming, the company introduced the Ryzen 7 7700X3D, an 8-core, 16-thread CPU equipped with 96MB of L3 cache using 3D V-Cache technology, which stacks memory directly on top of the processor die to drastically minimize data retrieval latency during gaming loads.
