AMD to Release Zen 7 "Florence" in 2028 and Zen 8 "Ravenna" in 2030
AMD unveiled its CPU roadmap at the Advancing AI event. Following the 256-core EPYC with Zen 6 Venice, Zen 7 "Florence" is set for 2028, and Zen 8 "Ravenna" for 2030, with a focus on AI enhancements and specialized CPU strategies.
AMD revealed its long-term CPU roadmap for the Zen microarchitecture at its Advancing AI event on July 24, 2026. Alongside the announcement of the Zen 6 “Venice”-based 256-core EPYC 9996, CEO Lisa Su disclosed plans for the Zen 7 “Florence” slated for 2028 and Zen 8 “Ravenna” targeting a 2030 release.
Zen 6 Venice Announcement
The highlight of the event was the 256-core EPYC 9996 processor powered by the Zen 6 architecture. AMD also mentioned the “Venice-X,” scheduled for release the following year. According to Tom’s Hardware’s Anton Shilov, these products are part of AMD’s strategy to provide predictable update cycles for data center platforms.
Plans for Zen 7 Florence in 2028
“In 2028, we will introduce Florence,” Su stated. “Florence will deliver the next-generation Zen 7 core and cutting-edge process technology. It will feature a new set of AI computing enhancements to ensure comprehensive AI capabilities and support the latest memory technologies.”
While Su did not disclose specifics about the process technology, core count, memory subsystem, or AI enhancements, she hinted that Florence would not be a standalone processor family. Instead, it would be offered alongside the Ferrara AI Host Node and Faenza Agentic Sandbox. This approach enables AMD to provide purpose-built CPUs for diverse applications rather than applying a single architectural configuration across the server market.
The Path to Zen 8 Ravenna
“And we won’t stop there,” Su continued. “We are already deeply engaged in the development of Ravenna. Based on Zen 8, this will be the 8th generation EPYC family, progressing smoothly towards a 2030 release.”
This long-term roadmap reflects AMD’s broader strategy to establish a consistent product release rhythm within data center platforms.
Evolution of Accelerators
In addition to CPUs, AMD also showcased plans for its accelerator products. The MI500 series of Instinct products, equipped with next-generation HBM, are designed for large-scale domains and incorporate new copper and optical interconnect technologies. This aligns with NVIDIA’s Rubin Ultra and Feynman initiatives.
“MI500 represents the largest intergenerational leap in Instinct’s history. It paves the way for over a 2,000-fold increase in inference throughput within just four years,” Su said.
The CDNA Next-based Instinct MI600 family is also under development, targeting a 2028 release. Alongside the announcement of the MI455X GPU, AMD provided a preview of the MI500X, expected to debut next year.
Industry Impact
The roadmap announcement signals intensifying competition between AMD and NVIDIA in the data center market. By presenting long-term plans for both CPUs and accelerators, AMD aims to facilitate investment decisions for its enterprise customers. Notably, the diversification strategy for the Florence family, tailored for AI inference workloads, could represent a pivotal shift from conventional general-purpose server CPUs.
Editorial Opinion
In the short term, the market entry of the Zen 6 Venice (256-core EPYC) in 2026 is expected to escalate competition with Intel in the server CPU space. This is likely to reignite the core count race, particularly in large-scale virtualization environments of cloud service providers, driving improved price-performance ratios. However, the lack of detailed AI enhancements for Zen 7 Florence remains a point of uncertainty for development partners.
From a long-term perspective, AMD’s strategy to develop CPUs and accelerators as distinct product families promotes architectural diversification across data centers. If the industry moves from single general-purpose CPUs to chips specialized for AI inference nodes and agentic workloads, it could reduce vendor lock-in risks during server procurement.
That said, the editorial team is curious about the lack of specific architectural details for Zen 8 Ravenna. How AMD’s 2030 product will integrate current advancements in AI enhancements and interconnect technologies remains to be seen.
References
- ” AMD reveals CPU architecture roadmap through 2028, following Zen 6 ‘Venice’ launch — Zen 7 ‘Florence’ to debut in 2028 alongside diversified product family, confirms Zen 8 ‘Ravenna’ in development ”, by Anton Shilov — Tom’s Hardware, 2026-07-24T09:27:05.000Z (ARR)
- Source URL: https://www.tomshardware.com/pc-components/cpus/amd-reveals-cpu-architecture-roadmap-through-2028-following-zen-6-venice-launch-zen-7-florence-to-debut-in-2028-alongside-diversified-product-family-confirms-zen-8-ravenna-in-development
Frequently Asked Questions
- When is Zen 7 Florence expected to launch?
- AMD has scheduled its release for 2028. The product family will also include offerings like the Ferrara AI Host Node and Faenza Agentic Sandbox tailored for diverse AI workloads.
- What is AMD's Zen 8 Ravenna?
- Zen 8 serves as the foundation for the 8th generation EPYC family, which is targeted for a 2030 release. Specific details have yet to be disclosed.
- How significant is the performance boost of the MI500 series?
- AMD claims that the MI500 series marks the largest generational leap in Instinct's history, achieving over a 2,000-fold increase in inference throughput within four years, thanks to advancements in HBM and interconnect technologies.
Comments