
AMD has officially confirmed the existence of its next-generation Zen 7 CPU architecture and the forthcoming Gorgon and Medusa processors, marking a major step in its long-term roadmap through 2027. The announcements came during the company’s latest Financial Analyst Day, where AMD outlined its multi-year strategy for CPUs, GPUs, and NPUs. The event underscored a clear corporate pivot: datacenter and AI workloads are now AMD’s top priorities, with consumer and client markets taking a secondary role in its business structure.
Chief Technology Officer Mark Papermaster revealed that Zen 7 development is already underway, continuing AMD’s model of alternating design teams for each generation. Although no specific release window was shared, Zen 7 will succeed Zen 6—set to launch in EPYC server chips later this year—and will likely drive future Ryzen desktop and mobile CPUs. AMD’s strategy continues to rely on its proven chiplet approach, enabling scalable designs for both performance and efficiency cores under a unified instruction set.
The roadmap also provided a glimpse into AMD’s graphics and AI acceleration plans. Future GPUs will focus on both desktop performance and broader compute applications, while NPUs will gain higher AI TOPS ratings and greater efficiency. Jack Huynh, head of AMD’s Computing and Graphics Group, confirmed the company’s growing focus on “edge AI” applications alongside gaming and workstation markets. His presentation revealed that Gorgon and Medusa are scheduled to launch in 2026 and 2027 respectively, following recent AI-oriented families like Strix Point and Krackan Point.
Huynh emphasized AMD’s commitment to what he called a “no compromise” PC philosophy, declaring that the future of AI PCs will be built on AMD’s platforms. With Zen 7 on the horizon and new processor families positioned to capitalize on the AI boom, AMD is clearly preparing for what it sees as a transformative era in computing—one defined not just by performance metrics, but by the intelligence and adaptability of its chips.

