Partner commitments and a Morgan Stanley forecast point to a sharp ramp for AMD's next server chip as AI data centers lean harder on CPUs
AMD says its upcoming EPYC Venice server processors have drawn the strongest early demand of any EPYC generation so far. Partners have committed to more than 240 platforms built around the chip.
As of September 29, 2026, AMD’s 2027 production capacity for Venice was reportedly fully sold out. The company is now said to be taking orders for 2028 delivery. That timeline matters because the full Venice rollout is still scaling through 2027.
What Venice is and who wants it
Venice is the name for AMD’s EPYC 9006 series, a family of processors aimed at data centers and AI host duties. AMD first announced the lineup in July 2026.
The rollout comes in two waves. SP7 models are planned to begin shipping in Q4 2026, while the broader SP8 lineup scales throughout 2027.
Venice is built on TSMC’s 2 nm process and offers up to 256 Zen 6 cores and 512 threads. It also supports 16-channel DDR5 memory and PCIe Gen 6.
Pricing spans a wide range. SKUs run from about $700 at the low end to $14,904 at the top.
Meta is the marquee customer. The social media giant has already deployed millions of EPYC processors from earlier generations. It is also playing a hands-on role with Venice, collaborating on development and helping refine testing and validation.
The numbers behind the hype
Morgan Stanley forecasts that 6.75 million EPYC Venice units will ship in 2027. That compares with approximately 1.25 million units expected to ship in 2026.
That works out to roughly 5.4x growth in a single year.
The same forecast points to around $51 billion in potential server CPU revenue. That figure is based on a blended average selling price of nearly $7,691 per unit.
Why CPUs are back in the AI conversation
The demand surge for Venice is being driven primarily by hyperscale data centers and AI infrastructure, especially agentic AI workloads. Agentic AI refers to systems that plan and carry out multistep tasks, calling tools and coordinating processes along the way. That coordination work, often called orchestration, leans heavily on CPU capability.
What this means for AMD and the server market
Commitments to more than 240 platforms, plus capacity reportedly sold through 2027, suggest the company enters the Venice cycle with strong momentum.
The SP7 launch in Q4 2026 will be the first real test. If those shipments land on schedule, the 2027 ramp gains credibility.
A fully booked year cuts both ways, since sold-out capacity can cap upside if demand exceeds what AMD can produce. Reliance on a leading-edge 2 nm process also ties AMD’s ramp to TSMC’s manufacturing execution. Customer concentration is another factor to weigh. Hyperscalers like Meta drive enormous volumes, but they also hold significant pricing leverage and can shift their spending as AI strategies evolve.
