AMD has secured all of its 2027 shipments for the Venice server processor, according to industry analysts who reviewed channel data. The company is already taking orders for the 2028 batch, a timeline that suggests confidence in meeting market demand.

The Venice line represents AMD’s sixth‑generation EPYC server chips, built on the Zen 6 architecture and manufactured using TSMC’s 2 nm process. Four variants are slated for release: SP7 in the fourth quarter of 2026, SP8 in the first half of 2027, Venice‑X in the second half of 2027, and Verano also in the second half of 2027.

The SP7 model can house up to 256 Zen 6c cores arranged as 32 cores per chiplet times eight chiplets, delivering 512 threads. A lower‑tier SP7 version offers up to 96 Zen 6 cores — 12 per chiplet times eight — while maintaining a memory bandwidth ceiling of 1.6 TB/s and supporting PCIe Gen 6 at speeds up to 64 Gbps.

Venice‑X is designed for higher‑core-count workloads, featuring up to 96 cores and 192 threads. In a dual‑socket configuration it can reach 192 Zen 6 cores, 384 threads, and a maximum boost clock of 5.15 GHz. It also provides 50 % more L3 cache per core, up to 1,152 MB, thanks to stacked 3D V‑Cache technology. These specifications position the part for AI‑agent sandbox processing.

Verano targets more modest server nodes, offering up to 72 cores and 144 threads, along with 112 Gbps xGMI connectivity and LPDDR5X‑based SOCAMM2 memory.

Analysts estimate that AMD could ship 6.75 million Venice units in 2027, up from roughly 1.25 million in 2026. If the company has already sold its entire 2027 allocation, that would lock in about $51 billion of projected revenue, assuming an average selling price of $7,691 across the 31 available SKUs.

Morgan Stanley noted that AMD’s Venice CPUs are likely to capture a larger share of the server market as AI workloads shift toward more CPU‑centric designs.

Meanwhile, NVIDIA’s internally developed CPUs are not expected to pose a significant challenge to x86 offerings from Intel or AMD, according to a Wukong analysis. The report cites cost inefficiency for certain agent‑style workloads and high migration costs associated with moving existing data‑center software to ARM architecture as key barriers to adoption.