Intel's Future CPU and Foundry Strategy as Described by Q3 2025 Earnings Call
At the Q3 2025 earnings call, Intel CEO Lip-Bu Tan fully elaborated on which companies have made future plans along the lines of client, server, and foundry; such discussions prompted the company to confirm upcoming launches for future CPUs such as Panther Lake and Nova Lake as well as go on to talk about the ever-important 18A manufacturing process implementation within future products.
Client CPU Roadmap: Panther Lake and Nova Lake
Intel reiterated its roadmap for client CPUs by stating that the first Panther Lake SKU is on schedule to be released by the end of 2025, with a full reveal at CES 2026 and PCs available to consumers by January. Additional Panther Lake SKUs will follow over the first half of 2026.
Further into the future, Nova Lake-S will go amid a launch on high-end desktop families in place of the next Arrow Lake refresh due in the second half of 2026. Significant architectural upgrades will be seen in Nova Lake, boasting up to 52 cores and new Xe3P Arc integrated graphics.
Server CPU Updates: Granite Rapids, Clearwater Forest, and Beyond
Granite Rapids (Xeon 6 P-Core) CPUs, meanwhile, are witnessing very healthy interest from customers on the server side in their having been deployed into every major hyperscaler. The company expects continued strength from its Data Center AI (DCAI) group to offset any projected downturn in the Client Computing Group (CCG) next quarter.
Key updates on the 18A process are as follows in the Xeon family roadmap:
- Xeon 6+ Clearwater Forest: Expected launch date: Approximately mid-2026 at the time of IoT server hijack.
- Xeon 7 Diamond Rapids: This would be a future P-Core family also on 18A.
- Coral Rapids: A next-generation SKU which would bring back Simultaneous Multi-Threading (SMT) to the Xeon architecture to drive higher performance.
Intel Foundry (IFS) - Processes 18A & 14A
Particularly, this call put strong emphasis on the advancement of Intel foundry. The 18A is ramping and making high-volume manufacture at Fab 52 in Arizona. One of the greatest confirmations was that Intel also explained that this family process will hence form the pillars on which the following three generations of client and server products will be based on.
The company is also proceeding with the work on 18A-P, a performance-optimized option, which is promising a 10 percent improvement and is initiating technology definition for the next 14A node. Lip-Bu Tan emphasized a disciplined approach to foundry investment, adding capacity only when there is committed external demand.
GPU and AI Accelerator Plans
Intel also has mentioned its strategy on GPUs and AI accelerators concerning the next few generations, which include high memory and bandwidth features that are more optimized for inference. Among those builds by Intel, there is also Crescent Island, the first Xe3P accelerator designed for AI inference workloads. This will launch in the near future, which is set to be succeeded by other generations in an annual cadence.
Source: Information acquired from Intel's Q3 2025 Earnings Call.
