Intel Goes All-In on 18A at COMPUTEX 2026: Panther Lake, Nova Lake (52 Cores), and 288-Core Clearwater Forest Xeon
Intel is previewing its entire next-era chip lineup at COMPUTEX 2026, with CEO Lip-Bu Tan headlining a full slate of 18A-node products: Panther Lake for AI handhelds, Nova Lake for desktops, and Clearwater Forest for servers. NVIDIA is also at the show with its N1 SoC for Windows-on-Arm.
COMPUTEX 2026 in Taipei is shaping up as the most consequential chip show in years. Intel is using it to preview three major product families, all built on its 18A manufacturing node—the process the company has staked its foundry ambitions on—while simultaneously hosting NVIDIA’s first Windows-on-Arm silicon announcement.
Intel’s lineup spans every computing tier. Panther Lake targets AI handhelds and ultra-portable devices with Arc G3 integrated graphics. Nova Lake is a 52-core desktop CPU, Intel’s response to AMD’s Ryzen Threadripper dominance in the high-core-count workstation market. Clearwater Forest is a 288-core Xeon for data center workloads—a chip count that would have seemed implausible on any Intel node before 18A. CEO Lip-Bu Tan delivers the main keynote on June 2, the first COMPUTEX appearance by an Intel CEO in several years.
18A is the story behind all three. Intel’s previous nodes—Intel 7, Intel 4, Intel 3—were competitive with TSMC’s N5 and N3 roughly speaking, but 18A is the company’s first process where Intel is positioning itself as a genuine foundry alternative. The node uses RibbonFET gate-all-around transistors and PowerVia backside power delivery, the same fundamental innovations TSMC is bringing to its N2 generation. Whether Intel’s 18A yields are competitive at volume scale is the open question; COMPUTEX previews give no yield data, but the willingness to show three product families simultaneously suggests the company has reached manufacturing confidence.
NVIDIA’s N1 SoC is the other major announcement at the show. The N1 pairs MediaTek CPUs with Blackwell GPU cores, targeting Windows-on-Arm devices for on-device AI inference. Specs: up to 6,144 CUDA cores and 128 GB of LPDDR5X memory. This is NVIDIA entering the integrated SoC market for Windows PCs—a direct play for the Copilot+ PC segment that has been dominated by Qualcomm’s Snapdragon X series since 2024. NVIDIA’s inference performance advantage in the Blackwell generation, if it translates to an integrated chip, would make the N1 the most capable on-device AI compute platform in a Windows laptop.
The COMPUTEX 2026 presenter list reads like an industry rollcall: AMD CEO Lisa Su previews Ryzen 9000 series updates, Arm CEO Rene Haas covers the Arm PC platform roadmap, and GIGABYTE and ASUS are both showing next-generation motherboards for both Intel and AMD platforms.
For developers, the immediate relevance is the direction of on-device inference. The convergence of NVIDIA Blackwell (N1), Apple M5, and Intel’s 18A-era NPUs means that by late 2026 or early 2027, a developer laptop will carry more inference compute than a mid-tier cloud VM did in 2023. The tooling implications—local model serving, offline agent runtimes, latency-free inference in dev environments—are significant.
Nova Lake and Clearwater Forest are scheduled for production in H2 2026. Panther Lake devices are expected at retail in Q4 2026.