Intel confirms Nova Lake on 18A-P, plans client capacity shift

By Sophie Lindqvist

What process node will Nova Lake use?

Intel confirmed that Nova Lake desktop processors will become the company’s first client product on the enhanced 18A-P process, according to CEO Lip-Bu Tan on the Q3 2026 earnings call. Tan said: “Nova Lake, our first Intel 18A-P client product, is tracking ahead of schedule, with design wins expecting to be twice those of Panther Lake at launch. We look forward to sharing more at CES in January.” The statement confirms the process choice, but the launch claim concerns design wins, not shipments or revenue. Intel supplied no unit forecast, pricing, or launch date to test that expectation.

Before confirmation, speculation alternated between Intel 18A and TSMC N2P. The current rumor assigns 18A-P to the SoC tile and TSMC N2P to the compute tile. Intel did not confirm that split, any TSMC involvement, die sizes, or production volumes. The earlier reporting considered Intel’s newest node on an SoC tile unlikely, though possible. That uncertainty leaves Nova Lake’s manufacturing arrangement unresolved.

Intel announced 18A-P risk production in June. At Hot Chips in August, the company revealed that next-generation Xeon 7 Diamond Rapids will be its first server products to use 18A-P. Those milestones confirm development activity, not commercial-scale output. Panther Lake and Wildcat Lake currently use base 18A, while Intel claims 18A-P delivers 9% more speed at iso-power, or cuts power 18% at iso-speed. Intel also calls 18A-P a drop-in replacement for 18A, requiring no additional engineering work. These performance figures remain vendor claims, and no independent benchmark accompanied them.

Does Intel have demand despite a weaker PC market?

Wildcat Lake products have appeared in devices including the Framework Laptop 12, while Tan says demand for Wildcat Lake now outstrips Intel’s available supply. Intel did not disclose the unit gap or how long constrained supply will last. The company says 18A products now generate one-third of client CPU revenue. That share rose from less than 5% in Q1, according to Intel. These data points show strong demand for selected 18A chips, not a broad recovery in PC orders. Tan said OEM partners are aggressively managing inventory to match softer sell-through. Intel interprets that as customers buying less to match end-market demand. The statements appear contradictory because they describe different points in the channel: constrained Wildcat Lake supply and cautious OEM purchasing. Neither figure quantifies total PC demand, so the broader forecast deserves more weight.

What does the PC forecast imply?

Intel CFO David Zinsner expects Intel’s client business to decline by more than 20% in the next quarter, a sharper contraction than the roughly 12% decline forecast for the broader 2026 PC market. He said: “We expect the 2026 PC unit [Total Addressable Market] to decline approximately 12%, with second half units down more than 20%. We do not expect a sustained recovery until memory availability improves and pricing begins to normalize.” Intel also plans for the 2027 PC unit total addressable market to decline roughly 10%.

Intel attributes the downturn to memory prices rather than a conventional component shortage. Intel described OEM and IT-department spending pulling back, reducing direct customer orders. That distinction matters because inventory reductions can suppress chip orders even when end users continue buying PCs. The 2026 and 2027 figures are management forecasts, not independent market estimates. Zinsner nevertheless expects Intel to gain share and describes a future PC renaissance despite the memory shortage. He offered no numerical share target, so the long-term claim remains unsupported by measurable commitment.

How much capacity moves to data center?

Intel’s chief financial officer said the company intends to take Panther Lake capacity offline to build more Granite Rapids data-center CPUs. Zinsner said Intel is taking 18A wafers generally assigned to Panther Lake offline to put Intel 3 into Oregon, freeing some capacity for additional data-center wafers. The move shifts production from client silicon toward server products, but Intel gave no wafer count, start date, incremental revenue, or Nova Lake allocation. That omission prevents a calculation of the trade-off between Panther Lake volume and Granite Rapids supply. It also leaves Nova Lake and Wildcat Lake priorities unclear within the remaining 18A capacity. The shift does not resolve near-term PC weakness; it redirects supply while Intel waits for memory prices and availability to improve. The commercial test through 2027 is whether memory availability and pricing normalize before Intel’s planned roughly 10% PC-unit decline further weakens the client business.