Intel’s next-generation client processor family, Nova Lake, will ship with a significantly more capable integrated graphics engine and split its manufacturing across two advanced nodes, according to a new report from Wccftech published in July. The leak describes a line-up that pairs a 12-core Xe3P iGPU with a dual-sourcing approach that sees some tiles fabbed on Intel’s own 18A-P process while others land on TSMC’s N2P.

Nothing about Nova Lake is official yet—Intel has not confirmed the architecture’s existence—but the details align with the company’s public roadmap for future client platforms and its aggressive push to regain process leadership. If accurate, the strategy would mark the broadest use of third-party foundry silicon in a mainstream Intel client product, directly affecting what millions of Windows PC buyers will find inside their next laptops and desktops.

What the Leak Actually Reveals

Wccftech’s report, which cites unnamed sources familiar with Intel’s plans, outlines several concrete claims about Nova Lake:

  • The integrated GPU will move from the current Meteor Lake / Arrow Lake Xe-LPG and Lunar Lake Xe2 architecture to a new Xe3P design. Early leaks suggest it will house 12 Xe cores, a substantial step up from the 8 Xe cores found in Meteor Lake’s top-end Arc graphics tile. Xe3P is expected to be a performance-tuned variant of Intel’s Xe3 IP, possibly sharing DNA with the discrete Battlemage—or “Celestial”—families.
  • The compute tile—the main CPU, GPU, and NPU complex—will reportedly be manufactured on Intel’s 18A-P node. This is the company’s most advanced internal process, featuring RibbonFET gate-all-around transistors and PowerVia backside power delivery. Volume production on 18A is expected to start in late 2024 for early products, but Nova Lake is a 2026 platform, giving Intel time to refine the node.
  • Other tiles, possibly including the SoC tile or an I/O die, will be manufactured by TSMC on its N2P node. N2P is TSMC’s second-generation 2 nm-class technology with enhanced power and performance, set to enter risk production in late 2025 and volume production in 2026. The split suggests Intel will use TSMC’s most advanced node for functions that benefit from extreme density or low leakage, while keeping the critical compute engine on its own latest technology.

None of these details have been verified by Intel. A company spokesperson declined to comment on the Wccftech story, citing a policy of not discussing unannounced products. That is standard, but it also means everything here is a rumour until an official announcement.

What Nova Lake Means for Your Next Windows PC

For everyday users who buy a mid-range or premium laptop in late 2026, the most noticeable change would be graphics performance. A 12-core Xe3P iGPU could double the silicon resources available today and, combined with a new architecture and faster memory, might finally deliver a playable 1080p gaming experience without a discrete GPU. Even mundane tasks—video playback, AI-enhanced photo editing, Windows Studio Effects—would see a snappier response.

Power users and IT administrators face a more nuanced picture. A split manufacturing strategy raises questions about supply stability and driver support. Historically, Intel has used TSMC for certain tiles (like the GPU tile in Meteor Lake) and managed to integrate them into its client platforms without major issues. Nova Lake would extend that to a more advanced node, but Intel’s ability to fine-tune drivers across two different foundry processes will be tested. The move could also affect system-on-chip power characteristics: the combination of Intel 18A’s high-performance focus and TSMC N2P’s efficiency may yield an unusually wide dynamic range, potentially improving battery life under light loads while boosting peak performance.

For developers, the split may be transparent. Nova Lake will run standard x86 code and, if Intel sticks with its hybrid architecture, could mix Performance-cores and Efficiency-cores in new ways. The larger GPU will be a welcome target for AI inference and media processing workloads that already tap into Intel’s oneAPI and OpenVINO stacks.

How Intel Arrived at a Two-Foundry Future for Client Chips

Intel’s journey toward external manufacturing began years ago, but it accelerated dramatically after the company’s multi-year struggle to move beyond 10 nm. In 2023, Meteor Lake became the first mainstream client processor to use a TSMC tile (the 5 nm-class N5 GPU tile). The following year, Lunar Lake went all-in: its compute tile was built on TSMC’s N3B, with only the Foveros packaging handled in Intel’s own facilities. Arrow Lake followed with a mix of TSMC N3B for the GPU tile and Intel 20A for the CPU tile. Nova Lake now appears to reverse the trend slightly, bringing the compute tile back to an Intel node (18A) while still leaning on TSMC for other critical pieces.

The timing matters. Intel’s 18A-P is a “performance” flavour of the base 18A node, presumably offering higher drive current at the expense of density or leakage. It may be analogous to TSMC’s N2P or to Intel’s own historical “P” variants like Intel 4-P. If Intel can ship 18A-P in volume by early 2026, it would demonstrate that the company’s five-nodes-in-four-years plan has kept it competitive—at least on paper—with TSMC’s best. At the same time, using N2P for a companion tile hedges against any delays or capacity constraints on 18A. It is a pragmatic, if complicated, strategy.

The Xe3P iGPU also has a lineage. Intel’s Xe graphics architecture debuted with Tiger Lake in 2020, then scaled up to the discrete Arc Alchemist cards and the integrated Xe-LPG in Meteor Lake. Battlemage (Xe2) is expected to arrive in discrete form soon, with Lunar Lake integrating a scaled-down version for ultraportables. Xe3, also called “Celestial,” is the architecture slated for 2025-2026 discrete cards. Putting a “P” (performance) variant of Xe3 into an integrated package would be a statement: Intel believes its graphics IP can now rival entry-level discrete GPUs even without dedicated VRAM.

What You Should Do Now (Spoiler: Very Little)

If you are in the market for a PC right now, the Nova Lake leak should not change your purchasing decision. The platform is not expected until 2026, and current Meteor Lake, Lunar Lake, and Arrow Lake systems are perfectly capable. Waiting more than 18 months for an unconfirmed product rarely makes sense.

For IT decision-makers planning enterprise refresh cycles in the 2026-2027 window, the leak is a useful data point. It suggests that Intel’s next big platform will deliver a meaningful integrated graphics uplift and that a dual-sourced supply chain might be more resilient than a single-node approach. It is worth monitoring official Intel communications, especially at events like Intel Innovation or Computex, for architectural disclosures that would validate or refute the Wccftech report.

Enthusiasts and early adopters should treat the leak as a signal that Intel is serious about catching up in graphics and about using TSMC’s most advanced nodes to stay competitive. But a lot can change between now and a 2026 launch—roadmaps shift, nodes are renamed, core counts are added or subtracted. File this one under “promising but unconfirmed.”

What to Watch Next

Intel’s next public update on its client roadmap is likely during its investor-focused events in early 2025. The company may share more about 18A yield progress and the role of external foundries in future products. Meanwhile, keep an eye on TSMC’s N2P timetable: any delay there could force Intel to reconfigure Nova Lake’s tile split, which might manifest as a change in the core counts or feature set. The Xe3P iGPU story is also tied to the discrete Battlemage launch—if Battlemage delivers on its promise, Xe3P becomes a more credible upgrade. If it stumbles, integrated graphics buyers may need to temper their expectations.