Intel Cuts TSMC Dependence, Launches Core Series 3 CPUs for Faster Production

Intel’s newest Core Series 3 silicon marks the first batch of 2 nm‑class processors that have left the company’s Oregon and Arizona fabs rather than Taiwan’s TSMC, a shift that Intel says will “ease reliance on TSMC” after the majority of its client portfolio was outsourced there in 2024 [Theregister]. The chips are built on Intel’s 18A node—a 2 nm‑class process that entered volume production at the Hillsboro and Chandler facilities late last summer—and are positioned as stripped‑down versions of the Core Ultra Series 3 parts unveiled at CES in January. Most SKUs combine two high‑performance “Cougar Cove” cores with four low‑power “Darkmont” efficiency cores, delivering up to six cores per package. By contrast, the flagship Core Ultra models retain a larger core count and a more aggressive GPU configuration.

The integrated graphics subsystem on the Core Series 3 parts is modest: each chip ships with two Xe³ GPU cores, two fewer than the base Core Ultra offering, and a “somewhat pokey” neural‑processing unit (NPU) rated at 15‑17 INT8 TOPS of local AI throughput. Intel notes that the combined CPU, GPU, and NPU can achieve up to 40 TOPS on the platform, but not concurrently, a limitation that disqualifies the parts from Microsoft’s Copilot+ certification [Theregister]. Clock speeds remain competitive, however: boost frequencies range from 4.3 GHz on the low‑end Core 7 304 to 4.8 GHz on the top‑spec Core 7 360, while GPU clocks sit between 2.3 GHz and 2.6 GHz. Memory support is also scaled for cost‑sensitive markets, with either up to 48 GB of LPDDR5‑7467 MT/s or 64 GB of user‑serviceable DDR5‑6400 MT/s, but only a single memory channel, halving bandwidth relative to the higher‑end Core Ultra line.

Performance claims focus on a clear upgrade path from Intel’s legacy 11th‑gen Tiger Lake CPUs. In Cinebench 2024 testing, Intel reports that the 15 W Core 7 360 delivers 47 % higher single‑thread and 41 % higher multi‑core scores than the 5.5‑year‑old Tiger Lake reference, despite the reduced core count [Theregister]. The chips also integrate a full I/O suite: two Thunderbolt 4 ports, two USB 3.2 ports, Wi‑Fi 7, Bluetooth 6, and a surprising number of legacy USB 2.0 connectors. Intel says more than 70 OEM partners have already qualified the silicon, with the first notebook designs slated for release on Thursday. The same silicon is being marketed for low‑power edge boxes, where Intel positions it against Nvidia’s Jetson Orin Nano for workloads such as object detection, image classification, and video analytics; edge shipments are expected later this quarter [Theregister].

From a strategic perspective, the move underscores Intel’s desire to re‑establish a domestic supply chain for its mainstream products. By fabricating the Core Series 3 line in the United States, Intel can sidestep geopolitical risks associated with Taiwan‑based foundries while still leveraging its own advanced 18A process. The trade‑off is a modest reduction in raw compute density—fewer GPU cores, a single memory channel, and a lower‑spec NPU—yet the company argues that the overall platform performance (up to 40 TOPS) remains sufficient for the target segments of budget laptops and edge AI devices. Analysts have not yet quantified the cost impact of the domestic fab versus TSMC outsourcing, but Intel’s public messaging emphasizes “home‑grown” production as a competitive differentiator in a market where supply‑chain resilience has become a purchasing criterion.

Finally, the Core Series 3 launch arrives amid intensifying competition from Apple’s upcoming MacBook Neo, which also employs a six‑core big‑little architecture. While Intel’s chips offer higher boost clocks and a broader I/O set, Apple’s silicon is expected to benefit from tighter integration with macOS and a more mature AI accelerator. Intel’s positioning therefore hinges on price‑to‑performance and the ability to ship quickly from its U.S. fabs. If the early OEM implementations meet the promised performance uplift over Tiger Lake, the Core Series 3 could become a viable bridge for customers awaiting the next generation of higher‑end Core Ultra or competing AMD offerings, while simultaneously reducing Intel’s exposure to external foundry capacity constraints [Theregister].