2 nm Chip Ramp: TSMC Starts Volume Production
TSMC says it has begun volume production of chips on its latest 2 nm-class process, marking the start of a manufacturing ramp that will shape leading-edge capacity planning through 2026. The update was highlighted in a Japan Times report and is also reflected in an undated statement on TSMC’s own process-technology site.
2 nm Chip Ramp: What TSMC Actually Announced
On its “2 nm Technology” page, TSMC states that its N2 process “has started volume production in 4Q25,” and positions N2 as its first generation of nanosheet (gate-all-around) transistors. In the same note, the company also points to packaging- and PDN-adjacent work such as low-resistance redistribution layers and high-performance metal-insulator-metal capacitors as part of the platform.
The company also names Fab 20 and Fab 22 as its 2 nm production facilities. That matters because “2 nm” headlines are often a blur of R&D, risk production, and pilot lines; here, the company is explicitly describing volume output and tying it to specific sites.
Performance, Power, Density, and the First GAA Node
From a technology standpoint, N2 is a bigger step than a routine shrink because it’s TSMC’s first move from FinFETs to gate-all-around nanosheets. In practical terms, TSMC’s published targets (as summarized by Tom’s Hardware from TSMC material) are in the usual triad: roughly 10–15% higher performance at the same power, or around 25–30% lower power at the same performance, alongside an approximate 15% density uplift versus N3E for mixed designs (with higher figures cited for logic-only designs).
None of this guarantees any single customer’s end-product gains—design rules, libraries, SRAM, IO mix, and power delivery quickly become the real story—but it does explain why the 2 nm chip ramp is closely watched by smartphone and HPC/AI silicon teams budgeting for their next major platform turns.
Capacity and Timing: Taiwan First, Faster Ramps Next
Tom’s Hardware reports that initial volume output appears to have started at Fab 22 in Kaohsiung, with Fab 20 expected to follow. It also notes TSMC’s public comments that the ramp should accelerate in 2026, driven by both mobile and HPC/AI demand.
Earlier in 2025, Reuters had already described TSMC’s Kaohsiung expansion activity and reiterated the company’s intent to keep scaling in Taiwan, while balancing overseas investment and customer supply-chain expectations. That backdrop matters because the 2 nm chip ramp is not only a process milestone; it is also an allocation and geopolitical planning exercise for customers that increasingly want second-source options—or at least second-location options—without rewriting their process strategy every cycle.
Looking beyond first-gen N2
TSMC’s cadence doesn’t stop at the initial N2. Tom’s Hardware also points to follow-on variants (including N2P) and the A16 node, with volume production timing discussed in the second half of 2026. If those schedules hold, 2026 becomes the year when the “first 2 nm” story turns into a “how quickly can you scale, tune, and allocate it” story across multiple product classes.
For related context, see earlier eeNews Europe coverage of TSMC’s 2 nm fab expansion: Earlier eeNews Europe coverage.
If you enjoyed this article, you will like the following ones: don't miss them by subscribing to :
