The most important number in China's chip story is not a chip. It is a wafer count from a foundry that still cannot buy the world's best machines.
SMIC (中芯国际), the mainland's largest contract chipmaker, just reported its strongest quarter on record and gave the clearest signal yet that its advanced-node climb — the N+2 process beneath Huawei's Ascend and a string of domestic AI accelerators — has crossed from laboratory curiosity to paying business.
Key takeaways
- Second-quarter 2026 revenue reached US$3.01B, up 20% sequentially and 36% from a year earlier, with a gross margin of 25.3%.
- Third-quarter guidance points to revenue up 2–4% sequentially and a gross margin of 26–28%.
- Morgan Stanley estimates SMIC's N+2 node capacity will reach about 40,000 wafers per month in 2026, expanding further in 2027.
The quarter that set the record
SMIC reported second-quarter revenue of US$3,005.6M, against US$2,505.5M in the first quarter and US$2,209.1M a year earlier. Gross profit was US$760.6M, lifting the margin to 25.3%, up more than five points sequentially. For the first half, revenue totalled US$5.511B, up 23.7% year on year.
Management guided the third quarter to grow revenue 2–4% sequentially with a gross margin of 26–28%, and said the industrial momentum from AI would persist into the second half, driving broad demand for integrated-circuit manufacturing. The foundry said it would flexibly allocate existing capacity and accelerate qualification of new capacity to ease supply constraints.
What N+2 actually is
SMIC does not market its advanced nodes by nanometre number the way rivals do. "N+2" is its in-house label for an enhanced node widely read by analysts as roughly 7 nanometres, achieved with deep-ultraviolet multi-patterning rather than extreme-ultraviolet lithography. It is the process under Huawei's Ascend 910C and a roster of domestic AI and mobile chips.
By one sell-side estimate, N+2 capacity will reach about 40,000 wafers per month in 2026. Reporting from Chinese financial media, citing sell-side research, suggests the node has passed a yield threshold and entered a profitable phase — the point at which volume finally covers the painful learning curve.
The N+3 question
Beyond N+2 sits N+3, an even tighter node. Publicly, this remains at the risk-trial stage rather than volume production. The honest read is that SMIC's leading edge is N+2 today; anything finer is progress, not product. Claiming otherwise would confuse a research milestone with a shipment.
Why pricing matters as much as process
SMIC's management has said it sees no basis for price cuts this year, and that the upward pricing trend could extend into 2027. In a foundry world where utilisation and price move the bottom line as much as technology, that comment matters. Tight advanced-node supply, driven by AI-adjacent demand, gives the foundry pricing power it lacked during the last downcycle.
The constraint that remains
None of this closes the gap with the leading global foundry, which runs far finer nodes with EUV and far higher yields. SMIC's answer is scale and availability: more usable wafers, qualified faster, for customers who cannot easily buy elsewhere. It is a strategy of compensation, not catch-up.
The record is worth holding next to that caveat: SMIC's scale is, in part, a feature of being the default foundry for a large, partly closed market. That is a durable advantage for the nodes it can reach, and a hard ceiling for the ones it cannot.
The equipment wildcard
The longer game depends less on SMIC alone and more on the toolmakers feeding it. Without EUV lithography, N+2 leans on DUV multi-patterning, and every additional layer raises cost and complexity. Domestic equipment names — in etching, deposition and cleaning — have lifted their share inside Chinese fabs, but the weakest links remain lithography and metrology, where local penetration stays low. That gap is why N+3 remains a risk trial rather than volume. SMIC's record quarter is real; the ceiling above it is set by what its suppliers can deliver, not by demand.
Honest limitations
SMIC's quarterly figures and guidance are taken from the company's official results released through its investor-relations channel. The 40,000-wafers-per-month N+2 capacity figure is Morgan Stanley's estimate reported via technology and financial media, not SMIC guidance; SMIC does not disclose node-level capacity or yield in its filings. The "profitable phase" characterization derives from Chinese financial-media reporting citing sell-side analysis and is presented as reported, not independently verified. All SMIC financials are reported in US dollars and require no currency conversion.
What readers can do now
- Track SMIC's gross margin quarter to quarter — it is the cleanest public proxy for how well the advanced node is absorbing its cost base.
- Read "N+2" and "N+3" as maturity labels (volume vs risk-trial), not as direct nanometre claims.
- Pair foundry data with accelerator and memory results to see the domestic supply chain as one system, not isolated wins.