Huawei's Ascend AI chip ecosystem scales up as China pushes for semiconductor independence
"The Kirin 9030 is manufactured using SMIC’s N+3 process, a scaled extension of its previous 7nm (N+2) node," wrote Rajesh Krishnamurthy, an analyst with TechInsights. "However, in absolute terms, N+3 remains substantially less scaled than industry 5nm processes from TSMC and Samsung. While SMIC has executed notable innovations in DUV-based patterning and DTCO techniques, the process is expected to face significant yield challenges, particularly due to aggressively scaled metal pitch using DUV multi-patterning."
TechInsights's conclusion does not reveal much, but at least the wording makes it clear that N+3 is not a true generational leap, but an incremental stretch of SMIC's existing 7nm-class technology. By calling it a 'scaled extension' of N+2, the analysts are signaling that front-end scaling is largely exhausted: fin pitch (FP), contacted poly pitch (CPP), and fundamental transistor geometry have not moved meaningfully. Instead, SMIC is extracting remaining gains through DUV-driven DTCO and back-end-of-line (BEOL) tricks (something the company has already used to make its N+1 a viable alternative for 7nm-class nodes ), rather than through a clean node transition comparable to TSMC or Samsung's 5nm-class processes.
The most important technical warning is the focus on aggressively scaled metal pitch using DUV multi-patterning, as multi-patterning is where the risk concentrates. Scaling BEOL with DUV requires several patterning steps that must line up extremely precisely, and each step adds to line roughness (via misalignment) and defect risk. Unlike FEOL, where performance degrades gradually, BEOL yield can collapse abruptly once overlay and variability budgets are exceeded, which must be a reason why TechInsights explicitly flags yield challenges rather than raw scaling limits.
In general, N+3 demonstrates that SMIC can still push density upwards without EUV, but only at rapidly rising cost and falling yield headroom. Such design decisions place the process firmly closer to a 7nm/6nm-class nodes rather than to 5nm-class nodes. Furthermore, it signals that SMIC's future progress will rely less on further lithographic shrink and more on design discipline via DTCO (which is not limitless, so to speak), innovative high-density libraries, and conservative clocks, because it looks like the foundry can only do this much on the FEOL level. Of course, advanced packaging remains the most viable path for SMIC's scaling going forward, but that does not really work for mobile and other low-power devices.
Get Tom's Hardware's best news and in-depth reviews, straight to your inbox.
Key considerations
- Investor positioning can change fast
- Volatility remains possible near catalysts
- Macro rates and liquidity can dominate flows
Reference reading
- https://www.tomshardware.com/tech-industry/semiconductors/SPONSORED_LINK_URL
- https://www.tomshardware.com/tech-industry/semiconductors/huaweis-latest-mobile-is-chinas-most-advanced-process-node-to-date-despite-using-blacklisted-chipmaker-huawei-kirin-9030-mobile-soc-made-on-smic-n-3-process-but-cant-compete-with-5nm-nodes#main
- https://www.tomshardware.com
- Framework raises DDR5 RAM upgrade prices by 50% amid DRAM shortage — only for Laptop DIY edition, says prices will likely rise again
- HDD prices spike as AI infrastructure and China's PC push collide — hard drives record biggest price increase in eight quarters, suppliers warn pressure will co
- Oracle reportedly delays several new OpenAI data centers because of shortages — tight material and labor supply frustrate expansion plans, possibly by a year or
- Photonic latch memory could enable optical processor caches that run up to 60 GHz, twenty times faster than standard caches — optical SRAM stores and outputs da
- Acer Predator X34 X0 34-inch Mini LED 200 Hz gaming monitor review: Ultra wide, ultra fast, and ultra bright
Informational only. No financial advice. Do your own research.