Semiconductors & Chips

China's semiconductor self-sufficiency campaign continues to advance at multiple speeds, all below an untouched leading-edge ceiling, and the new documents add texture, not a turning point. The one genuinely new central-level signal is an Aug 26 MIIT statement in which planning chief Yao Jun promised "super-conventional measures" to push integrated circuits, machine tools, and other strategic fields toward "decisive breakthroughs" ¹. The phrasing is more muscular than earlier ministerial language, but it is a goal statement, not a policy package: no new funding vehicle, no export-control response, no revised roadmap. The strongest concrete evidence of progress remains provincial: Guizhou reports 81.2% of its intelligent-computing capacity now runs on domestic chips, attached to a total compute scale above 176 EFLOPS, with AI-type computing more than 90% of it ². Together these suggest the center is still setting rhetoric while localities carry the de facto industrial policy.

The operative policy layer is provincial and local financial vehicles. Shanghai's '15th Five-Year' strategic emerging industry plan (Aug 27) again names integrated circuits among the three leading sectors, targets 2.1 trillion yuan in strategic-emerging value-added by 2030, and lists high-end chips, AI-empowered design, advanced packaging, and new device architectures ³. The same day, Hubei Party Secretary Guan Zhi'ou toured Dinglong Semiconductor Materials Park, urging key-materials R&D and a full R&D-production-supply-service chain ⁴. Liaoning held a 'patient capital' matchmaking event on Aug 27, listing semiconductors among emerging sectors ⁵. On the financing side, Mochu (Muchuang) Integrated Circuit closed a Series B of hundreds of millions of yuan, with Lenovo Capital as strategic investor alongside Guangzhou and Guanghua government funds ⁶. Nothing here is a Big Fund-scale intervention; the pattern is that the center articulates goals, provinces and cities assemble capital.

Foundry results reaffirm the mature-node story. SMIC's Q2 sales crossed $3 billion, up 36.1% YoY, with China-region revenue at 90.2% and net profit +261.7%; Hua Hong set a record $717.5 million at 102.8% utilization ⁷. Both guided Q3 higher, and both show consumer electronics as the largest end-market — meaning AI demand is inflating mature fabs as much as it is advancing leading-edge logic. These figures are consistent with the view that state-backed demand is soaking up capacity at 28nm and above, not at the cutting edge.

On the design track, Xiaomi's Aug 24 launch of three Xuanjie chips — the O3 SoC, O100 AI accelerator, and D100 AD chip — confirms product cadence. The O3 starts shipping in September with the Xiaomi 18 Fold and is the first flagship with domestic CXMT LPDDR6 ⁸. But the logic die's leading-edge process still depends on external foundry capacity, so the design win does not change the lithography constraint. Design localization is real; manufacturing independence is not.

Equipment and materials remain in the mature segment. Tuojing Technology topped out its Shenyang base in August, an ~18-billion-yuan project planned for 1,000+ thin-film deposition tools a year, with products already in about 100 domestic chip lines ⁹. That is industrial scale-up in deposition, not a lithography shift. Materials similarly deepen incrementally: Hubei's Dinglong push underscores consumables localization, while Caixin's Tianfu Communication note shows the bottleneck persisting one level up — Q2 profit recovered but upstream optical chips remain constrained ¹⁰.

The cost environment cuts both ways. Storage vendors' price increases — SK Hynix DRAM ASP up roughly 65% QoQ in Q1 and 30% in Q2, NAND up 75% and 55% — squeeze phone margins, and Huawei's Yu Chengdong warns of broad handset price rises ¹¹. Global equity markets are wobbling: memory stocks fell on Aug 24 (Micron -5.83%, SK Hynix -4.92%), the Philadelphia Semiconductor Index dropped 2.70%, and Nvidia logged a seventh straight losing session ahead of earnings ¹². The AI boom that finances China's build-out also inflates its input costs and market volatility.

None of this changes the standing judgment. MIIT's "super-conventional measures" is the closest thing to a central-level signal in this batch, but it lacks implementation detail. Watch whether it translates into Big Fund-style money or binding procurement in the coming quarters; also watch SMIC's Q3 earnings and capex, CXMT's LPDDR6 ramp beyond Xiaomi, and whether Shanghai's IC plan becomes binding. Until then, the summary holds: China's chip story is design-plus-foundry-plus-equipment-plus-materials, moving at different speeds, all below an untouched ceiling.