China’s Ministry of Industry and Information Technology issued a development plan for the information and communications industry for the 15th Five-Year Plan period (2026 to 2030) on September 7, and the numbers inside it are not incremental. The plan targets 9,800 exaflops of intelligent computing capacity by 2030 and calls for ¥3.8 trillion (roughly $532 billion) in cumulative information infrastructure investment over the 2026 to 2030 period. It also sets a target of ¥4.1 trillion in total industry revenue by 2030.
Reports circulating in China’s corporate filings and local market commentary have cited 2,185 exaflops (FP16) of intelligent computing capacity as of end-June and rapid year-over-year growth. MIIT’s September 7 plan does not appear to publish that June baseline in the public-facing summary, so treat the “4.5 times” framing as directional rather than definitive until the full plan text is reviewed.
Press reporting on the MIIT plan says it calls for the “orderly deployment” of intelligent computing clusters with 10,000 graphics processing cards or those using 100,000 or more cards, alongside inference computing facilities tailored to different applications. It also emphasizes adapting infrastructure to home-grown computing chips. That last line is where the trading angle sits.
Stocks in Focus
Expansion at SMIC and allied fabs is better read as policy-supported capacity building than a pure end-demand signal, meaning additions can persist even when near-term economics look strained. For SMIC, that is structurally bullish volume, though margin pressure remains a risk if yields stay constrained.
Cambricon is the name getting the most attention. Prior reporting, attributed to Bloomberg and summarized by multiple outlets, has said the chip designer is targeting roughly 500,000 AI accelerators in 2026, including as many as 300,000 units of its Siyuan 590 and 690 chips, built primarily on SMIC’s N+2 (7nm-class) process. Other industry commentary has warned that competition for SMIC’s most advanced capacity could limit how quickly Cambricon can execute a ramp of that size.
AI chip bifurcation is no longer theoretical: Chinese cloud and AI labs are deploying a mix of domestic accelerators, including Huawei Ascend and other local suppliers, creating a parallel cost and performance curve that will keep diverging from the Nvidia and AMD frontier as export controls tighten. For Alibaba, Tencent, and Baidu, the plan strengthens the odds of a government-supported runway for cloud and compute buildouts through 2030, even if macro conditions soften.
On the Western side, Nvidia and ASML face a more complicated read. In August, CGTN reported that China had switched on an AI supercluster built from 100,000 domestically produced chips at the Zhengzhou node of the national supercomputing network. Every cluster that goes live on domestic silicon is one Nvidia cannot sell into, and every meaningful step China takes toward domestic lithography capability narrows ASML’s long-run addressable market in the country.
The Summit Variable
President Donald Trump is expected to host Xi Jinping at the White House around September 24, 2026, according to think-tank and media reporting, and the timing would coincide with the UN General Assembly window. Public U.S. and China readouts from the May 2026 Trump-Xi meeting emphasized trade, while the harder edges of competition in AI, export controls, and digital sovereignty remained unresolved.
That unresolved tension makes the MIIT plan politically relevant this week. Beijing published a document quantifying its intent to build a domestic AI stack at massive scale, weeks before the leaders are expected to meet. Whether chip export policy loosens, tightens, or stays frozen will reset valuations for every name on this list faster than any earnings report. Watch the summit communique language on technology for the first signal.
The Cheat Sheet
- Top Theme: China’s five-year AI infrastructure mandate formalizes a domestic chip buildout that sidelines foreign suppliers and accelerates local beneficiaries.
- Stock to Watch: Cambricon, where the production ramp and policy tailwind converge, but SMIC yield constraints are the binding variable.
- Sector to Watch: Semiconductor equipment and domestic Chinese AI hardware, with diverging trajectories on each side of the export control line.
- Biggest Risk: The expected September 24 Trump-Xi meeting produces tighter chip restrictions, hitting Nvidia and AMD on the day while lifting domestic Chinese names.
- Biggest Opportunity: Any summit outcome that signals export control relief could produce a sharp rally in Nvidia, which has been pricing in continued exclusion from China’s largest data center builds.
- One Thing to Remember: ¥3.8 trillion in state-mandated infrastructure spending over the 2026 to 2030 period does not slow down because a summit goes poorly. The domestic build is policy, not sentiment.
