Record highs show where capital is concentrating

The market's message is fairly clear: this is first a policy-and-survival trade, and only second an AI-growth trade. The Shanghai Composite at its highest since June 30, 2015 is the backdrop, but the more important signal is leadership. The CSI Semiconductor Index surged 6.3 per cent to a record high, while the CSI AI Index climbed 3.2 per cent. Investors are paying up now for the infrastructure layer of self-reliance, rather than waiting for cleaner earnings proof.

Breadth is selective, not broad-based

That distinction matters because the rally is concentrated in technology and advanced hardware, even as weakness in financial, consumer, and property sectors shows how fragile the rest of the market still is. Broader concerns about domestic consumption and growth remain, and Hong Kong faced heavier selling pressure. Capital is not buying China's entire equity story; it is crowding into the parts tied to strategic control of compute and manufacturing.

The split view: early buildout or overcrowded trade?

That is why the next move depends more on policy follow-through than on slogans. Bulls see an early buildout phase: Beijing is pushing to strengthen domestic innovation and reduce dependence on foreign technology, and investors are front-running that shift. Bears see an overcrowded trade. They point to positions heavily concentrated in chipmaking and AI, which leaves the sector more exposed to sentiment reversals or any sign that policy support is losing momentum.

Why the rally has an infrastructure-layer logic

What matters now is the mechanism underneath the move.

Restrictions can turn substitution into a forced buildout

The bull case starts with simple bottleneck logic: when external supply is constrained, domestic substitution stops being a niche theme and becomes a forced buildout. China has already signaled that direction, with policy aimed at strengthening domestic innovation and reducing dependence on foreign technology. In that setup, semiconductors are treated less like a normal cyclical sector and more like a strategic layer for AI and manufacturing self-reliance.

That helps explain why investors may be valuing the sector on durability of demand rather than just near-term margins. The market is starting to treat China's chip and AI complex as a strategic compute stack, with semiconductor manufacturing and advanced hardware viewed as central to the country's longer-term strategy. Once a sector is seen that way, capital allocation can follow different logic: sustained capex, policy support, and faster adoption in protected markets.

China's Semiconductor Rally Is Betting Big on AI Self-Reliance-But the Next Move Depends on Policy, Not Hype

From substitution to capex to adoption

This is where the S-curve argument gets interesting. Restriction creates the initial demand shock. Domestic foundries, equipment makers, memory players, and related suppliers then absorb some of that demand. Over time, that can become a real adoption curve if capex stays firm and technology improves.

Globally, the semiconductor complex has already shown what can happen when AI demand meets a capital-intensive supply chain: the Philadelphia Semiconductor Index surged by over 200% from 2020 to 2024. In China, a similar instinct is showing up, with the CSI Semiconductor Index surging to a record high and the CSI AI Index climbing alongside it as optimism about AI infrastructure intensified. The key point is not the price action by itself, but what investors think comes next: a domestic compute stack built to carry AI workloads inside China.

State backing helps demand, but execution still matters

Bulls argue that state backing can turn substitution into a long-duration build cycle. Bears argue that policy support is not the same as execution. Recent pressure in the sector shows why that debate is still live: during a tech sell-off, positions were heavily concentrated in chipmaking and AI, and one trigger was news that state semiconductor funds were reducing their stakes.

That is the real fault line. If state support remains consistent and companies convert it into usable capacity and better yields, the sector can keep rerating as an infrastructure-layer winner. If execution lags or subsidies prove uneven, investors may stop seeing durable infrastructure and start seeing a capex-heavy trade with weak profitability.

How investors can participate-and where the trade can break

The practical question is access, not ideology.

The cleaner passive route

For most investors, the cleaner route is not stock-picking Beijing's next winner. It is buying the index. The Global X China Semiconductor ETF is built to track the FactSet China Semiconductor Index, which can be a more direct way to express the domestic substitution theme than trying to guess which company gets the next order. In ETF form, the exposure is simple and transparent: it gives investors a way to own China's chip complex as defined by the benchmark.

But simple does not mean safe. The same product disclosure warns that the fund has concentration risk because it tracks a single region and a sector-heavy index. Semiconductors can be more volatile than a diversified portfolio, and A-share exposure through Stock Connect adds liquidity, currency, political, legal, and taxation risks.

Entry timing still matters

There is also a reason to be selective on entry. ETF 512760 was at 1.600, inside a 1.029 to 1.869 52-week range, and carried a P/E ratio of 48.15. That is not a cheap valuation, and it suggests the market is already pricing in a strong self-reliance narrative.

The downside risk is real too. Semiconductor ETFs have seen steep drops in bearish periods. This is a high-volatility way to own the theme, so timing and position sizing matter.

What could keep the rally going

What could break it

  • Tech stops leading and semiconductor breakout strength fails.
  • Concentration turns into a stampede, with positions heavily concentrated in chipmaking and AI.
  • State backing appears to wind down, or policy support stops translating into real capacity and adoption.