China's Semiconductor Earnings Are Growing 1,000%. Their Stocks Were Down 40% Last Week. The Disconnect Is the Trade.

While global investors were debating whether the AI bubble had burst, the companies actually building the infrastructure were reporting numbers that suggest the boom is just beginning. The selloff was about positioning. The recovery will be about fundamentals.

The first half of 2026 has delivered one of the most striking divergences between stock-price action and underlying business performance in recent Chinese market history. Between July 13 and July 17, the semiconductor complex—chips, equipment, and materials—was routed. Industry leaders fell more than 10% in a week. Smaller names lost over 40% in a month.

And then the earnings reports landed.

The Numbers: An Industry on Fire

Yuanjie Technology (源杰科技, 688498.SH), a leading designer of optical communication chips, reported first-half 2026 net profit guidance of 600–650 million yuan, representing year-over-year growth of 1,197% to 1,305%. For context, that is more profit in six months than the company generated in its first three years as a listed entity combined.

In the semiconductor equipment segment, the growth is equally extreme. Multiple equipment firms guided to upward revisions in expected growth rates, with some projecting triple-digit percentage increases. Industry research firm Omdia projected that China's domestic semiconductor market would expand 93% this year, driven by AI data-center construction and domestic substitution demand.

Upstream in materials and components, the story is the same. Firms across the supply chain reported order backlogs extending into 2027, with product roadmaps moving from mature nodes to advanced processes at an accelerating pace.

The common thread across every earnings release: AI-driven data-center construction is pulling demand for semiconductor components at a pace that has surprised even bullish analysts.

Why the Stocks Crashed Anyway

If the fundamentals are this strong, why did the stocks collapse? The selloff was driven by three factors, none of which relate to actual demand:

1. Global de-risking and de-crowding. The Philadelphia Semiconductor Index entered a technical bear market on July 17, down 20.2% from its June 22 high. As global thematic funds reduced AI exposure across all geographies, Chinese semiconductor names—held by many of the same pools of capital—were sold indiscriminately.

2. Margin unwinding. Chinese retail investors and domestic funds had crowded into the semiconductor trade during the first-half rally. When the correction began, leveraged accounts faced margin calls, triggering forced selling that fed on itself.

3. Fear of cyclical peak. Some investors worry that the AI capital-expenditure boom is peaking and that 2026 will represent a demand high-water mark. The earnings reports suggest the opposite: order backlogs extend into 2027, and product roadmaps are moving from 1.6T to 2.4T to 3.2T and beyond in optical modules.

"The market's concerns about an AI bubble, overcapacity, and new technology substitution are clearly misread," the head of a listed semiconductor company told Securities Times. "AI computing-power-driven demand has high certainty. The industry is still in its early stages."

The Structural Differentiator: Who Wins

Not all semiconductor companies are created equal, and the earnings dispersion makes this clear. Yuanjie's projected 600–650 million yuan in first-half profit dwarfs smaller competitors by an order of magnitude. The reason is customer concentration at the top of the AI supply chain.

Nvidia, Google, Microsoft, and Meta—the "hyperscalers" driving global AI capex—maintain tightly controlled supplier lists for their data-center components. Chinese firms that have broken through the qualification barriers are entrenched as primary or secondary suppliers. Smaller players struggle to break through and are relegated to lower-margin, lower-volume products. The industry is consolidating around a handful of winners.

The Domestic Substitution Wildcard

Looking further out, a policy shift may deepen the moat of China's semiconductor leaders. At the 2026 World Artificial Intelligence Conference, industry participants highlighted the acceleration of domestic computing cluster construction as a pathway to reduce reliance on foreign chips. Tencent Cloud announced plans to deploy NPO (Near-Package Optics) super nodes in the fourth quarter of 2026, using large-scale domestic computing power to push inference costs to their lowest possible level.

If domestic AI infrastructure spending scales from niche to mainstream over the next three to five years, the companies with existing semiconductor expertise—particularly the equipment and chip leaders—will be best positioned to capture the value.

Implications for Global Portfolios

• The earnings-fundamentals gap is a tactical opportunity. Stocks trading at 40% discounts to recent highs while delivering quadruple-digit earnings growth are rare. The current dislocation is driven by sentiment and positioning, not by demand destruction.
• Focus on the top of the pyramid. Companies with the scale, customer relationships, and technology roadmaps to sustain outperformance are the names to own. Smaller players face margin pressure and customer-concentration risk.
• Order visibility is the key metric. Companies with order backlogs extending into 2027 and deep relationships with both global hyperscalers and domestic AI clusters offer the most defensible demand profiles.

Risk Factors

• U.S. earnings guidance: If Nvidia or the hyperscalers guide down capex growth in their July-August earnings calls, the semiconductor-demand narrative could face a reset.
• Trade policy: U.S. restrictions on advanced semiconductor manufacturing equipment could constrain the ability of Chinese firms to produce the most advanced nodes.
• Capacity overbuild: While current demand is robust, aggressive expansion plans across the supply chain could create oversupply by late 2027 if AI capex growth decelerates faster than expected.

Bottom Line

The semiconductor sector is experiencing a classic market overreaction: stocks are being priced as if the AI demand cycle is ending, while the companies themselves are reporting that it is accelerating. For global investors with a 12- to 18-month horizon, the current correction offers a lower-risk entry point into what remains one of the highest-conviction growth stories in the China technology complex.

The disconnect between price and fundamentals is the trade.