AI Investment "Matching" Public Funds Focus on Computing Power Hardware Performance Realization

Author: Hao Jian

In recent times, Wall Street “Big Short” Michael Burry bought put options on Nvidia while adding to his positions in Microsoft and Adobe. In China’s A-share market, public fund managers have continued to increase their holdings of individual stocks such as Zhongji Xuchuang, Xinyisheng, and Hengtong Optoelectronics in the first quarter. The two lists of holdings point in markedly different investment directions. With AI investment “on target,” the disagreement behind it reflects a set of pragmatic logic in domestic technology investing.

**　　Divergent Investment Directions**

Burry’s latest changes in his holdings show a complete strategy toward AI: buying put options on Nvidia, Palantir, the SOXX ETF, and the Nasdaq 100, establishing long positions in Microsoft, increasing stakes in software and payment-sector stocks such as PayPal and Adobe, and building a portfolio of “long software, short hardware.” Burry said that the market’s current enthusiasm for AI is very similar to the final stage before the internet bubble burst in 2000.

But the first-quarter reports of domestic public funds show a different trend.

According to data from Tianxiang Investment Advisory, at the end of Q1 2026, Zhongji Xuchuang continued to rank as the largest overweight holding among public funds, with a market value of 73.896 billion yuan. Xinyisheng ranked as the third-largest overweight holding. AI computing power hardware became the main direction for public funds to increase holdings in the first quarter. Specifically, Hengtong Optoelectronics was the stock with the largest increase in market value among public funds in Q1; Changfei Fiber, Yuanjie Technology, Zhongtian Technology, and others also saw increases of more than 7 billion yuan each.

The holdings lists of high-performing public funds overlap to a high degree. As of May 13, Wind data shows that GF Vision Intelligent Selection’s return since the beginning of this year exceeded 110%. At the end of Q1, the fund’s top holdings included Changfei Fiber, Zhongtian Technology, and Hengtong Optoelectronics, among others. In its Q1 report, fund manager Tang Xiaobin said: “Even if the valuations of global technology stocks face pressure, the certainty of domestic computing power performance remains clearly evident.” Guoshou Anbao Digital Economy A posted a return of over 96% in the same period. In Q1, fund manager Yan Kun made a significant increase to holdings in Zhongji Xuchuang and Xinyisheng, with a high concentration in the top ten holdings. Pudong Ansheng Digital Economy A’s return since the beginning of this year exceeded 87%. Fund managers Zheng Minhong and Li Fan said: “We remain firmly optimistic about core AI computing power assets represented by optical communications.”

On one side are Wall Street bigwigs taking bearish positions on hardware; on the other are domestic public fund managers heavily overweighing hardware. Between the two holdings lists, it seems there is no compromise.

**　　Focusing on Performance Realization**

Domestic fund managers’ concentrated positioning in the hardware direction is supported by performance data.

Benefiting from strong demand for AI computing power infrastructure, the optical module industry continued to show robust growth in Q1 2026. Based on disclosed results, in Q1 leading companies such as Zhongji Xuchuang saw substantial growth in operating revenue and net profit, and core suppliers such as Tanfeng Communication and Huagong Technology also maintained steady growth rates. In addition, companies such as Guangju Technology and Yuanjie Technology reported year-over-year net profit growth rates that were all above market expectations.

ICBC Credit Suisse Fund said that AI data center demand is continuously intensifying, driving a global storage supply-demand imbalance. The rise in the storage sector is “not a traditional cyclical rebound, but a shift in industrial logic driven by AI computing power demand,” and the duration for which storage prices remain elevated at high levels may exceed what is typical for traditional cycles. Tang Xiaobin said that the allocation logic should further focus on sub-sectors where supply is constrained and demand grows in an “inflation-like” manner, including storage, CPO, optical modules, liquid cooling, AI power supplies, and fiber optic cables, among others. His view is: “There is no need to count on the dividend from long-term concept themes; the current fundamental outlook already forms a more favorable investment direction.”

The scope of hardware focus is also extending outward. In Q1, Ping An Fund’s fund manager Lin Qingyuan focused on core components of gas turbine equipment and targets related to power grid equipment. In his Q1 report, he said that power is the most critical “hard constraint” in the current expansion of AI. Orders from major global gas turbine companies have already been booked through 2030, making this the part of the entire AI industry chain with the largest supply-demand gap.

Software and application directions face a different situation. A team from CICC Research believes that: first, since 2023, U.S. software companies have basically not delivered scalable AI revenue of a size that satisfies the market; second, the commercialization of Agent products launched by industry giants such as Salesforce remains in an early stage and has not contributed actual revenue; third, technological progress has instead triggered market concerns about the moat of enterprise SaaS. The team concluded: “High-valuation sectors are naturally fragile. Once long-term barriers are questioned, the discounting phenomenon appears quickly and in a concentrated way.” McKinsey’s latest report also said that although AI has huge commercial potential, it faces challenges in realizing monetization.

**　　Trading Crowding Cannot Be Ignored**

However, as consensus moves toward extremes, risks are accumulating in parallel. Although fund managers firmly believe that AI computing power is the core line of the mid-term, at the market level the trading congestion caused by extreme crowding cannot be ignored. Some fund research teams warned that compared with multiple past instances of crowded positioning in the A-share market, this cycle’s style appears even more extreme; potential risks such as trading congestion, elevated valuations, and technological iteration are building up, so it is necessary to prevent in advance the liquidity shocks that could be triggered by pullbacks under extreme consensus. A research team at Haitong Securities said that AI hardware has broadly entered an overbought range, and the probability of a short-term adjustment is rising.

A research team at CICC Research believes that the AI rally has not yet reached the typical “bubble” stage. However, “front-running” in investment exists objectively, and this is also the main reason why the AI market has advanced through twists and turns over the past few years. Legendary U.S. investor Paul Tudor Jones said that the current AI-driven rise shares similarities with the period before the internet bubble burst, but he also believes the market may still have further upside potential.

Some fund managers have also noticed changes that are taking place. In their Q1 reports, Zheng Minhong and Li Fan said that for North American large-model companies represented by Anthropic, ARR performance is showing “exponential growth.” Meanwhile, domestic companies such as Zhipu and Kimi are showing improving financial data. “Industry growth is gradually moving away from an arms race and toward a new business model driven by diverse applications.” Yan Kun also noted that once the source of AI capital expenditure shifts from debt leverage to a revenue closed loop, the market’s concerns about AI investment returns will be corrected.

Fuguo Fund’s view is that AI computing power remains the core theme in the medium term, but the market may experience short-term volatility. The market’s structure is “a differentiation between business momentum and performance validation.”

Burry’s short positions reflect caution about AI hardware valuations, while domestic public fund managers’ heavy holdings reflect a judgment about the certainty of computing power performance. The core divergence between the two may not lie in the direction, but in the timing.

(Editor: Xu Nannan)

Keywords: