Zhao Feng's "Cold Arrow": AI Capital Expenditures Are Swimming Naked
The overwhelming momentum of AI has not only left value-oriented fund managers “on pins and needles,” but also prompted typically balanced fund managers to “voice their opinions.”
The latest second-quarter fund reports show that some balanced fund managers have begun to question and discuss the current AI market. They are expressing views and even doubts about the sustainability of industry investment and returns, the market’s investment logic and valuations, and the commercial fundamentals of the industry.
Zhao Feng, manager of the Ruoyuan Balanced Value Three-Year Holding Hybrid, highlighted the logical discrepancy of some AI industry chain companies in the Q2 2026 report.
Similarly, Dong Chunfeng, Qin Wei, and Wu Fei—co-managers of the Ruoyuan Research Selected Balanced Three-Year Holding Hybrid Initiated Fund—devoted substantial discussion to the current value distribution within the AI industry chain; Zhang Jialu, manager of Ruoyuan HK Stock Access Core Value Fund, stated operational conditions more directly.
The second-quarter reports of these three Ruoyuan funds are not focused on “whether AI is important,” but rather on more specific investment issues:
Who ultimately receives the value? When does investment convert into returns? By what valuation standard should investors trade?
AI value is almost entirely concentrated at the hardware investment end
The Ruoyuan Research Selected Balanced Three-Year Holding Hybrid Initiated Fund is the most direct in questioning the current AI value allocation system.
In the second-quarter report managed by Dong Chunfeng, Qin Wei, and Wu Fei, it states that as the AI investment scale expands rapidly, it has become the largest variable influencing the macro economy and market, yet their portfolio maintains a cautious stance in selecting relevant targets—mainly due to doubts about the current value allocation in the AI chain.
“Currently, value is almost entirely concentrated at the hardware investment end. This state is unlikely to last, and among hardware companies, those truly meeting our requirements for quality, business model, and competitive barriers are rare. Some suboptimal candidates have significant valuation discrepancies compared to our standards.”
Dong Chunfeng and others further cite overseas storage companies as examples, analyzing the potential risks of the logic of “moving up the AI industry chain, seeking tight links, and earning profits from price elasticity.”
They explain that chasing price elasticity may underestimate the stimulating effect of price increases on supply. In fact, price increases and excessive profits have already attracted new entrants in some supposedly “moated” industries.
Secondly, due to inventory behaviors, upstream shipment fluctuations are greater than downstream. Even a minor revision in capital expenditure expectations can cause inventory changes and sharp price declines.
Therefore, on valuations, their conclusion is:
“Since price rises lead to negative feedback from supply and demand, profits from price increases should be discounted much more than from quantity growth.”
Rather than chasing short-term price elasticity of upstream products, Dong Chunfeng and team prefer “companies that leverage their competitiveness to enter the AI core supply chain and enjoy industry’s long-term incremental gains.” In Q2, their portfolio increased allocation to “consumer electronics leaders with stable main business, solid barriers, and clear future AI-related business growth.”
Their attitude towards AI is not closed. The report states the portfolio “always maintains an open attitude towards AI,” and will actively seek “companies with a long-term advantageous position in future AI value distribution.”
There is a “possibility” of pressure on long-term demand and pricing
Zhao Feng first affirms AI’s long-term industrial status.
In the Q2 report for Ruoyuan Balanced Value Three-Year Holding Hybrid Fund, he states, “We believe AI is the next generation of infrastructure, the core variable in the next productivity revolution. Its impact on productivity and social organization is still early-stage, and future potential remains vast.”
But he immediately adds a constraint:
“If the current speed of AI value realization can't catch up with the rate of capital expenditure and hardware inflation, it will negatively impact the industry’s demand growth and pricing power.”
He also analyzes this issue in the context of historical technology revolutions. From core breakthrough to productivity fulfillment, it usually takes 50-60 years; each revolution sees “returns first to capital, then to labor,” with a lag, and the end of the installation phase nearly always causes financial bubbles and crashes. Bubble bursting doesn’t end the technology, but triggers a “deployment phase” in which tech dividends spill across society through asset repricing.
This irrationality may be corrected to some degree in the future, though the reasons are unpredictable, but the high returns implied by low valuations are definite—shown either through rising stock prices or shareholder returns. Of course, research tracking on AI will be intensified, as investment opportunities from future industry developments remain abundant, whether from tech changes or application layers.
Back to the current market, Zhao Feng observes that Q2 saw extreme market divergence. The tech stock enthusiasm brought by the AI tide has continually siphoned funds from other sectors, widening the valuation gap.
On one hand, by consensus, many tech companies’ 2027 valuation is already tens or even hundreds of times PE; on the other hand, many traditional industry companies and “AI victim companies” have EV/FCFE ratios down to single digits, meaning investors can recoup principal in a few years with stable free cash flow.
He concludes: “Aside from industries benefiting from AI, other industries still face fundamental pressure in Q2, with unclear prospects. Given the gap between decades-long and years-long payback periods, and the difference in risk between long-term and near-term profit forecasts, we believe current market pricing is irrational.”
This valuation doubt is not isolated.
He mentions that while he’s followed AI tech since ChatGPT's debut, unlike industry insiders' belief in Scaling Law, there is a certain lag in tracking its industrial development. When relevant stocks’ appreciation exceeds research understanding, and ever-higher profit forecasts are needed to support prices, investing becomes difficult.
He adds that many AI beneficiaries in this rally have low competitive barriers. Without the short-term supply-demand imbalances, these companies wouldn’t have such high profitability; once demand-supply dynamics shift, their bargaining power will inevitably decline.
He believes relying on demand forecasts for future profit estimation isn’t a strong method—first, fast-growing demand predictions are uncertain; second, when company competitive barriers are weak and industry structure is poor, converting revenue into profit is highly uncertain, thus these waves can’t be fully covered by such investment methods.
Yet Zhao Feng does not deny AI’s ongoing investment potential. He says research tracking will intensify, with many opportunities still ahead, both in tech shifts and application growth.
“Reasonable valuation range” comes only after “periodic adjustment”
Ruoyuan HK Stock Access Core Value Hybrid, managed by Zhang Jialu, is also following AI.
The report says AI was “the standout core theme” of Q2, whose strong money-attracting effect “almost drained liquidity from all traditional industries.” As Hong Kong stocks offer few local AI investment targets, they haven’t benefited from this wave, suffering obvious “bloodletting” pressure. Funds have continued to flow to Korean and Japanese semiconductor hardware leaders and related A-share stocks, further diverting Hong Kong liquidity.
Operationally, the product continues to invest in value stocks with stable cash flows and attractiveness after long-term correction; it also filters “hardware firms benefiting from the global AI computing power chain and still in reasonable valuation range” for allocation.
Thus, operations adhere to value assets at the core, using stable cash flow and low volatility to provide solid portfolio returns. At the same time, it will closely track periodic adjustments in AI hardware, deploying only in reasonable valuation ranges, seeking both offense and defense in a balanced strategy to deliver excess returns amid complex, volatile markets.
For the second half, Zhang Jialu judges, “the AI industry trend is far from over, but internal differentiation will intensify, shifting from pure concept-driven to rigorous performance delivery scrutiny.”
Her operating principle: while continuing value asset core allocation, “closely track periodic adjustment in AI computing hardware, and allocate in reasonable valuation ranges.”
But “periodic adjustment” and “reasonable valuation range” together set clear allocation criteria—she’s not chasing AI hardware at any price, but waiting for price and valuation to enter her acceptable zone.
It’s worth noting that as of quarter's end, New Yisen and Changfei Optical Fiber Cable were new entries among top holdings. This shows the fund isn’t just waiting—it has already allocated some hardware firms; but the quarterly report continues to stress “periodic adjustment” and reasonable valuation for future additions.
Bringing investment questions back to cash flow, competitive moat, and buy price
From these three Q2 reports, Ruoyuan Fund’s long-term outlook on AI is not pessimistic.
But doubts from these three products fall on three areas:
First, value created by AI is almost entirely concentrated at the hardware end—can this allocation persist? Second, can AI value realization keep pace with expanding capital expenditure and hardware inflation, and is there irrational pricing between assets with decades-long and years-long payback periods? Third, even with a long-term industry trend, AI hardware needs periodic adjustment and should be allocated only once in reasonable valuation range.
Ruoyuan Fund managers are “picking holes” in AI, not denying the tech revolution, but bringing investment questions back to cash flows, competitive moats, and purchase price: Who can secure AI-created value for the long term, how long before this value is realized, and how many years of growth expectations are baked into today’s prices.
Risk warning and disclaimerThe market has risks; investment needs caution. This article does not constitute personal investment advice, nor does it consider individual users’ special investment goals, financial situation or needs. Users should consider whether any opinions, views, or conclusions contained in this article suit their particular circumstances. Invest accordingly, at your own risk.