This year's AI-driven market rally is a "bottleneck trade" fueled by massive infrastructure investment, more akin to the 2006-07 bull market driven by investment and capital-intensive companies rather than the dot-com bubble. Rate hikes are unlikely to affect the valuations of "AI-cyclical" stocks unless they genuinely impact end-demand for AI, commercialization assumptions, and the pace of capital expenditure growth. Globally, the rate-hiking cycle is first affecting sectors with relatively weaker demand growth, with the K-shaped divergence between AI and non-AI sectors holding true worldwide. However, with the simultaneous return of a strong US dollar narrative and a market structure dominated by existing capital reallocation, non-AI cyclical sectors in the A-share market are significantly weaker compared to their overseas counterparts. While a K-shaped divergence exists, the breadth of the A-share rally is relatively narrower than overseas. To change the weakness in non-AI sectors, they need their own narratives to show positive developments or a shift in capital flows, rather than waiting for an AI sector correction.

Comparing This AI Rally to the 2006-07 Bull Market

If we analogize this AI-driven rally to China's 2006-07 bull market, then "rate hikes" will not be the turning point that ends the rally. Firstly, this rally is a "bottleneck trade" driven by massive investment, more similar to the 2006-07 bull market than the dot-com bubble. The global AI theme this year has very clearly played out as a "bottleneck trade." In the US market, hardware segments like memory, optical communication, and semiconductor equipment leaders have performed excellently, while traditional platform companies and AI application pioneers have generally underperformed. Compared to the 2000 peak, the Nasdaq 100 forward P/E reached as high as 60x, whereas it is currently only 25x; the average trailing P/E of the "Four Horsemen" then soared to 82.7x, while the current "Magnificent Seven" stands at 38x. Asia-Pacific regional markets also primarily reflect "supply-demand drivers" rather than just valuation drivers. Year-to-date, core Asia-Pacific indices have significantly outperformed US stocks, especially the KOSPI 50 (+160%) and Nikkei 225 (+42%), which are deeply embedded in the AI supply chain, with corresponding P/E (TTM) ratios of only 10x and 22x respectively. Similar characteristics are evident domestically. On a macro level, the current spillover from AI infrastructure investment and China's technology industry upgrade is strikingly similar to the 2006-07 shift of global industrial demand eastward and the explosion of China's foreign trade exports. At the market level, cyclical-leaning sectors that have performed well this year, such as capacitors, fiberglass, copper foil, IGBT, silicon carbide, and silicon wafers, were still viewed as typical cyclical stocks under traditional supply-demand frameworks two years ago, with industry leaders' P/B ratios generally between 1-3x. However, driven by the红利 of North American AI Capex and the narrative of domestic substitution, these sectors have undergone a deep re-rating of their valuation centers, shifting to PEG or even PS valuations.

Secondly, rate hikes are unlikely to affect the valuations of "AI-cyclical" stocks unless they genuinely impact AI's end-demand, commercialization assumptions, and capital expenditure growth. Federal Reserve Chair Waller's debut this week broke the forward guidance mechanism introduced by Bernanke in 2012, as he declined to submit personal rate forecasts, emphasizing a focus on inflation and employment to restore Fed credibility. The market currently expects and is pricing in no rate cuts this year and a 25 basis point hike in October. However, the impact path of rate hikes differs significantly between growth-stock rallies driven by valuation and cyclical-stock rallies driven by supply-demand gaps. The Fed began raising rates in June 1999, and the Nasdaq index peaked and turned bearish 7 months later. For the 2004-2007 bull market, the Fed's first rate hike was in June 2004, and the bull market did not end until 38 months later (October 2007). For this "bottleneck trade" driven by massive investment, unless rate hikes truly affect end-demand and capital expenditure growth, or unless the competitive landscape and AI's monetization narrative face significant challenges, rate hikes (especially in the early stages) are unlikely to directly impact the valuations of "AI-cyclical" stocks.

Global Rate Hikes and the AI vs. Non-AI Divide

The easing expectations from 2024 have shifted in recent months, with the US, Europe, and the UK moving from easing cycles back towards neutrality. The rate-hiking process is first affecting sectors with relatively weaker demand growth, and the K-shaped divergence between AI and non-AI sectors holds true globally. AI has been the nearly sole market theme across countries this year, with AI universally outperforming non-AI globally, but the "nature" of this divergence varies starkly by country. In terms of performance, the cumulative excess return of AI over non-AI year-to-date is: South Korea +157%, China +143%, Japan +134%, and the US +96%, with divergence generally greater in Asia than in the US. However, the AI/Non-AI P/E ratio calculated using the aggregate method (Σ market cap ÷ Σ earnings) is only 1.48x for the US, 1.45x for Japan, 1.27x for Korea, while for A-shares it is as high as 9.8x (AI basket ~92x P/E, non-AI only ~9x), a full order of magnitude higher than developed markets. In other words, South Korea and Japan are experiencing "explosive profit growth with still-cheap valuations" (AI P/E was even lower than non-AI at the start of the year), driven by profits and relatively healthy. The US shows moderate profit and valuation premiums, while A-shares' rally has the highest proportion of "valuation expansion."

Strong Dollar Narrative Returns, Pressuring Non-AI Sectors in A-Shares

Judging from Federal Reserve Chair Waller's statements at the June FOMC meeting, the Fed emphasizes restoring credibility and focusing on its core mandates of inflation and employment. On the economic and industrial front, according to the latest corporate earnings guidance, some large tech companies' capital expenditure scale in the AI field for 2026 approaches $700 billion, mostly initiated by the North American private sector and implemented in North America, establishing a monopolistic competitive advantage relative to the globe. Under this combination, the US dollar is reasserting its strength, with the "Great Moderation" narrative of the 1990s returning. This is evident from the recent divergence between the US Dollar Index and long-term US Treasury yields. The 10-year Treasury yield has fallen from 4.6% a month ago (May 18) to around 4.45% on June 18, while the US Dollar Index rose from 99.0 to 100.8 over the same period. The recent dollar strength can no longer be explained solely by initial dollar liquidity shortages during the Middle East conflict. This may partly explain why A-share non-AI cyclical sectors have been weaker relative to overseas markets since May. In fact, for many globally priced industrial goods, global AI-driven infrastructure demand is still growing, China's "Six Networks" are steadily advancing, global defense spending has entered a new expansion cycle, and traditional resource-rich countries in the Middle East, Latin America, and Southeast Asia are increasing their industrialization investments. Demand-wise, a recession narrative should not be applied. Earnings expectations for many A-share industrial manufacturing sectors have generally been revised upwards this year, completely different from the ongoing downward revisions in the consumer sector, yet their stock price performance has converged and is明显 weaker than overseas. A similar situation occurs in the brokerage sector. This may involve factors like提前 pricing negative narratives and some liquidity issues. Continued net redemptions of broad-based ETFs are suppressing non-AI sectors lacking clear narrative catalysts. During the previous index adjustment phase (May 28 – June 12), net redemptions of broad-based ETFs明显 slowed, and related sectors experienced a brief recovery. However, with the index反弹, this week (June 15 – June 18) the four CSI 300 ETFs held by Central Huijin, as shown in the 2025 annual report, continued to show large net outflows, with cumulative net redemptions of 42.2 billion yuan. The strong US dollar and ETF net redemptions are currently key factors suppressing non-AI cyclical sectors.

Catalysts Needed for Non-AI Sectors, Not Just Waiting for AI to Correct

To change the weakness in non-AI sectors, they need their own narratives to show some positive changes in the future, rather than waiting for an AI adjustment. This change could come from an unexpected drop in oil prices after海峡通航, lowering inflation expectations, or from a synchronized recovery in global non-AI industrial production and social activity (e.g., supply recovery and increased operational rates in Asian countries as oil and gas prices fall). After all, the global environment we are in is not a recession; industrial production is expected not to weaken amid AI, defense, and accelerating industrialization in developing countries. It just requires more evidence for the market to see. In fact, the narrative discussion about the silicon-based vs. carbon-based divergence is relatively common only within China, possibly stemming from确实 weaker domestic consumption growth data globally. China's fixed asset investment is weak but structurally倾斜 towards advanced tech manufacturing, which is positive in the long run. Slower capital accumulation in traditional industries reduces internal卷, while emerging manufacturing industries continue to increase investment to enhance competitiveness. This corresponds to stronger profit-creating potential for A-share technology and manufacturing sectors. In terms of allocation, we still recommend adhering to an AI + Energy/Chemicals structure. On the AI side, we remain看好 on some品种 with relatively lower筹码拥挤度, such as memory, gas turbines, diesel generator sets, and some semiconductor equipment and materials. On the Energy/Chemicals side, for electric power and new energy, we看好业绩兑现 for varieties like electrolytes and additives, separators, etc. For chemicals, the decline in the center and volatility of oil prices brings restocking and operational demand, and peaking macro liquidity expectations could be potential节奏 points later. Currently, we are more看好 on varieties with significant cost reduction potential, relatively rigid demand, and low valuations, such as refrigerants, phosphorus chemicals, spandex, dyes, and large-scale refining. For non-ferrous metals, we recommend computing power metals that have some AI exposure but whose valuations are temporarily suppressed by the加息 narrative on a macro level, such as tin, copper, and some AI minor metals (tungsten). Additionally, we continue to recommend increasing allocation to undervalued brokerages. Current瑕疵 like liquidity suppression may gradually消退 starting in the second half of the year, and mid-year earnings previews could also serve as a catalyst.

Risk Factors

Intensified friction between China and the US in technology, trade, and finance; domestic policy intensity, implementation effectiveness, or economic recovery falling short of expectations;超预期 tightening of macro liquidity domestically and overseas; further escalation of regional conflicts such as Russia-Ukraine and the Middle East; slower-than-expected digestion of real estate inventory in China.