As the US-Iran peace talks and Trump's visit to China draw closer, supply and demand analysis becomes increasingly inadequate, and macroeconomic assumptions may be temporarily absent. Macroeconomic speculative funds will retreat in stages, potential entrants will adopt a wait-and-see approach, and allocation-oriented funds will reduce their positions on rallies; existing funds will exhibit a clear pattern of selling at highs and buying at lows, with rallies at low levels being a behavioral outcome rather than a signal.
Both the US-Iran peace talks and Trump's visit to China are events where market sentiment is nearing its peak. Given the strengthening of existing structural market trends, proactively reducing volatility is a better option.
Despite the complex short-term pace, we believe that in the medium term, AI and energy will remain the main sources of the supply-demand gap; this year's AI + energy is equivalent to the AI + dividend in 2023-2024, and AI + resources in 2025.
As the US-Iran peace talks and Trump's visit to China draw closer, supply and demand analysis becomes increasingly inadequate, and macroeconomic assumptions may be temporarily absent.
1) The timing of the US-Iran peace talks and Trump's visit to China will constrain the investment decisions of bullish investors in commodity and equity markets.
For commodity participants, during the window of time when the outcome of the negotiations is about to be revealed, every release of information on the progress of the negotiations means that existing positions face a sharp directional repricing risk. Even if there is a bullish intention, there is no need to bear the risk of position retracement. For example, recently, the net long positions of managed funds in crude oil futures under the CFTC have decreased by 19.6% compared with April 21. This does not reflect a bearish consensus, but rather a structural disconnect of "having a directional judgment + no additional positions".
Bullish and bearish opinions are not necessarily reflected in the pricing of financial assets. Rashly attempting to deduce the strength of supply and demand fundamentals from price signals at this stage can easily lead to directional misjudgments in the future. A similar logic applies to actual inventory replenishment in the real economy.
Since the outbreak of the current conflict between the US, Israel, and Iran, domestic social inventories of major chemical products have undergone a passive destocking cycle of approximately two months. According to data from JLC Network Technology Co., Ltd. on representative liquid bulk chemical inventories, as of the week ending April 28, social inventories of liquid bulk chemicals, represented by methanol, ethylene glycol, styrene, benzene, and toluene, have decreased by an average of about 26% compared to before the conflict, and the turning point has yet to appear.
This unusual inventory behavior is essentially consistent with the logic of the financial market: having already passively reduced inventory for two months, and with the outcome of the negotiations approaching, there is no need to proactively replenish inventory and risk drastic price fluctuations. Consequently, genuine end-user demand in the economy is also partially absent.
2) The lack of multiple financial institutions coupled with the lack of inventory replenishment in the real economy directly led to a "lame" macroeconomic analysis.
At this stage, we can only analyze the supply logic, but cannot effectively assess the true situation on the demand side—are there buyers willing to take delivery in the spot market? How large is the actual supply-demand gap? When will the restocking cycle begin? Is end-consumer demand being systematically suppressed?
These questions cannot be reliably answered until the critical events take place.
If the signals from the demand side are distorted, then macroeconomic analysis based on demand projections, judgments of the monetary environment, and even corporate profit forecasts will also be in a state of "absence".
Simply relying on supply for supply and demand analysis, whether for commodities or stocks, limits the slope of price movements and the height of market trends.
Macro-speculative funds will retreat in stages, potential entrants will adopt a wait-and-see approach, and allocation-oriented funds will reduce their positions on rallies.
Based on the information disclosed so far regarding the US-Iran peace talks, the window for the talks to be finalized may be in mid-to-late May, and Trump's potential visit to China could become another focal point.
Following typical game theory logic, before Trump's visit to China, the two sides were likely to maintain a controllable and moderate stance in their game; after the visit, the issues became more complex, and the probability of escalating friction increased marginally. If this is compounded by the positive factor of a US-Iran peace agreement, late May could be the peak of market sentiment and risk appetite for the year.
Since April 9 this year, domestic broad-based ETFs have continued to experience net redemptions. As of May 7, there have been net redemptions for 18 consecutive trading days, with a cumulative net redemption scale of approximately 262.5 billion yuan. In particular, since the Shanghai Composite Index stabilized above 4,100 points, the redemption scale of broad-based ETFs has continued to expand.
Considering the current macroeconomic environment approaching a major event window and the profit-making effect that the Shanghai Composite Index has been accumulating since late March, we believe that the recent redemption trend in the ETF market is likely due to a large amount of allocation funds reducing their positions.
We can choose to ignore such short-term game-level disturbances, but we must be fully prepared for the amplification of volatility before and after the event window. After all, there are still a large number of participants in the market using quantitative and trend-following strategies. Their mechanical reactions to volatility and momentum signals may significantly amplify the two-way price fluctuations before and after the event.
The pattern of existing funds shifting from high to low levels will be quite pronounced; price increases at low levels are a result of behavior rather than a signal.
When the inflow of allocation-oriented funds is limited while market trading remains active, frequent high-to-low rotations are very likely to occur.
Market activity remains high recently, with turnover exceeding 3 trillion yuan in the first three trading days of May.
The recent investment strategy of active funds is to continuously look for low-priced themes or new catch-up themes and switch between high and low. Commercial aerospace and robotics are low-priced stocks, while themes such as token factories and token agents are new themes.
As for more conservative funds, their investment strategy is to look for sectors with low valuations and potential for growth or marginal changes, such as gaming, Hang Seng Technology (cloud), and even real estate (recently, real estate ETFs have shown very stable performance, and the share prices of leading Hong Kong-listed companies have broken through previous highs, mainly reflecting the partial stabilization of the secondary housing market in some first-tier cities and the easing of real estate policies in more cities). This high-to-low characteristic is very evident in the ETF market.
Given the strengthening of the existing structural market trend, proactive volatility reduction is a better option.
Recently, the momentum of technology stocks in China, the US, Japan, and South Korea has been continuously strengthening.
However, even the strongest industry trends can be followed by sharp fluctuations if the slope is too steep in the short term. We have constructed the Momentum Strength Index (MSI) to measure the momentum strength of recent performance in major global technology indices.
Currently, the momentum strength of the South Korean KOSPI has entered the overheated peak zone, while the Nasdaq 100 and Nikkei 225 are in the strong acceleration zone, and the ChiNext index is very close to the threshold of the strong acceleration zone. Specifically, the latest MSI reading of the South Korean KOSPI index is 92.4, indicating that market momentum is nearing its limit, and there is even a possibility of a costly "short squeeze rally".
Historically, excessively high MSI readings have often corresponded to a short-term peak, with a high probability of subsequent sharp fluctuations. The latest MSI readings for the Nasdaq 100 and Nikkei 225 are 86.3 and 80.0 respectively. While this doesn't directly mean momentum will quickly run out, the momentum effect clearly exceeds the historical normal range, indicating limited room for further significant gains and potentially making them highly sensitive to negative news.
As for the ChiNext Index, the latest MSI reading is 64.4, which is still within a reasonable historical range. It is worth mentioning that before the pullback on May 8, the reading on May 7 was 71.9, which was very close to the critical value of 75 for the strong acceleration zone.
We believe that even though AI-related industry trends are strong, the unusually strong momentum and steep upward slope in the short term likely indicate increased volatility in the future. In this context, proactively reducing portfolio volatility may be a better option for existing funds.
The short-term situation is becoming more complex, while in the medium term, AI and energy will remain the main sources of the supply-demand gap.
1) The rapid iteration of AI has created new supply and demand gaps in different fields for four consecutive years.
The supply-demand gap for computing cards in the first phase (2023-2024) was mainly driven by the scaling law at the training end.
The second phase (2025) of the Scaling Law for inference will drive new demand for computing power, storage, and power facilities.
The third phase (2026) begins with the explosion of agents such as Claude Code and Codex, where AI upgrades from "usable" to "easy to use." The demand in real-world work scenarios drives a significant increase in token usage. The weekly token consumption on the OpenRouter platform surged from 8.25 trillion at the end of January to over 20 trillion. Anthropic's annualized revenue and usage in the first quarter of 2026 increased 80 times.
Strong demand continues to reinforce the necessity of computing infrastructure, allaying market concerns that AI investments may be difficult to commercialize.
More importantly, the Agent has achieved significant differentiation and established a certain level of user stickiness. This is the foundation for monopoly and sustained excess profits, unlike the fierce competition of homogeneous models.
During this phase, the storage gap gradually expanded from HBM to traditional DRAM and even flash memory, and supply and demand gaps also emerged for other AI server components such as CPUs. With the increasing demand for cost-effective models and the maturation of domestically produced models, domestic cloud services also experienced supply shortages and price increases.
Looking ahead, the AI industry trend is relatively certain, and the continuous surge in demand is exposing new supply and demand gaps. The main contradictions in the market lie in the slope, pace, and valuation.
2) Supply constraints or contraction may lead to a sustainable energy supply-demand gap.
In contrast to AI, the energy and chemical sector, another major supply-demand gap this year, has been primarily driven by supply-side contraction rather than demand-side expansion. Looking at the petrochemical chain as a whole, the supply elasticity of crude oil will be significantly higher than that of processed products in the future, forming the basis for sustainable price spreads and excess profits.
The regions suitable for expanding petrochemical energy processing capacity in the future are quite limited.Europe faces high refining costs and a real need for energy diversification and green transformation. Driven by energy security and dual-carbon goals, China has implemented a strategy to proactively reduce petrochemical energy consumption. The Middle East was originally the most suitable region for downstream petrochemical and chemical industries, but geopolitical conflicts have cast a shadow over future private sector investment, and investment in the Middle East has consistently lagged behind Asia in terms of both pace and efficiency. Japan, South Korea, and other emerging Asian market countries have exposed serious supply chain instability issues in the recent US-Iran conflict, which may drive subsequent energy transformation and reduce reliance on a single energy source.
Long-term supply constraints may create a sustainable supply-demand gap, but due to the overall stable demand, the energy and chemical chain is more inclined towards dividends or value attributes, and its explosive growth potential is relatively low.
We believe that this year's AI+energy and chemical industry is equivalent to the AI+ dividend of 2023-2024, and the AI+resource industry of 2025.
This barbell-like structure indicates that the K-shaped differentiation across various sectors globally is intensifying, with very few sectors truly possessing sustainable supply-demand gaps and unexpected profits.
AI is a demand-driven offensive option, while energy and chemicals (new energy, traditional energy, refining, and basic chemicals) are more suitable to replace the dividends of 2023-2024 and resources in 2025 as a relatively stable return option. If you observe the trend of oil and petrochemical ETFs, the trend before and after the peak of the energy and chemical market in March 2026 is highly consistent with the trend of the four major banks around May 2023.
In terms of configuration, the underlying logic remains the reassessment of pricing power in China's advantageous manufacturing industries, most notably new energy, chemicals, non-ferrous metals, and power equipment. We continue to closely monitor the progress of domestic AI. The explosive growth in hardware volume remains the area with the largest expected gap in the AI chain, while advancements in domestic models are expected to drive both volume and price increases in cloud services. We are optimistic about domestic computing power and cloud platforms.
In addition, it is recommended to continue to increase allocations to some undervalued stocks, with a focus on securities firms and insurance companies.For cyclical products experiencing price increases, the prosperity of cyclical growth products such as AIDC chain and lithium battery chain continues, but the expectation gap is relatively smaller, and the trend of shrinking market is faster.
It is recommended to focus on the most tightly supplied segments, which is reflected in the recent frequency of price increases. These segments mainly include copper clad laminates, fiberglass, high-speed silicon, electronic specialty gases, optical fibers, MLCCs, chromium, lithium carbonate, rare earth elements, and carbon fiber.
For traditional cyclical commodities, it is recommended to focus on commodities where actual capacity clearing or supply is absolutely constrained, such as phosphate chemicals, MDI, spandex, dyes, glyphosate, urea, rubber, and refrigerants.
This article is sourced from:
