Strategy Focus | Volatility Reduction and Awaiting Negotiations and Visits
As the U.S.-Iran negotiations and the potential Trump visit to China approach, supply-demand analysis becomes increasingly lopsided, with macro assumptions temporarily absent. Macro-driven trading funds are expected to retreat in phases, potential entrants are likely to adopt a wait-and-see approach, and allocators may reduce positions at higher price levels. The rotation from high to low positions among existing capital will be quite pronounced, with low-position rallies being a behavioral outcome rather than a signal. Both the U.S.-Iran talks and a potential Trump visit are events that could push risk sentiment toward a peak. Against the backdrop of strengthening structural market trends, actively reducing portfolio volatility is the preferable choice. Although short-term dynamics are becoming more complex, we believe that, from a medium-term perspective, AI and energy/chemicals remain the primary sources of supply-demand gaps. This year's combination of AI and energy/chemicals is analogous to the AI-dividend play of 2023-2024 and the anticipated AI-resource theme of 2025.
As the U.S.-Iran negotiations and potential Trump visit to China draw nearer, supply-demand analysis becomes increasingly lopsided, and macro assumptions are temporarily absent.
1) These two upcoming events constrain investment decisions for long positions in commodities and equity markets. For commodity participants, during the window leading up to the negotiation results, every piece of progress information implies a risk of severe directional repricing for existing positions. Even with a bullish view, there is little incentive to bear the risk of position drawdowns. For instance, the net long positions in crude oil futures held by managed money (CFTC data) have fallen by 19.6% since April 21. This reflects not a bearish consensus but a structural disconnect of "having a directional view but not adding positions." Bullish or bearish views may not be accurately reflected in financial asset pricing during this phase. Attempting to deduce supply-demand fundamentals from price signals now could easily lead to directional misjudgments later. Similar logic applies to real inventory replenishment in the physical economy. Since the outbreak of the latest U.S.-Israel-Iran conflict, social inventories of major domestic chemical products in China have undergone about two months of passive destocking. According to Jinlianchuang statistics on representative liquid bulk chemical inventories, as of the week of April 28, average social inventories for products like methanol, ethylene glycol, styrene, pure benzene, and toluene have fallen by approximately 26% compared to pre-conflict levels, with a turning point yet to materialize. This anomalous inventory behavior is essentially consistent with financial market logic: having endured two months of passive destocking, and with negotiation results imminent, there is even less reason to initiate active restocking and bear the risk of sharp price volatility. Consequently, real terminal demand in the economic system is also partially absent.
2) The absence of financial longs combined with the absence of physical restocking directly leads to "lopsided" macro analysis. At this stage, we can only analyze supply-side logic but cannot effectively assess the true state of demand. Questions like whether buyers are willing to accept physical goods in the spot market, the true size of the supply-demand gap, when the restocking cycle will begin, and whether terminal consumption is systematically suppressed cannot be reliably answered before key events conclude. If demand-side signals are distorted, then macro analysis, monetary environment assessments, and even corporate earnings forecasts based on demand extrapolation are also in a state of "absence." Relying solely on supply for supply-demand analysis, whether for commodities or stocks, limits the potential slope and height of price movements.
Macro-driven trading funds are expected to retreat in phases, potential entrants are likely to adopt a wait-and-see approach, and allocators may reduce positions at higher price levels.
Judging by currently disclosed information on U.S.-Iran negotiations, a resolution window could open in mid-to-late May, with a potential Trump visit to China becoming another focal point. Under typical博弈logic, before a Trump visit, both China and the U.S. are likely to maintain a controlled and moderate博弈姿态. Post-visit, issues may become more complex, with a marginally higher probability of摩擦escalation. If this is coupled with the positive factor of a U.S.-Iran agreement being reached, late May could represent a阶段性peak for market sentiment and risk appetite this year. Since April 9, domestic broad-based ETFs have seen continuous net redemptions, with 18 consecutive trading days of net outflows as of May 7, totaling approximately RMB 262.5 billion. Notably, redemption规模has expanded since the Shanghai Composite Index consolidated above the 4100-point level. Considering the current macro environment临近major event windows and the赚钱效应accumulated since late March in the Shanghai index, we believe the recent pattern of redemptions at highs in the ETF market likely reflects large-scale allocators reducing positions. While we can choose to ignore these short-term博弈disturbances, we must be fully prepared for放大volatility around event windows, given the significant presence of quantitative and trend-following strategies in the market. Their mechanical reactions to volatility and momentum signals could significantly amplify two-way price swings around these events.
The rotation from high to low positions among existing capital will be quite pronounced, with low-position rallies being a behavioral outcome rather than a signal.
When allocator inflows are limited but market trading remains active, frequent high-to-low rotations are common. Recent market trading enthusiasm remains high, with turnover exceeding RMB 3 trillion for each of the first three trading days in May. Active capital's近期investment strategy involves continuously寻找low-position themes or new catch-up themes for rotation. Commercial aerospace and robotics属于low-position sectors, while themes like token factories and token agents属于new题材. As for more conservative capital, their strategy is to寻找low-position sectors with earnings or marginal changes, such as gaming,恒生科技(cloud), or even real estate (the recent steady performance of real estate ETFs and港股leading companies breaking previous highs mainly reflect reactions to partial stabilization in二手房prices in some top-tier cities and further宽松policies in more cities). These high-to-low rotation characteristics are very evident in the ETF market.
Against the backdrop of strengthening structural market trends, actively reducing portfolio volatility is the preferable choice.
The momentum in tech stocks across the U.S., China, Japan, and South Korea has been strengthening. However, even the strongest industry trends can experience剧烈volatility if short-term slopes become too陡峭. We constructed a Momentum Strength Index (MSI) to measure the momentum intensity of recent global major tech indices. Currently, South Korea's KOSPI MSI has entered the overheated topping zone. The Nasdaq 100 and Nikkei 225 are in the strong acceleration zone. The ChiNext Index is very close to the threshold of the strong acceleration zone. Specifically, the latest MSI reading for the KOSPI is 92.4, indicating market momentum is nearing extremes, potentially even leading to a "short-squeeze" style rally regardless of cost. Historically, excessively high MSI readings correspond to临界points for short-term tops, with a very high probability of subsequent剧烈volatility. The latest MSI readings for the Nasdaq 100 and Nikkei 225 are 86.3 and 80.0, respectively. While not directly implying imminent momentum exhaustion, they are clearly beyond historical norms, suggesting limited room for further significant gains and heightened sensitivity to negative news. As for the ChiNext Index, the latest MSI reading is 64.4, still within a historically reasonable range overall. It is worth noting that before the pullback on May 8, the reading on May 7 was 71.9,非常接近the strong acceleration zone临界value of 75. We believe that even with strong AI-related industry trends,异常momentum intensity and陡峭price slopes in the short term likely预示increased subsequent volatility. In this context, for existing capital, actively reducing portfolio volatility may be the better choice.
Short-term dynamics are becoming more complex, but from a medium-term perspective, AI and energy/chemicals remain the primary sources of supply-demand gaps.
1) Rapid AI iteration has created new supply-demand gaps in different fields for four consecutive years. The first stage (2023-2024) supply-demand gap for computing power cards was primarily driven by scaling effects in the training端. The second stage (2025) will see推理端scaling laws drive new demand for computing power, storage power, and power facilities. The third stage (2026) will begin with the爆发of agents like Claude Code and Codex,升级AI from "usable" to "useful." Demand from real work scenarios will drive a大幅increase in token usage. Weekly token consumption on the OpenRouter platform surged from 8.25 trillion at the end of January to over 20 trillion. Anthropic's annualized data shows an 80-fold growth in revenue and usage for Q1 2026. Strong demand continues to reinforce the necessity of computing power infrastructure, dispelling market concerns about the commercial viability of AI investments. More critically, agents have shown clear differentiation, creating user stickiness, which forms the basis for垄断and sustained超额profits, unlike the fierce competition among homogeneous models. During this stage, the storage gap gradually expands from HBM to traditional DRAM and even flash memory, while supply-demand gaps for other AI server components like CPUs也逐渐emerge. With growing demand for cost-effective models and the maturation of domestic models, domestic cloud services are also experiencing supply tightness and price increases. Looking ahead, the AI industry trend has high certainty, with demand爆发continuously exposing new supply-demand gaps. The main矛盾lies in the slope,节奏, and valuation of the行情.
2) Supply constraints or收缩could create sustainable supply-demand gaps in energy/chemicals. In contrast to AI, energy/chemicals, as another major supply-demand gap this year, are primarily driven by supply-side收缩rather than demand-side expansion. Looking at the overall petrochemical chain, future crude oil supply elasticity will be significantly higher than that of processed products, forming the basis for sustainable price spreads and超额profits. Regions suitable for石化energy processing capacity expansion are quite limited: European refining and chemical costs are high, with real needs for energy diversification and green transition; for energy security and dual-carbon goals, China has implemented a strategy to actively reduce石化energy consumption; the Middle East was originally the most suitable region for extending into石化and chemical下游, but geopolitical conflicts cast a shadow over attracting private sector investment, and investment progress and efficiency in the region have historically lagged behind Asia; Japan, South Korea, and emerging Asian markets have暴露serious supply chain stability issues during this U.S.-Iran conflict, which may推动subsequent energy transition and reduce dependence on单一energy sources. Long-term supply constraints may form sustainable supply-demand gaps. However, due to overall stable demand, the energy/chemical chain leans more towards dividend or value attributes, with relatively lower爆发性.
We believe this year's AI + energy/chemicals combination is analogous to the AI + dividend play of 2023-2024 and the anticipated AI + resource theme of 2025.
This barbell-like structure represents the加剧of K-shaped differentiation across various fields globally, with very few areas possessing truly sustainable supply-demand gaps and超预期profits. AI serves as the进攻端choice driven by demand, while energy/chemicals (new energy, traditional energy, refining/chemicals, basic chemicals) are more suitable to replace the dividends of 2023-2024 and the resources of 2025 as a choice for relatively稳健returns. Observing the trend of oil and石化ETFs, the price action around the March 2026 peak in energy/chemicals行情closely resembles that of the四大banks around May 2023. In terms of配置, the underlying logic remains the重估of pricing power in China's优势manufacturing, with the most representative industries being new energy, chemicals, non-ferrous metals, and power equipment. Continue to closely monitor the progress of domestic AI. On the hardware side, the爆发of the "volume" logic remains the direction with relatively larger预期差on the AI chain, while advancements in domestic models are expected to推动量价齐升in cloud services. We are看好domestic computing power and cloud platforms. Additionally, it is建议to continue增配some low-valuation品种, focusing on securities and insurance. For cyclical涨价品种,景气度remains持续for周期成长品like the AIDC chain and lithium battery chain, but the预期差is relatively smaller, and the trend of缩圈is faster.建议关注the tightest supply-demand links, reflected in recent涨价frequency, mainly including CCL, fiberglass, high-speed silicon, electronic specialty gases, optical fiber, MLCC, chromium, lithium carbonate, rare earths, and carbon fiber. For traditional周期品,建议focus on品种where real产能出清has occurred or where there are absolute supply constraints, such as磷化工, MDI, spandex, dyes, glyphosate, urea, rubber, refrigerants, etc.
Risk Factors Escalation of摩擦in Sino-U.S. 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 conflicts in regions like Russia-Ukraine and the Middle East; slower-than-expected消化of China's real estate inventory.
Comments