Navigating the second half of 2026 begins with a rebound after micro-liquidity risks ease; further index upside depends on a thaw in dollar liquidity. By July, credit spreads and turnover rates had already fallen to low levels, with broad-based ETF inflows surging, while the STAR 50 and ChiNext indices sit near volume-weighted annual support lines. Recent cooling in Fed rate hike expectations for September suggests dollar liquidity may improve; overall, US inflation expectations have not risen markedly, and Fed Chair Warsh's stance appears hawkish in tone but dovish in action. Three factors suggest the bull market may persist: valuations have corrected significantly, the earnings trend has not reversed, and the AI industry trend remains intact. Breaking previous highs hinges on incremental policy implementation—whether the index can surpass the mid-May Shanghai Composite high of 4,258 points depends on whether earnings recovery can sustain into 2027 after current valuation adjustments. Fundamentals data shows this year's quarterly economic trajectory closely mirrors 2024; if September-October delivers incremental policies similar to Q3 2024, the sustainability of the A-share earnings recovery could be reinforced.
Looking to 2027, where will incremental funds come from? Insurers offer relatively certain but smooth inflows; leveraged capital acts as a sentiment amplifier under heightened risk controls; household deposit migration provides lagging inflows after wealth effects spill over; and public funds contribute repair-driven inflows once returns normalize. This year's growth momentum stems primarily from exports and PPI-driven volume and price gains; looking to next year's earnings, domestic demand, cyclical and new energy sectors may shift from drags to drivers, while tech manufacturing growth could accelerate. However, if a rate hike cycle begins, liquidity-sensitive tech growth names could face significant pullbacks.
Can the AI trend continue? Three short-term key questions: First, how to resolve America's old problem becoming AI's new trouble? Financing pressure rising marginally makes it unlikely cloud vendors will cut near-term capex; the solution path—economic cooling and policy support—ultimately leads to monetary easing. Second, have trapped positions eased? Public fund crowding in TMT remains in the mid-to-late stage; after July's sharp correction, AI's market weight and fund holdings have clearly declined. Third, is this a sharp peak or an M-top? Current AI dynamics resemble an M-top—fundamentals validate AI's prosperity continuation, participation includes more diverse capital, and technological progress could further lift valuation ceilings.
Beyond AI, what else to allocate? Cyclical resources offer price elasticity under PPI normalization, with AI's industrial wave creating new growth poles and near-term suppressing factors clearly easing, strengthening valuation recovery momentum. High-dividend names serve as core positions reducing volatility and improving Sharpe ratios. Leading companies outperform during inflation and earnings recovery periods. Consumption shows moderate income and confidence repair.
Navigating H2 2026
July's low volume corresponds to low prices; a rebound follows after micro-liquidity risks ease. Credit spreads and turnover have retreated to lows: as of end-July, the A-share turnover rate (based on free-float shares) bottomed at 3.1%, near the April 2026 US-Iran conflict low of 3.0% and the April 2025 tariff shock level of 2.8%. Credit spreads continue declining this year—the AA vs AAA corporate bond yield differential now stands at 8.6bp, below the pre-September 2024 low of 12.5bp. Broad-based ETFs saw massive inflows: cumulative net buying from July 13-31 reached 301.3 billion yuan, surpassing the September 2024 period's 280 billion yuan, with product scope expanding from CSI 300 and CSI A500 to include STAR 50 and ChiNext—large-scale market stabilization inflows clearly support bottom-level chip positioning.
The STAR 50 and ChiNext indices sit at volume-weighted annual support lines. Compared to traditional MA250, volume-weighted annual lines more effectively reflect market equilibrium points formed by buying and selling. As of August 3, the STAR 50's intraday low of 1,549 points approached its volume-weighted annual line of 1,531, while the ChiNext's July 30 intraday low of 3,158 points broke below its volume-weighted annual line of 3,296. Historically, the ChiNext showed strong support near volume-weighted annual lines during bull phases in May 2014, July 2014, March 2021, and August 2021.
If the index rises further within the year, dollar liquidity recovery becomes necessary. Recent unexpected contraction in US nonfarm payrolls alongside moderate core PCE declines have reduced rate hike urgency; market expectations for a September Fed hike fell significantly from 67% on August 3 to 33% by August 19, with expectations now pricing only one hike in 2026 and rates held in 2027.
Breaking down the past three months of Treasury yields, the rise is more driven by real rates, reflecting a still-strong US economy where sustained AI capex provides support; meanwhile, inflation expectations measured by the yield gap have declined from 2.5% in May to the current 2.2%-2.3%. Trump's election prospects and Treasury payment pressures transmit to Fed Chair Warsh; despite hawkish statements and communication frameworks, his actual orientation—no dot plots, opposition to excessive forward guidance—leans dovish.
Three Reasons the Bull Market May Persist
Valuations have corrected significantly. The A-share median P/E has fallen from a February 2026 high of 31 times to a low of 23.5 times, approaching the 23 times level seen during the April 2025 tariff shock when the Shanghai Composite was at 3,100 points. The earnings trend has not reversed: industrial enterprise profits for the first six months grew 18.7% year-over-year, with full-year non-financial net profit growth optimistically projected at 15%-20%; since May, the earnings upgrade/downgrade structure has improved markedly. Across sectors over the past six months, 2026 forecast net profit growth revisions show clear improvement in overall A-share non-financial earnings expectations; notably, AI-related tech manufacturing and cyclical resource sectors—building materials, nonferrous metals, chemicals, electric equipment, electronics—show the largest upward revisions.
The AI industry trend remains intact. Cloud vendor capex continues upward revisions; late July disclosures from four major North American cloud vendors raised combined annual capex to 720-745 billion US dollars (median ~732.5 billion) to maintain large-scale data center and computing investments, confirming the overseas AI trend has not ended. Additionally, Kimi K3 topped open-source models in late July with strong domestic computing demand—AI fundamentals remain difficult to falsify.
Breaking the previous high requires incremental policy. The core question for surpassing the mid-May Shanghai Composite high of 4,258 points is whether earnings recovery can sustain into 2027 after valuation adjustments complete. The key lies in whether domestic demand and traditional sectors see a right-side inflection point in fundamentals. We believe the possibility of policy stimulus similar to September 2024 cannot be ruled out within the next six months. Fundamentals data shows this year's quarterly trajectory most closely resembles 2024—strong Q1, declining Q2, Q3 awaiting observation. July manufacturing PMI recorded 49.2, entering contraction territory with super-seasonal decline compared to recent years.
The July 2024 Politburo meeting noted "insufficient domestic effective demand, diverging economic operations, persistent key area risks, and growing pains in新旧动能转换," while emphasizing "accelerating full implementation of established policies and promptly reserving and appropriately launching incremental policy measures." Subsequently, August 2024 accelerated existing policy implementation: new special bond issuance notably accelerated; 300 billion yuan of "two new" funds were fully allocated by end-August, with trade-in programs lifting September consumption to positive territory. By September 24, incremental policies concentratedly launched with dual easing of monetary and fiscal policy: 0.5pct RRR cut, 20bp policy rate reduction, average 0.5pct cut in existing mortgage rates, pioneering swap facility at 500 billion yuan initial tranche, and stock buyback and refinancing lending at 300 billion yuan initial tranche.
From a policy perspective, this year also resembles 2024. This July's Politburo meeting similarly mentioned "attaching high importance to difficulties and challenges in economic operations, strengthening confidence, and facing difficulties head-on," while proposing "fully leveraging existing policy effectiveness and timely planning practical incremental policies." If incremental policies emerge in September-October, both fundamental K-shaped divergence convergence and earnings growth sustainability could strengthen, making index breakout above previous highs possible.
K-shaped divergence reflects earnings divergence. Industry differentiation after Q1 report season shows clear earnings-pricing characteristics. Since May 1, A-share pricing has been based on earnings revisions, with industry gains and losses reflecting fundamental K-shaped divergence—AI leading gains while domestic consumption and property chains lag; cyclical resources saw earnings upgrades but underperformed on external liquidity factors. Since the July 20 rebound, some resource products previously subject to "true siphoning" have experienced fundamental re-pricing; overall market pricing remains earnings-based with AI and resources as main lines.
Under weak restocking, style tilts toward growth: policy easing favors growth, as seen in 2013-2014 when TMT tech sectors led gains.
Is 2027 Still a Bull Market?
Where will incremental funds come from? Insurers provide relatively certain but smooth inflows; compared to index levels, insurers focus more on asset valuation attractiveness and dividend yields. While quarter-end profit-taking may cause seasonal net buying slowdowns, the long-term trend of increasing equity allocation ratios remains unchanged—overall holdings growth shows drip-like quarterly increases rather than pulse-style surges. Foreign capital provides sustained strategic allocation increases as Chinese asset advantages stand out; with delayed Fed rate hike expectations, stable RMB appreciation, and geopolitical easing, foreign inflows may accelerate. Private funds offer performance-driven structural elasticity; over the next six months, the core trigger for accelerated marginal inflows may be performance-driven positive feedback effects. After July's adjustment and gradual liquidity risk relief, the market has entered a counterattack trend; AI tech leading gains is gradually rebuilding private fund performance positive feedback, driving incremental capital entry. Leveraged capital acts as a sentiment amplifier under upgraded risk controls; after July's rapid deleveraging, margin trading sentiment remains in slow repair; over the next six months, in a market recovering with oscillation, margin funding likely shows moderate total expansion with more cautious structure. ETFs serve as catalysts for industrial trend consensus; scale ETFs as medium-to-long-term market stabilization tools have recently shown counter-cyclical characteristics. The market has passed July's worst moment; upward momentum accumulates while external support gradually diminishes—future incremental space may shift from medium-to-long-term funds toward industry/theme ETF-led industrial trend capital. Household deposit migration provides lagging inflows after wealth effects continuously spill over; as August's market counterattack begins, household risk appetite and willingness to enter markets are slowly repairing. Public funds offer repair-driven incremental inflows as wealth effects return; new issuance may resume upward trends, but volatility from market conditions means public fund inflows appear more as repair-driven rather than steady-state additions, naturally weaker in sustainability than insurer-led medium-to-long-term capital.
M1-PPI-corporate earnings have gradually transmitted to the earnings end. M1 recovered from -7.4% in September 2024 to 3.6% in July 2026 (old caliber); PPI recovered from -3.6% in July 2024 to 3.5% in July 2026; A-share non-financial cumulative net profit attributable to parent recovered from -13% in December 2024 to 11% in March 2026. This year's growth sources include PPI and exports: PPI monthly year-over-year recovered from -1.9% in December 2025 to 3.5% in July 2026, driving A-share non-financial gross margins from 17.7% at end-2025 to 18.6% in Q1 2026; export value monthly year-over-year surged from 6.4% in December 2025 to 23.9% in July 2026.
Looking to next year's earnings, new growth drivers: domestic demand, cyclicals and new energy shift from drags to drivers, tech manufacturing accelerates. In the next two years' net profit growth outlook, one category includes past 2-3 year drags expected to see bottoming recoveries: food and beverage, pharmaceuticals, light manufacturing within domestic demand; cyclical resources (coal, steel, chemicals, building materials); and new energy chains (electric equipment). Another category includes electronics, defense, computers, and machinery—tech manufacturing sectors where growth may further accelerate.
If a rate hike cycle begins, liquidity-sensitive tech growth names could experience significant pullbacks. Across three rate hike cycles since 2000, US tech represented by Nasdaq declined notably within 1-3 months after hikes, averaging 10.5% declines over three months. A-shares may react insufficiently in the first month but transmit effects within 1-3 months after hikes, with liquidity-sensitive growth, high-valuation, and tech directions typically falling deeper.
Can the AI Trend Continue?
Short-term key question one: America's old problem becoming AI's new trouble. The recent macro environment—rapid long-end Treasury yield increases, gold rising, dollar weakening—is not unprecedented but represents America's old problem, similar to Q2 2025 markets. However, this round differs: AI has transformed from bystander to participant in US fiscal credit issues, turning America's old problem into AI's new trouble. Previously, AI achieved separate pricing from US government credit through strong earnings and abundant cash flows, rising independently of fiscal turmoil. This round, AI has shifted from cash flow-driven to more debt-financing dependent; long-end rate increases now transmit directly to AI fundamentals through financing costs.
The core reason AI assets got drawn into US fiscal credit issues is cloud vendors' capex sources shifting toward debt financing. Since August, macro conditions resemble Q2 2025 but AI performance diverges clearly. Regarding competition for long-term capital, this resembles "price crowding out" rather than "quantity crowding out": US Treasury absolute yield levels remain attractive, and AI can still secure sufficient market subscription through higher credit spreads. While concerns grow that financing pressure could reduce AI capex and affect industrial trends, the transmission chain currently remains at the financial level—marginal financing environment deterioration unlikely triggers near-term cyclical inflection in AI itself. The solution: economic cooling and policy support both ultimately lead to monetary easing. Compared to AI companies, some traditional real estate/durable consumer sectors with weaker balance sheets and higher financing dependence may face pressure earlier under tightening credit conditions; AI possesses industrial trends—it may be "dragged down" but not "drown." This contradiction may resolve through two paths: natural economic cycle cooling or proactive policy front-loading. Overall, current US fiscal credit issues may follow H2 2025's path—shifting from acute shocks to chronic constraints after economic cooling and rate cut expectations open; during this process, quality AI assets with solid fundamentals may regain capital favor.
On demand-side capex health: comparing to the tech bubble, demand came from startups with extremely short histories in internet/software, capex dependent on external equity financing, forming "VC→IPO→secondary market" capital chains from venture capital through IPOs to post-listing follow-on offerings. Demand-side capex lacked self-sustaining operational cash generation; when capital markets fluctuated, "burning cash for growth" internet companies lost continuous financing ability. Current AI: demand comes from North American cloud vendor leaders with strong profitability and stable cash flows; current capex mainly derives from operating cash flows. Risk signals appear as Amazon, Meta and other leaders begin debt financing to support capex; liquidity expectations and credit spreads may affect borrowing costs, potentially causing capex declines.
Short-term key question two: Have trapped positions eased? Historical public fund crowding formations and collapses typically accompany price contribution rates showing "V-shaped" changes through five stages: price dominance, position adding dominance, position adding and price resonance, peak stage, and crowding collapse. Referring to historical experience, price contribution rates typically exceed 100% at crowding peaks, followed by price contribution peaking and crowding collapse; current public fund crowding in TMT may still be in the mid-to-late stage. After July's sharp AI correction, AI's market weight and public fund holdings have clearly declined: in the CSI 800 benchmark index, June's AI rally drove electronics weight up sharply from 19.6% on May 31 to 25.5% on June 30; after July's AI decline, electronics weight fell back to 19.2% by July 31.
Short-term key question three: Sharp peak or M-top? Sharp peak industries show rapid short-term capex increases: typical examples include nonferrous metals, where urbanization-driven demand volume growth accelerated supply expansion, with industry capex/(depreciation+amortization) rising from 2.6 in H1 2006 to 3.5 in H2 2008; the 2015 internet similarly showed capex/(depreciation+amortization) rising from 1.5 in H2 2014 to 3.1 in H2 2016. M-top industries show relatively prudent capex with technological breakthroughs creating new supply-side drivers: M-top industries' growth space is not statically given but continuously redefined through technological progress, product iteration, or application scenario expansion; after the first prosperity cycle materializes, new technology paths (higher efficiency processes, next-generation product forms) reopen profit space, maintaining profit growth at high levels after peaking, providing market basis for secondary industry valuation pricing.
The sharp peak versus M-top distinction fundamentally stems from earnings sustainability and growth space certainty. Sharp peak industries possess clear, measurable growth space, meaning market perceptions of future industry value are "one-time complete"; M-top industries differ in that after earnings peak, high levels maintain or even expand growth space through technological progress catalysts. Comparing current AI: it more resembles an M-top. First, fundamentals: North American cloud vendor interim reports validate AI prosperity continuation. Second, capital structure: this round's AI participation includes more diverse capital, not solely trend or leverage capital pricing. Third, with technological progress and large model capability breakthroughs, AI industry valuation ceilings may further rise. From PCB and optical module industrial trends, supply-side capacity will continuously release over the next year, with gross margins potentially further increasing.
As high prosperity attracts accelerated capital return, a new round of capex expansion cycles gradually begins. Overall, the next two years will see domestic and international hardware manufacturers' capacity gradually releasing; supply-side capacity release may flatten gross margin expansion slopes and face peak-return pressure. Referencing the previous new energy cycle: the lithium battery industry began capacity expansion in Q2 2020, with capex/(depreciation+amortization) rising from 1.9 to 4.9 by Q4 2022; as capacity released and supply-demand patterns reversed, industry gross margins peaked at 22% in Q2 2022 after rising from 18% lows, with stock prices peaking as early as Q4 2021. The photovoltaic industry showed capex/(depreciation+amortization) rising from 1.9 in Q2 2020 to 3.5 in Q4 2022, maintaining 3.4-3.5 high expansion for the following year; gross margins peaked at 25% in Q1 2023 after rising from 20% lows, with stock prices peaking in Q3 2022.
Global AI industry chain net profit flows: the US and Korea capture most shares, with mainland China accounting for less than 5%. Global AI core profits remain highly concentrated among a few US and Korean tech leaders and upstream key segments: according to Gougu Big Data forecasts, the 2026 global AI industry chain estimated net profit pool reaches 637 billion US dollars, showing high concentration: the US at 314 billion (49% global share), with Nvidia alone accounting for 207 billion; Korea at 223 billion (35% share), mainly contributed by SK Hynix and Samsung memory businesses. Mainland China has only about 26 billion, less than 5% share, mainly flowing to Zhongji Innolight, Foxconn Industrial Internet, Eoptolink and other optical module and OEM segments.
A-share AI trading logic shifts from price beta to positioning alpha. Focus on domestic AI hardware chain positioning advantages (chips/data centers/fiber optics/semiconductor equipment/lithography/EDA). Within hardware, subsequent screening requires greater emphasis on industry chain positioning. On one hand, focus on industries with less full future capacity release and smaller supply shocks, measured by CAPEX/EBITDA (observing capex expansion relative to profit expansion); on the other hand, focus on industries with strengthening chain bargaining power, measured by net increases in (advances receivable - advances payable) relative to revenue. After screening AI hardware segments, currently noteworthy sub-sectors concentrate in chip design, optical modules, EDA, fiber optics, semiconductor equipment, computing power leasing, and lithography equipment.
What Else to Allocate Beyond AI?
Cyclical resources: price elasticity release under PPI normalization. With PPI turning positive, reflation shifts from expectation to reality; this round's PPI recovery scope continues expanding, driven by global supply rigidity constraints. Historically, cyclical sector earnings continuously repair after PPI turns positive; 1-3 quarters after PPI normalization often mark institutional capital acceleration into cyclical markets, such as nonferrous metals/chemicals in the 2021 new energy cycle and steel/coal/building materials in 2016's supply-side structural reform. Nonferrous metals: global supply rigidity. Industrial metal supply rigidity is a global issue: major copper mine ore grades continue declining with insufficient long-term capex causing scarce new projects; aluminum faces domestic capacity ceilings and "dual carbon" policy constraints with little room for new capacity; overseas, high electricity costs after Europe's energy crisis weaken restart and expansion willingness. Basic chemicals: capacity contraction and supply advantages. Domestic tightening environmental and energy policies combined with permanent European high-energy-capacity exits significantly slow new capacity release, with supply constraints increasingly rigid. AI's industrial wave creates new growth poles that may intensify metal supply-demand tensions: global AI data center and supporting power infrastructure large-scale construction uses copper far more intensively than traditional infrastructure, sustained throughout the computing expansion cycle, creating triple resonance with traditional grid investment and EV consumption, further widening supply-demand gaps.
Near-term resource-suppressing factors have clearly eased, strengthening valuation recovery momentum. Gold benefits simultaneously from US fiscal credit concerns, long-term dollar credit revaluation, and potential future real rate declines—possessing relatively complete upside logic under both fiscal risk continuation and monetary policy shifts scenarios. Copper/aluminum and other nonferrous metals benefit from data center, grid, and power equipment investment growth; supply constraints help maintain price centers, with demand logic continuing until AI capex clearly declines.
High-dividend: lowering volatility and improving Sharpe ratios as core positions. We believe lowering expected returns, reducing portfolio volatility, and improving Sharpe ratios should be relatively common market demands for H2. Combining dividend yields and volatility (weekly return volatility over the past year), focus on higher dividend and low volatility sectors: banks, expressways, white goods, baijiu, telecom operators, traditional Chinese medicine, and ports. If inflation weakens, barbell allocation may return; public fund holdings have hit historical lows. One end of the barbell is pure dividend low-volatility assets, the other small caps; historical experience shows barbell allocation typically underperforms large-cap growth styles during inflation returns; during inflation weakening, large-cap growth EPS elasticity expectations underperform small-cap growth while dividend yield contributions underperform large-cap value. Active equity public funds' dividend sector holdings have continuously declined to 9.3% by Q2 2026, at historical lows; once later market style rebalances, incremental funds' potential re-entry space is notable.
Leading companies: during inflation and earnings recovery periods, large-cap leaders outperform. Historical experience shows large-cap leader styles typically outperform during inflation returns; inflation's return further drives listed company earnings recovery, and in previous earnings uptrend cycles, large-cap leaders generated relative returns versus small-mid cap companies. From enterprise quality perspectives, large-cap leaders typically possess more solid moats and stable ROE levels. As the economy gradually enters the stock phase, the ROE stability demonstrated by A50 and CSI 300 indices over the past five years is expected to continue.
Consumption: income and consumer confidence show moderate repair. From income and confidence indicators, household pressure shows some stabilization signs, with transmission to actual consumption growth still lagging. Per capita disposable income wage income cumulative year-over-year rose slightly from 4.9% in Q1 2026 to 5.3% mid-year; consumer confidence index income sub-component and consumption willingness sub-component have continuously repaired from 2022 lows to near-neutral 100 levels, indicating marginal improvement in household income expectations and consumption willingness.
Risk Disclosures
1. Macroeconomic recovery may fall short of expectations. 2. Tariff policies may exceed expectations, potentially impacting related industry chains and domestic exports. 3. Historical experience does not guarantee future results: due to changing market environments and other factors, historical data-derived experiences may fail in the future. 4. Incremental funds represent only ideal-assumption market paths and do not constitute inevitable market forecasts. Market advancement highly depends on multiple conditions including corporate earnings repair, domestic liquidity environment, foreign risk appetite, and household risk appetite repair; if fundamentals disappoint, external macro disturbances occur, or household capital entry willingness is insufficient, markets may fail to complete all four-stage transmissions; market volatility, geopolitical conflicts, and other black swan events may also alter capital rhythms.
Comments