Deleveraging Nears Its End, Creating a Reallocation Window for Chinese Technology Stocks

Overview

This report assesses whether Chinese technology stocks have entered a stage suitable for reallocation after the sharp correction in July. UBS judges from margin financing balances, collateral ratios, fund holdings, ETF flows, global leverage changes, earnings forecasts, valuations, and global AI demand that the worst phase of technical selling and deleveraging may have passed. Meanwhile, U.S. hyperscale cloud providers' results, cloud business order backlogs, AI-native enterprise demand, and enterprise AI spending all indicate an improved global AI narrative, while the catch-up of China's domestic technology supply chain and improving domestic GPU supply are expected to accelerate AI data center construction in the second half of 2026.

Core views

First, the sharp correction in China AI technology hardware mainly reflects global technology selling and deleveraging factors, while the core narratives and fundamentals of related companies have not deteriorated in tandem. Second, margin financing balances and share prices have both returned to levels close to April, and the margin collateral ratio remains healthy, indicating that forced deleveraging pressure may have eased significantly. Third, valuation declines and upward earnings forecast revisions together improve the risk-reward profile, making the current period suitable for gradually reallocating to technology. Fourth, uncertainty remains around AI monetization, and recent volatility will also curb investors' willingness to crowd into technology, so market breadth is expected to expand in the second half. Fifth, within technology, preferences are for semiconductor equipment, networking chips, and advanced packaging; outside technology, preferences are for internet, power equipment, non-ferrous metals, and overseas expansion stocks.

Analysis framework

The report uses a combination of top-down and bottom-up approaches: it first evaluates technical conditions and crowding through A-share margin financing balances, collateral ratios, relative fund positions, ETF flows, and overseas leveraged ETF size; then tests fundamentals through earnings forecast revisions, valuation ranges, and global AI demand indicators; and finally forms sector preferences, model portfolio adjustments, and stock screening results accordingly.

Methodology notes

• Market technicalsLeverage and deleveraging tracking

  Assess whether technical selling pressure is nearing its end through margin financing balances, collateral ratios, leveraged ETF size, and share price drawdowns.

  A-share margin financing balances fell from RMB 3 trillion to RMB 2.6 trillion, close to the April level; South Korea's margin financing balance is also close to its early-2026 level, while the size of leveraged ETFs in South Korea and the U.S. has declined by about 50% and 30%, respectively, jointly supporting the view that global technology deleveraging is relatively advanced.

• Fundamental analysisEarnings forecast revisions

  Compare EPS trends and recent analyst forecast revisions for China and global AI technology hardware.

  Despite falling share prices, EPS forecasts for China AI technology hardware are still being revised upward, indicating that the market pullback was not driven primarily by a contemporaneous deterioration in earnings expectations.

• Thematic researchAI demand and monetization validation

  Validate the AI investment cycle using hyperscale cloud providers' results, cloud business order backlogs, AI-native enterprise demand, and enterprise AI spending.

  Microsoft's and Amazon's results, increased cloud business order backlogs, and a 25% quarter-on-quarter increase in average enterprise AI spending are all used to support the improvement in the global AI demand and monetization narrative.

• Valuation analysisRelative valuation and historical range comparison

  Compare forward P/E and P/B ratios for China and global AI technology hardware, and assess the risk-reward profile in conjunction with historical averages.

  After the correction, valuations of China AI technology hardware are only slightly above historical averages, forming a relatively attractive combination with earnings forecasts that are still being revised upward.

• Company valuationDiscounted cash flow, Gordon growth, and relative valuation

  Use discounted cash flow models, Gordon growth models, as well as P/E, enterprise value multiples, and P/B methods for different Hong Kong and A-share companies.

  The report states that individual stock target prices do not use a single model, but select applicable absolute or relative valuation methods based on industry and company characteristics.

• Asset allocationSector preferences and model portfolio

  Form sector rankings and adjust the model portfolio by integrating fundamentals, valuations, crowding, and earnings growth.

  The most preferred areas are technology hardware, internet, overseas expansion stocks, non-ferrous metals, and power equipment; the least preferred are construction, autos, consumption, and software.

Asset mapping & comparison

Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).

• China AI technology hardware

  Core overweight direction

  Strengths

  Technical deleveraging may be nearing its end, valuations have fallen significantly, earnings forecasts are still being revised upward, and the catch-up of domestic supply chains and AI data center construction provide growth support.

  Weaknesses

  Recent volatility has been high, investor risk appetite has been impaired, and uncertainty remains around AI commercialization delivery.

  Comparison

  Compared with global AI technology hardware, related Chinese stocks corrected in tandem during the global selloff, but their domestic supply chain fundamentals did not deteriorate to the same extent.

  Risks

  Renewed deleveraging in global technology stocks, slowing AI capital expenditure, domestic GPU supply improvements falling short of expectations, and renewed valuation compression.

• Semiconductor equipment, networking chips, and advanced packaging

  Most preferred directions within technology

  Strengths

  They are parts of the AI theme with relatively weaker cyclicality and benefit from semiconductor localization and AI data center expansion.

  Weaknesses

  Some companies still have high valuations and require strong order and earnings delivery.

  Comparison

  Compared with other AI hardware segments that are more volatile or have more uncertain commercialization paths, the report believes these areas have better earnings visibility.

  Risks

  Industry capital expenditure volatility, technology iteration, intensified competition, and downward earnings forecast revisions.

• China internet

  Priority allocation direction outside technology

  Strengths

  Improving earnings trends, a better AI narrative for hyperscale cloud providers, and cheap valuations.

  Weaknesses

  Growth and AI monetization still need ongoing validation.

  Comparison

  Compared with some AI hardware with higher valuations, internet combines lower valuations with potential for earnings improvement.

  Risks

  Weak consumer demand, intensified competition, regulatory changes, and AI investment returns falling short of expectations.

• Power equipment

  Cross-sector allocation benefiting from AI data center expansion

  Strengths

  Cheap valuations, while benefiting from AI data center power demand and the trend toward energy independence.

  Weaknesses

  Project cycles are long, and revenue realization may lag thematic trading.

  Comparison

  Compared with direct AI hardware, power equipment provides more indirect AI infrastructure exposure with less valuation pressure.

  Risks

  Slowdown in AI data center construction, project execution falling short of expectations, and industry competition.

• Non-ferrous metals

  Priority allocation direction driven by earnings trends

  Strengths

  Strong earnings trends and potential benefits from power and data center infrastructure demand.

  Weaknesses

  Earnings are sensitive to commodity prices and macro cycles.

  Comparison

  Compared with highly volatile technology stocks, non-ferrous metals provide different earnings drivers and portfolio diversification.

  Risks

  Commodity price declines, weakening global demand, and rising costs.

• Overseas expansion stocks

  Allocation direction for broadening market opportunities in the second half

  Strengths

  The impact of previous RMB appreciation is expected to gradually fade, while overseas revenue and market expansion can provide sources of growth.

  Weaknesses

  Operating results are vulnerable to exchange rates, overseas demand, and the trade environment.

  Comparison

  Compared with purely domestic-demand assets, overseas expansion stocks have a broader geographic revenue base.

  Risks

  Trade frictions, exchange rate fluctuations, overseas regulation, and localization execution risks.

• BABA.N

  New internet stock added to the model portfolio

  Strengths

  Low valuation, improving earnings trends, and an improved global cloud provider AI narrative; the target price of 195.00 is above the closing price of 126.81 in the table.

  Weaknesses

  FY26 EPS growth in the table is -32%, and UBS's earnings forecast is 2% below consensus.

  Comparison

  As an internet allocation, it can reduce the portfolio's concentrated dependence on pure AI hardware.

  Risks

  AI monetization, competition in core businesses, consumer environment, and earnings recovery falling short of expectations.

• 3908.HK

  New stock added to the model portfolio

  Strengths

  Rated Buy by UBS, with FY26 EPS growth of 39% in the table and UBS's earnings forecast 6% above consensus.

  Weaknesses

  Business performance is relatively sensitive to capital market activity and the trading environment.

  Comparison

  Provides financial market-related exposure outside the technology theme, helping to broaden the portfolio's sector exposure.

  Risks

  Cooling market turnover, weaker capital market activity, and changes in the regulatory environment.

• 603308.SS

  New stock added to the model portfolio

  Strengths

  Rated Buy by UBS, with FY26 EPS growth of 67% in the table, UBS's earnings forecast 14% above consensus, and the target price of 100.00 significantly above the closing price of 48.52.

  Weaknesses

  The FY26 P/E ratio in the table is 48.0x, requiring strong growth delivery.

  Comparison

  Among the newly added stocks, it has relatively high expected earnings growth and substantial target price upside.

  Risks

  High-valuation compression, order delivery falling short of expectations, and downward earnings forecast revisions.

• 600584.SS

  New advanced packaging-related stock added to the model portfolio

  Strengths

  Fits the report's preferred advanced packaging direction; FY26 EPS growth in the table is 72%, and UBS's earnings forecast is 23% above consensus.

  Weaknesses

  The FY26 P/E ratio in the table is 51.7x, and year-to-date return has reached 107%, indicating high crowding and volatility risks.

  Comparison

  Fundamental expectations are strong, but valuation and prior gains are also higher than most newly added names.

  Risks

  Advanced packaging demand falling short of expectations, valuation correction, industry cycle changes, and profit-taking.

Key data

• July decline in China AI technology hardware stocks-32%Average monthly share price performance of the sample tracked by UBS.
• Share of sample with declines of 40% or more36%Reflects the breadth and intensity of the July technology hardware selloff.
• A-share margin financing balanceRMB 2.6 trillionHas fallen from a high of RMB 3 trillion and is close to the April level.
• A-share margin collateral ratio280%The report believes the overall collateral level remains healthy and can provide some buffer against further technical selling.
• Average enterprise AI spendingUp 25% quarter-on-quarterUsed to support the view that enterprise AI adoption is accelerating.
• Change in South Korean leveraged ETF sizeDown about 50%Shows that leverage levels in overseas technology trades have also declined significantly.
• Change in U.S. leveraged ETF sizeDown about 30%Together with South Korean data, forms evidence of global deleveraging.
• BABA.NClosing price 126.81, target price 195.00Added to the model portfolio, rated Buy by UBS; based on the prices in the table, the target price is about 53.8% above the closing price.
• 3908.HKClosing price 21.66, target price 31.40Added to the model portfolio, rated Buy by UBS; based on the prices in the table, the target price is about 45.0% above the closing price.
• 603308.SSClosing price 48.52, target price 100.00Added to the model portfolio, rated Buy by UBS; based on the prices in the table, the target price is about 106.1% above the closing price.
• 600584.SSClosing price 77.75, target price 135.00Added to the model portfolio, rated Buy by UBS; based on the prices in the table, the target price is about 73.6% above the closing price.
• Finalization time of research view2026-08-09 22:53 GMTThe finalization time of the research recommendation disclosed in the report.

Impact & implications

The investment implication is to move from broadly avoiding technology to selectively reallocating to it, but investors should not simply replicate the highly concentrated AI trade seen in the first half. Within technology, priority should be given to semiconductor equipment, networking chips, and advanced packaging, where earnings visibility is higher and cyclicality is relatively weaker; at the portfolio level, industry diversification can be improved through internet, power equipment, non-ferrous metals, and overseas expansion stocks. If margin financing balances stabilize, earnings forecasts continue to be revised upward, and domestic GPU supply improves, the technology sector may receive support from both valuation recovery and earnings growth; if AI monetization is questioned again, broader cross-sector allocation will have greater defensive value.

Risks

• A hard landing in China's property market could hurt economic growth and equity risk appetite.
• Capital outflows related to RMB depreciation could depress market valuations.
• Slow progress on structural reforms could limit medium- to long-term growth expectations.
• If policies fail to adequately address risks related to property, capital flows, and reform, they could trigger market shocks.
• Excessive stimulus could hinder the economy's transition from investment-driven to consumption-driven growth and increase debt at the government and state-owned enterprise levels.
• Uncertainty remains around AI commercialization and returns on capital investment, which may limit the short-term performance of technology stocks.
• Technology stocks have been highly volatile recently; if global leveraged trading heats up again and then reverses, technical selling could reappear.
• Valuations of some preferred technology stocks remain high, and earnings or orders falling short of expectations could lead to rapid valuation compression.

What to watch

• Whether A-share margin financing balances can stabilize around RMB 2.6 trillion, and whether financing flows into the technology sector return to net inflows.
• Whether the A-share margin collateral ratio continues to remain healthy, avoiding a renewed rise in forced liquidation pressure.
• Earnings forecast revisions over the next three months and FY26 EPS trends for China AI technology hardware.
• The pace of improvement in domestic GPU supply and the progress of AI data center construction in the second half of 2026.
• Cloud revenue, AI capital expenditure, and order backlogs of hyperscale cloud providers such as Microsoft and Amazon.
• Growth in enterprise AI spending and the actual monetization of AI applications.
• Fund flows into technology-related ETFs such as the STAR 50 and ChiNext, and the level of government support.
• The size of leveraged ETFs, margin financing balances in South Korea and the U.S., and crowding in global technology trades.
• Whether market breadth spreads from AI hardware to the internet, power equipment, non-ferrous metals, and overseas expansion sectors.
• Earnings delivery, target price adjustments, and valuation changes for the newly added stocks in the model portfolio.