AI Growth Cannot Offset Property Drag; Dual K-shaped Divergences Intensify

Overview

This Nomura report presents a dual narrative of China’s current economy: on one hand, emerging sectors led by AI are experiencing explosive growth, supporting the economy through investment, exports, and productivity gains; on the other, the property bubble that burst in 2021 continues to deepen, severely impacting domestic demand, local government finances, household wealth, and the credit system. The core thesis is that the AI boom and property crisis are intertwining to create two mutually reinforcing K-shaped divergences—one across labor and wealth holders (population-level), and another between a few 'smart cities' and numerous 'declining cities' (geographic-level). The report argues that the AI boom is too concentrated and insufficient in scale to fill the massive void left by the property downturn, and its benefits may further exacerbate inequality and suppress aggregate demand, posing complex challenges for policymakers.

Core views

**The Real Support—and Limits—of the AI Boom** The report confirms that the AI boom is providing tangible support to China’s economy. Nomura estimates that AI-related fixed asset investment (FAI) will contribute about 0.3 percentage points to GDP growth in 2026. Driven by the global AI supercycle, Nomura has revised its 2026 export growth forecast upward from 4.0% to 8.6%. However, the report cautions against overestimating this support for four key reasons: First, China remains heavily reliant on imported advanced chips for large models and data centers, diverting part of AI investment demand overseas. Second, the export acceleration is primarily driven by price effects—about half of April’s 14.1% export growth came from higher prices for chips and electronics, limiting its real GDP contribution. Third, as a net chip importer, China sees its terms of trade deteriorate when chip prices surge, since import costs rise even faster than export revenues. Fourth, AI could widen gaps within the population and across cities, and combined with the property crisis, these dual K-shaped divergences may further dampen consumption and investment. **The Ongoing Property Crisis and Its Ripple Effects** The report describes China’s property sector as undergoing an unprecedented collapse. Prior to 2021, real estate was the largest engine of growth, contributing roughly 25% of GDP, 38% of fiscal revenue, and 60% of household wealth at its peak. Since then, it has undergone a sharp adjustment: according to NBS data, new home sales volume and value fell by 43.8% and 43.9% respectively from 2021 to 2025; CRIC data—which better reflects market reality—shows that contract sales by the top 100 developers plummeted by 72.7% in value and 80.2% in area over the same period. In early 2026, the slump continued, with NBS-reported new home sales down 14.6% year-over-year in the first four months. The property crash has cost 14 million migrant workers their construction jobs, and housing price declines in lower-tier cities have disproportionately hurt low-income households. The crisis has triggered cascading effects: local governments’ land sale revenues have collapsed (estimated to be down over 80% after excluding purchases by LGFVs), tightening fiscal space; non-performing debt chains are entangled, weakening credit expansion; and the economy is stuck in a demand-driven deflationary environment with falling interest rates, as seen in the declining yield on 10-year government bonds. **Two Interlocking K-shaped Divergences** The report’s most insightful contribution is identifying two intertwined K-shaped divergences. The first is demographic: the AI boom is splitting society into two groups—the “upper arm” comprising capital owners, high-skilled white-collar workers, and asset holders who benefit from AI, and the “lower arm” of workers displaced or partially replaced by AI. The property crisis exacerbates this divide, as housing prices in lower-tier cities have fallen more sharply, eroding the wealth of low-income groups more severely. The second is geographic: unlike the 2000–2021 property boom, which broadly created wealth, the AI supercycle is highly concentrated. Entering the AI era requires massive computing power, proprietary data pools, and top talent—all of which are clustered in a few first-tier cities (Beijing, Shanghai, Hangzhou, Shenzhen). These “smart cities” extract significant economic rents from “declining cities” through models like MaaS (Model-as-a-Service), creating a resource-siphoning effect. Critically, these two divergences reinforce each other: as top talent and capital flow into smart cities, displaced workers are pushed into gig or low-end service jobs, depressing wages in declining cities, worsening service-sector deflation, and further weakening aggregate consumption. Additional structural factors—rapid aging, a surge in university graduates (projected at 12.7 million in 2026), and a highly unequal pension system (55% of retirees receive only RMB 244/month)—entrench these divides. **Geographic Concentration of the AI Economy and Implications for Property Markets** The report details which cities stand to benefit most from the AI boom. Beijing, Shanghai, Hangzhou, and Shenzhen are the clear winners, serving respectively as the “brain” of the AI ecosystem (Beijing), the hub for advanced chip manufacturing and packaging (Shanghai), the center for large models and AI+ applications (Hangzhou), and the nexus for chip design and embodied intelligence (Shenzhen). Some second-tier cities like Hefei, Wuhan, and Suzhou may indirectly benefit through specialization in robotics, memory chips, or optoelectronics, but the highest-value IP returns still flow to first-tier cities. Meanwhile, western regions hosting “East Data, West Computing” data centers receive substantial infrastructure investment, but these facilities are low-labor-intensive and fail to generate enough white-collar jobs to revitalize local property markets. The report concludes that a genuine property market recovery may be limited to a few smart cities—especially those that have lifted home purchase restrictions—while most of China, particularly smaller cities, will continue facing severe headwinds due to outmigration and deeply pessimistic expectations. **Policy Dilemmas and Potential Solutions** The report argues Beijing faces a difficult dilemma: it needs export growth to offset the collapsing property market, but the resulting trade imbalances are unsustainable. To address the dual K-shaped divergences, the report suggests several policy directions: first, avoid complacency from the AI boom (as earlier optimism about the “New Three” exports proved misplaced); second, proactively monitor and mitigate the growing gap between the “upper” and “lower” arms of the K-curve—for example, by supporting local governments in developing “sovereign AI” to help lower-tier cities share in AI gains; third, strengthen the social safety net, especially by reforming the highly fragmented pension system (where fiscal subsidies heavily favor formal retirees); fourth, consider slowing the adoption of certain technologies (e.g., full self-driving/FSD) to protect blue-collar jobs. But most critically, despite the AI boom, Beijing may ultimately need to significantly ramp up policy efforts to clean up the property sector’s mess and accelerate fiscal reforms to provide local governments with a more stable tax base.

Analysis framework

The report employs a hybrid 'macro narrative + structural decomposition' analytical framework. It begins with macro aggregates, quantifying the separate contributions and drags of the AI boom and property crisis on GDP, investment, exports, fiscal health, and household wealth, presenting a clear picture of two opposing forces. Then, moving beyond simple aggregation, it introduces the 'K-shaped divergence' lens to analyze how the AI boom inherently exacerbates inequality—across labor markets (by skill/income) and regions (through urban resource reallocation)—rather than benefiting everyone equally. In analyzing the AI sector, the report avoids conceptual vagueness and instead breaks down its economic contribution into three quantifiable channels: AI-related fixed capital formation, AI service revenue (APIs, cloud services), and AI-linked exports. Particularly in export analysis, it constructs a 'three-layer growth framework'—price cycle layer (memory chips, server components), structural layer (PCBs, power infrastructure), and emerging layer (optical interconnects)—to distinguish sources and sustainability of growth. Crucially, the report highlights that 'China remains a net chip importer,' showing that while rising chip prices boost export values, they increase import costs even more, worsening China’s terms of trade—a more nuanced approach than past analyses of the 'New Three' exports. For the property sector, the focus is on indirect macroeconomic spillovers via negative wealth effects, local fiscal stress, non-performing debt chains, and deflationary spirals. Finally, the report synthesizes these threads into policy implications, drawing lessons from Japan’s handling of non-performing loans.

Methodology notes

• Industry/Sector Analysis FrameworkSupply-demand framework

  Supply-Demand Framework

  The report evaluates AI’s actual GDP contribution by dissecting both supply-side factors (chip capacity, power supply) and demand-side drivers (AI investment, exports, service revenue), rather than discussing the AI hype generically.

• Industry/Sector Analysis FrameworkVolume-price decomposition

  Volume-Price Decomposition

  When analyzing China’s IC exports, the report separates export value growth into 'volume' and 'price' components. It finds that of the 99.6% year-over-year export growth in April 2026, a staggering 92.6 percentage points came from price increases, not higher shipment volumes—explaining why nominal exports look strong but contribute little to real GDP growth.

• Company Fundamentals & Financial FrameworkEarnings Quality Analysis

  Terms of Trade and Trade Balance Analysis

  The report takes a step back to examine absolute trade balance changes: although China’s chip export growth far outpaces import growth, because the absolute import value is 2.2 times larger than exports, rising chip prices actually increase the trade deficit and worsen China’s terms of trade. This goes beyond superficial export growth metrics.

Asset mapping & comparison

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

• Baidu

  Benefits from development of large AI model (ERNIE Bot) and commercialization of cloud services

  Strengths

  Headquartered in Beijing, Baidu is a core player in China’s AI ecosystem, with top talent, computing resources, and data advantages.

  Risks

  Faces intense competition from ByteDance, DeepSeek, and others; commercialization path remains exploratory.

• ByteDance

  Benefits from large AI model (Doubao) development and massive AI capex

  Strengths

  Leader in AI ecosystem; raised 2026 AI capex plan by 25% to over RMB 200 billion, with strong application deployment capabilities.

  Risks

  Fierce competition in large models; uncertainty around returns on high capex.

• Alibaba

  Benefits from open-source AI model strategy (Qwen) and Alibaba Cloud infrastructure services

  Strengths

  Attracts developers via open-source models, then monetizes through Alibaba Cloud offerings like compute leasing and private deployments—similar to Google’s Android ecosystem model.

  Risks

  Must navigate data sovereignty and compliance requirements; pace of enterprise service market expansion.

• Huawei

  Benefits from surging demand for domestic AI chips; holds ~28% share of China’s AI accelerator market

  Strengths

  Ascend chips integrated with DeepSeek V4 set industry benchmarks; core player in China’s AI supply chain.

  Weaknesses

  Still constrained by lack of access to advanced process nodes and EUV lithography, with yields lagging TSMC.

  Risks

  Risk of tighter U.S. export controls; execution risks for domestic chips at advanced nodes.

• SMIC

  Benefits from surging domestic AI chip demand; handles advanced process manufacturing

  Strengths

  Shanghai is a global hub for AI chip manufacturing and packaging; SMIC is expanding capacity to meet autonomous AI chip demand.

  Weaknesses

  Lags TSMC significantly in advanced nodes (e.g., EUV-related), with notable yield gaps.

  Risks

  Risk of delays in advanced node breakthroughs; escalating U.S. technology blockade.

Key data

• Cumulative decline in new home sales by top 100 developers (2021–2025)72.7%By value, CRIC data
• Cumulative decline in new home sales area by top 100 developers (2021–2025)80.2%By area, CRIC data
• YoY change in new home sales, Jan–Apr 2026-19.7%Top 100 developers (CRIC); NBS figure is -14.6%
• Estimated contribution of AI-related FAI to 2026 GDP growth0.3 percentage pointsNomura estimate
• Revised 2026 export growth forecast8.6%Up from 4.0%, per recent Nomura update
• Estimated contribution of IC price effects to April export growthApproximately halfOf the 14.1% overall export growth in April, about half came from higher chip and electronics prices
• Number of migrant workers unemployed due to property crisis14 millionCumulative construction job losses since 2021
• Core size of China’s AI industry (2025)Over RMB 1.2 trillion
• Estimated 2026 AI capital expenditure (broad definition)Approximately RMB 1.2 trillionEquivalent to 0.8% of GDP, about one-third of the U.S. share relative to GDP
• YoY growth in IC exports, April99.6%Of which 92.6 percentage points came from price increases; volume grew only 3.7%
• Youth unemployment rate (ages 16–24), April16.3%Significantly higher than the overall urban unemployment rate of 5.2%
• Size of flexible employment workforce (end-2024)240 millionAbout 30% of total national employment, 50% of urban employment
• Average monthly pension for rural and migrant workersRMB 244Only 3.8% of public institution retirees’ pensions and 7.3% of enterprise retirees’

Impact & implications

The report argues that the confluence of the AI boom and property crisis will profoundly reshape China’s economy. First, growth will become highly segmented: high-end manufacturing, AI-related services, and export-oriented sectors will continue to benefit, while local services dependent on land finance, low-end manufacturing, and traditional infrastructure investment will face severe pressure. Second, consumer markets will exhibit K-shaped recovery: high-net-worth individuals will maintain strong spending due to tech innovation and capital market gains, but the majority—middle- and low-income groups, especially AI-displaced white-collar workers and unemployed construction laborers—will save more and spend less amid deteriorating income expectations, fueling deflationary pressures. Third, the property market will show stark geographic divergence: premium housing markets in a few smart cities (Beijing, Shanghai, Hangzhou, Shenzhen) may stabilize or rebound first due to high-income inflows and relaxed purchase restrictions, but properties in most other cities will struggle with oversupply and falling prices, becoming illiquid assets. Fourth, local government finances face unprecedented challenges: land sale revenues have fallen over 80% from their peak, and the concentrated nature of the AI economy means tax gains accrue to a few cities, leaving many 'declining cities' with depleted revenue bases—urgently requiring central-local fiscal reform. Finally, on the social front, without timely pension reforms and job protection measures, widening K-shaped divergence could threaten social stability. The report stresses that the AI boom cannot replicate the property sector’s role as a broad-based growth engine, and policymakers must remain clear-eyed and act proactively.

Risks

• If the AI capex boom cools prematurely or chip prices normalize in H2 2026, the price-cycle layer’s export contribution will weaken, narrowing the trade deficit but also reducing economic support.
• Persistent property market weakness—despite multiple rounds of conventional easing by Beijing—could fuel a loss of policy credibility if the downward spiral continues.
• Massive AI-driven displacement of white-collar jobs could push youth unemployment higher (already at 16.3% for ages 16–24), while gig platforms become saturated with displaced professionals, depressing real wages.
• Rapid commercialization of embodied AI technologies like full self-driving (FSD) could disrupt blue-collar employment, further straining social cohesion.
• The highly fragmented pension system (55% of retirees receive only RMB 244/month) may be unable to cope with mass unemployment and early retirements triggered by AI and the property crisis, potentially causing systemic fiscal stress.
• Tighter U.S. export controls on advanced semiconductors could impose hard constraints on China’s ability to train cutting-edge large models, even if application-layer AI remains unaffected.

What to watch

• Track quarterly data on China’s AI capital expenditure (narrow and broad definitions) to verify its share of GDP and growth contribution against expectations.
• Monitor volume-price breakdowns of IC exports; if price contributions narrow and volume rebounds, terms-of-trade pressure may ease—but nominal export growth would also slow.
• Closely watch monthly trends in new home sales (especially top 100 developer data) and prices to assess property market stabilization, particularly the divergence across city tiers.
• Track youth unemployment and flexible employment figures as key indicators of K-shaped divergence in the labor market.
• Watch for Beijing’s potential rollout of 'sovereign AI' support policies for lower-tier cities, progress on pension system reforms, and regulatory stance on FSD and similar technologies.
• Observe whether the performance gap between Chinese and U.S. AI models continues narrowing (per Arena rankings) and if domestic chip procurement reaches the projected ~80% by 2027.