AI Boom Fails to Offset Real Estate Slump, China Faces K-Shaped Divergence Challenge

Overview

This macro research report by Nomura Securities delves into the core contradictions of China's current economy: the coexistence of an AI boom and a real estate slump, forming a K-shaped recovery pattern. The report argues that while AI indeed supports China's economy through fixed asset investment, exports, and productivity improvements, its scale is insufficient to offset the persistent negative impact of the real estate slump. More critically, the AI boom may exacerbate two K-shaped divergences—income and wealth inequality at the demographic level, and the divergence between smart cities and traditional cities at the geographic level—which will weaken aggregate demand. The report maintains its Q2 GDP growth forecast at 4.1%, below the market consensus of 4.7%, and cautions that Beijing cannot assume AI will cure the economic problems caused by real estate, potentially necessitating stronger policy responses.

Core views

The economic support from AI is real but overstated. The report estimates that AI-related fixed asset investment will contribute approximately 0.3 percentage points to GDP growth in 2026 and has revised its 2026 export growth forecast from 4.0% to 8.6%. However, the positive impact should not be exaggerated for four reasons: first, China still heavily relies on imported advanced chips to run AI models and data centers, causing some AI investment demand to leak to other economies; second, the export acceleration driven by AI is primarily price-driven, with about half of the 14.1% export growth in April 2026 coming from soaring chip and electronics prices, contributing little to real GDP growth; third, China remains a net importer of chips, and rising chip prices simultaneously boost imports and exports, worsening trade conditions; fourth, AI may further exacerbate inequality among populations and cities, which, combined with the real estate slump, could weaken consumption and investment demand. The real estate slump persists with profound impacts. Before 2021, real estate was the largest pillar of China's economy, contributing about 25% of GDP, 38% of fiscal revenue, and 60% of household wealth. However, new home sales (by value) declined by 43.8% from 2021-25, with sales of the top 100 developers dropping by 71.6%-72.7%. In the first four months of 2026, new home sales still fell by 14.6%-19.7% year-on-year, showing no clear signs of recovery. The real estate collapse has weakened domestic demand, damaged the balance sheets of governments, businesses, and households, led to massive bad debts, and plunged China's economy into a deflationary spiral. Two K-shaped divergences are forming and reinforcing each other. On the demographic K-shape, the real estate slump has caused 14 million migrant workers to lose construction jobs, with home prices falling more sharply in lower-tier cities, exacerbating income and wealth inequality; the AI boom further divides society into two groups—those who benefit from talent, job security, and wealth, and those who are fully or partially replaced by AI. On the geographic K-shape, the 2000-21 real estate boom was a broad-based wealth creator, but the AI supercycle is a concentrated event, with success dependent on computing density, data pools, and top-tier talent, which are highly concentrated in a few first-tier cities. A handful of 'smart cities' capture most of the national benefits of AI development, even draining resources from other regions. The two K-shapes reinforce each other, as displaced workers are forced into the gig economy or low-end services, depressing wages in traditional cities, exacerbating service-sector deflation, and weakening overall consumption demand.

Analysis framework

The report employs a macro framework to analyze the structural contradictions in China's economy, with its core methodology being a quantitative comparison and transmission analysis of the impacts of the AI boom and real estate slump. First, it quantifies the contribution of AI to the economy from the demand side, including AI-related fixed asset investment, exports, and service income, estimating its pull on GDP. Second, it contrasts the peak contribution of real estate to GDP, fiscal revenue, and household wealth with the current slump to assess its drag on the economy. Third, the report introduces a K-shaped divergence framework to analyze the distribution effects of AI and real estate, tracking resource flows across two dimensions: demographics (income and wealth inequality) and geography (urban divergence). Fourth, it evaluates China's AI development advantages (algorithm efficiency, open-source strategies, talent base, industrial deployment, energy infrastructure) and constraints (U.S. export controls, chip manufacturing gaps) through the lens of U.S.-China AI competition (G2 duopoly). Finally, the report combines Japan's 2002-04 experience in handling bad loans to speculate on Beijing's potential policy response paths, including cleaning up bad debts between financial and non-financial entities and accelerating fiscal reforms to provide local governments with a more solid tax base.

Methodology notes

• Industry/Sector Analysis FrameworkSupply-demand framework

  AI supercycle is a concentrated event, real estate boom is a broad-based wealth creator

  The report uses the supply-demand framework to analyze resource allocation characteristics of different economic cycles: real estate booms benefit provinces and lower-tier cities through rising land and property values, while AI success depends on scarce assets like computing density, data pools, and top-tier talent, which are highly concentrated in a few cities, leading to resource centralization in smart cities rather than broad distribution.

• Cycle and Sentiment FrameworkSentiment Inflection Analysis

  K-shaped divergence analysis framework

  The report uses the K-shaped divergence framework to track structural inequality in economic recovery: the upper arm (capital owners, AI-enhanced elites) captures productivity gains but has low marginal propensity to consume, while the lower arm (graduates, gig workers) has a high marginal propensity to consume but eroded purchasing power, creating persistent deflationary pressure.

• Company Fundamentals and Financial FrameworkThree Statement Articulation

  Balance sheet recession analysis

  The report analyzes how the real estate collapse damages the balance sheets of governments, businesses, and households, leading to intertwined chains of bad debts, loss of trust, slowed credit growth, and weak aggregate demand—a typical balance sheet recession transmission mechanism.

• Macroeconomic frameworkCredit/debt cycle

  Deteriorating terms of trade analysis

  The report notes that China is a net importer of chips, and rising chip prices simultaneously boost imports and exports, but the larger import base means absolute import increments exceed exports, worsening trade conditions and suppressing net exports—a typical external account transmission in credit/debt cycles.

• Industry/Sector Analysis FrameworkUpstream-Midstream-Downstream Transmission

  Model-as-a-Service (MaaS) economic suction effect

  The report analyzes how the MaaS model causes wealth transfer from lower-tier cities to coastal smart cities: when businesses in third- and fourth-tier cities adopt AI agents to optimize operations, they pay recurring subscription fees to coastal platforms, and wealth generated by local industrial activities is continuously siphoned into the treasuries of smart city firms.

• Event Gaming and Behavioral FinanceExpectation Gap/Expectation Management

  AI boom expectation-reality gap

  The report warns markets and policymakers against assuming the new AI economy can cure economic problems caused by real estate, as excessive optimism may lead to insufficient policy responses, necessitating close tracking of the upper and lower arm gaps in K-shaped divergence and preventive measures.

• Valuation methods

  Cross-country comparison of AI capital expenditure as a percentage of GDP

  The report compares China and the U.S. using AI capital expenditure as a percentage of GDP: the U.S. at ~2.5%, China at ~0.8%, with China's scale about one-third of the U.S., a relative indicator that better reflects the true gap in AI investment intensity between the two countries.

• Industry/Sector Analysis Framework

  Price effect vs. quantity effect decomposition

  The report decomposes IC export growth into price and quantity contributions: in April 2026, IC export value grew 99.6% but volume only 3.7%, with price contributing 92.6 percentage points, indicating export growth was mainly driven by price increases rather than capacity expansion, fundamentally different from the 'new three' export cycle in its trade balance impact.

Asset mapping & comparison

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

• Beijing

  Core of AI ecosystem, benefits the most

  Strengths

  Hosts over 2,500 AI firms, leads in registered large models (183), headquarters of top AI providers like Baidu, ByteDance, Zhipu AI, and Moonshot AI, and is an advanced logic chip design center

  Comparison

  One of the four smart cities alongside Shanghai, Shenzhen, and Hangzhou, but leads nationally in registered large models

• Shanghai

  High-end silicon manufacturing and packaging hub, significant beneficiary

  Strengths

  Zhangjiang High-Tech Park is a global hub for AI chip manufacturing and advanced packaging, with SMIC doubling advanced node capacity here; home to domestic GPU leaders like Moore Threads and Biren Technology

  Comparison

  Positioned as China's 'body' (Beijing as the 'brain'), with core status in chip manufacturing

• Hangzhou

  LLM and AI+ hub, beneficiary of MaaS economy

  Strengths

  Headquarters of DeepSeek and Alibaba (Cloud and DAMO Academy), earns AI subscription revenue from provincial SMEs via MaaS economy; humanoid and bionic robots are flagship focuses

  Comparison

  Most mature in MaaS economy model, turning intelligence into recurring utility fees

• Shenzhen

  Chipmaker and embodied AI hub

  Strengths

  Headquarters of Huawei (HiSilicon), dominates edge computing and NPU markets for robotics, drones, and autonomous systems; leads in embodied AI

  Comparison

  Unique advantages in hardware-smart integration, paired with the Greater Bay Area's agile hardware prototyping ecosystem

• Hefei

  Secondary beneficiary city, hardware node

  Strengths

  Leverages 'Hefei Model' state equity stake to play key roles in robotics and memory manufacturing (e.g., ChangXin Memory)

  Weaknesses

  Highest-value IP rents often flow back to Beijing

  Comparison

  More 'industrialized' than 'sovereign,' a vital hardware node but not a core hub

• Wuhan

  Secondary beneficiary city, optics and sensors hub

  Strengths

  Successfully transformed 'Optics Valley' into an AI+ optics and sensors hub, with electronic manufacturing sector growing 62.4% in early 2026

  Comparison

  Occupies specialized node status in autonomous driving and sensor tech

• Suzhou

  Secondary beneficiary city, service robotics leader

  Strengths

  Service robotics output grew 6.0x year-on-year in early 2026, successfully integrating AI into high-end equipment manufacturing base

  Comparison

  Emerging as a leader in service robotics

• Traditional cities (lower-tier cities)

  Losers, reduced to clients in MaaS economy

  Weaknesses

  Face capital flight to coastal areas, local property sector reduced to illiquid assets, accelerated hollowing out of young, educated population

  Comparison

  Stark contrast to smart cities, unable to retain fruits of AI-driven prosperity

  Risks

  Persistent real estate market slump, worsening population outflow

Key data

• AI-related capital expenditure contribution to GDP growth (2026)0.3 percentage pointsDriven through fixed asset investment channels
• 2026 export growth forecast8.6%Up from 4.0%, AI-driven chip upturn contributes 3.7-4.4 percentage points
• Real estate share of GDP at peak (2021)25%Also contributed 38% of fiscal revenue and 60% of household wealth
• Top 100 developer sales decline (2021-25)72.7%By value, still down 19.7% year-on-year in the first four months of 2026
• Migrant workers unemployed due to real estate slump14 millionMainly affecting household wealth in lower-tier cities
• China's AI capital expenditure as a percentage of GDP (2026)0.8%About one-third of the U.S. at 2.5%
• April 2026 IC export value growth99.6%But volume only up 3.7%, price contributed 92.6 percentage points
• Q2 GDP growth forecast4.1%Below market consensus of 4.7%
• Youth unemployment rate (April 2026)16.3%Significantly higher than the 5.2% urban unemployment rate
• Developer total debt (2024)83 trillion RMBDebt owed to upstream and downstream firms about 25 trillion

Impact & implications

The report argues that the AI boom alone is unlikely to resolve the economic woes caused by the real estate slump. The impact on consumption is structural: the upper arm (capital owners, elites) captures productivity gains but has a low marginal propensity to consume, while the lower arm (graduates, gig workers) has a high marginal propensity to consume but eroded purchasing power, delaying major life milestones like marriage and children due to financial insecurity, creating persistent deflationary pressure. The impact on real estate is geographic: the AI supercycle as a value concentrator drives sharp divergence in asset values and fiscal health along geographic lines, with only a few smart cities (especially the four first-tier cities) potentially seeing genuine real estate market recovery, while most cities, unable to retain the fruits of AI-driven prosperity, will face suppressed aggregate demand in the medium to long term. The policy implications are urgent: Beijing may need to significantly strengthen measures to clean up the real estate sector's mess, including more decisively clearing bad loans between financial and non-financial entities, drawing from Japan's 2002-04 experience; accelerating fiscal reforms to provide local governments with a more solid tax base, as land sales revenue losses exceed 80%; and potentially introducing policies to support more local governments' 'sovereign AI' to help lower-tier cities share in AI prosperity benefits, while improving social safety nets to mitigate the negative impacts on unemployed and underemployed workers.

Risks

• AI capital expenditure slows earlier than expected or chip prices normalize in H2 2026
• Youth unemployment worsens further, already at 16.3% in April 2026
• Persistent weak consumption, K-shaped divergence weakening aggregate demand
• Real estate recovery limited to a few first-tier cities, most cities remain sluggish
• Trade surplus narrows moderately from 2025 record highs
• Social safety nets insufficient to handle unemployment waves from AI displacement and real estate slump

What to watch

• Whether Beijing will introduce policies to support more local governments' 'sovereign AI' to help lower-tier cities share in AI prosperity benefits
• Social safety net improvements to mitigate impacts on unemployed and underemployed workers
• Policy intensity in cleaning up bad loans in the real estate sector
• Progress in fiscal reforms to provide local governments with a more solid tax base
• Whether the upper-lower arm gap in K-shaped divergence widens
• Whether adoption pace of technologies like autonomous driving that may displace blue-collar jobs slows