The domestic economy has displayed pronounced K-shaped divergence this year, characterized by robust external demand contrasting with weak internal consumption, and a thriving technology sector set against a sluggish traditional industrial base. The critical question emerges: can the AI-driven "silicon-based" economy revitalize its "carbon-based" counterpart? Following recent policy meetings emphasizing economic stability and a strong start to the 15th Five-Year Plan, the urgent need to stimulate domestic demand has become paramount, especially as external demand may soften in the latter half of the year.

Economic Landscape: Pronounced Divergence and Fragile Domestic Demand

The K-shaped divergence has become the defining characteristic of China's economic trajectory this year. Production data reveals a widening gap, with industrial output above designated size growing 5.3% in the first seven months, yet high-tech manufacturing surged 13.8%. Notably, midstream technology and high-end manufacturing significantly outperformed upstream and downstream sectors, with output of 3D printing equipment, lithium-ion batteries, and industrial robots growing 52.3%, 40.2%, and 28.5% year-on-year respectively. Concurrently, the real estate and construction sectors continue their contraction, suppressing related industries such as building materials and home furnishings.

On the demand side, external markets are booming while domestic consumption falters. Domestic demand's contribution to economic growth sits at historic lows, with consumption and investment contributing merely 2.2 and 1.7 percentage points respectively. Exports, however, stand out as the sole bright spot, driving GDP growth by 2.6 percentage points in the first half. Export growth of 18.5% in the first seven months stems primarily from computing power demand, with electronic products like integrated circuits and automatic data processing equipment contributing 10.8 percentage points. This reflects the "silicon-based prosperity" of the global AI race, as US cloud giants allocate an estimated $650-700 billion in capital expenditure this year, predominantly toward data centers, creating supply shortages in memory and optical modules.

The weakness in domestic demand deserves serious attention. The simultaneous contraction of both consumption and investment is unprecedented since the reform and opening-up period. Retail sales growth turned negative at one point, with social retail sales growing only 1.2% in the first seven months; adjusting for 0.9% CPI inflation, real growth approaches zero. May even recorded -0.6% growth, marking the first simultaneous contraction of consumption and investment excluding pandemic periods. This consumption slump fundamentally reflects households' active deleveraging driven by pessimistic income expectations, evidenced by a record cumulative decline of 827.1 billion yuan in household loans this year.

Investment weakness continues to deepen. After fixed asset investment's first annual decline (-3.8%) last year, this year's policy targets "stabilization and recovery." Yet the first seven months show the decline widening to -6.7%. While real estate investment remains the primary drag, infrastructure and manufacturing investment also present concerns. Sector-wise, only a few industries including transportation equipment, computer communications, and electrical machinery maintain positive growth, while traditional sectors generally contract, highlighting the prosperity gap between "silicon-based" and "carbon-based" economies.

As a super-large economy, China's high-quality development must ultimately rest on domestic demand. International comparison reveals China's household consumption rate is notably low at 40.0% in 2025, significantly below not only developed nations like the US (68.1%) and Germany (51.6%), but also developing countries such as Vietnam (53.2%) and India (56.7%). Historical comparison shows similar disparity: when the US, Japan, and Germany had comparable per capita GDP around $14,000, their consumption rates typically ranged between 50-60%, substantially higher than China's current level.

Medium-Term Outlook: Technology-Driven Transformation and Supply Creating Demand

The current "silicon-based" revolution is experiencing the "creative destruction" phase at the front of the "J-curve," where substitution effects precede incremental gains. AI partly displaces existing jobs while diffusion frictions may temporarily slow productivity growth. Only after this adjustment completes, with new business models, scenarios, and employment emerging, will technological dividends fully materialize. Since ChatGPT's release in November 2022, labor markets have faced significant disruption. US effective recruitment information indices declined 30.4%, with high-AI-exposure sectors like data analytics particularly affected. Non-farm payroll data shows slower employment growth in AI-intensive industries such as information, finance, insurance, and manufacturing compared to low-penetration sectors like accommodation, food services, construction, and retail.

The J-curve inflection point occurs when AI transitions from the virtual to the physical world. Currently, AI remains primarily in model iteration, with investment concentrated in computing and algorithms. However, as models mature toward multimodality, AI will inevitably enter physical applications. Changes are already emerging: since ChatGPT's launch, AI applications have expanded from information-intensive sectors like technology, media, and finance to physical industries. US manufacturing AI usage exceeds 60%, with even lower-exposure sectors like accommodation and food services reaching around 32%. NVIDIA projects AI-transformed physical industries including robotics, autonomous driving, factories, and edge computing will reach approximately $50 trillion by 2030.

Smart vehicles and humanoid robots represent the most critical manifestations. Autonomous driving is transitioning from premium options to mass adoption, with three key milestones anticipated: L4 autonomy becoming reality by 2028, scaled applications following the surge in vehicle computing power during 2029-2030, and by 2035, vehicles potentially evolving into "third intelligent spaces" beyond homes and offices. Humanoid robots show equally immense potential, with Bank of America projecting shipments reaching 1.2 million units by 2030, twelve times current levels, and surpassing 100 times by 2035.

China holds significant advantages for the "second half" marked by physical AI, potentially completing its old-new growth engine transition around 2029. Chinese engineer compensation runs approximately one-quarter to one-fifth of US levels, token prices are one-fifth to one-tenth of US costs, combined with a complete industrial system and scaled applications. Smart vehicles serve as natural carriers of physical AI, offering substantial scale and extensive supply chains; evolving from transportation tools into "mobile intelligent spaces" could activate massive replacement demand. Humanoid robots, while still iterating technologically, present tremendous market potential given diverse application scenarios, potentially becoming flagship next-generation products of physical AI.

China's manufacturing may experience its "2003 moment" anew. After 2003, Chinese construction machinery and home appliances went global, driving mid-range manufacturing growth. This round of smart manufacturing could generate fresh export dividends, improving corporate ROE and expanding investment and consumption. Currently, China's auto industry revenue accounts for approximately 8% of GDP, 3-5 percentage points below traditional powers like Germany and Japan. If Chinese auto industry share expands to comparable levels, the new economy would transform quantitatively and qualitatively, surpassing the old economy. Based on US industrial data trends, if manufacturing AI penetration continues its current trajectory, near-comprehensive coverage could approach by 2029; given China's manufacturing scale and efficiency advantages, 2029 represents a conservative timeline for China's growth engine transition.

Short-Term Policy Priorities: Supporting Domestic Demand Through Coordinated Consumption and Investment

The immediate challenge involves potential external demand declines requiring domestic demand to fill the gap. Export growth may decelerate in the second half, with three inflection points emerging in overseas capital expenditure: first, major tech companies' free cash flow turning negative as AI arms race spending intensifies; Google's operating cash flow already turned negative at -$5.86 billion in the first half. Without clear monetization pathways, convincing capital markets for continued investment becomes difficult. Second, debt financing costs face upward pressure; as cash flows tighten, companies must seek external financing amid persistently high interest rates, with 30-year US Treasury yields recently hitting 20-year highs. Under high leverage and interest rate conditions, markets are beginning to price default risks, with CDS spreads rising rapidly. Third, computing power demand growth may peak; demand concentrating at the high end could create oversupply in mid-to-low-end computing, as evidenced by H100 rental prices plateauing after June, while global manufacturing PMI peaks suggest moderating "silicon-based" enthusiasm.

As a super-large economy, China cannot rely solely on external demand. Should overseas tech giants shift capital expenditure, export dividends would rapidly diminish; policies to support domestic demand should be implemented immediately. Stabilization policies should center on "six networks" infrastructure and "investing in people." Macro policy must address immediate concerns while planning long-term development, preventing the spiral of declining investment and weak consumption during transition while directing resources toward key sectors and industries for improved capital efficiency, avoiding outdated approaches.

On the investment side, increased investment in the "six networks," particularly domestic computing power networks, is recommended. As the top-level infrastructure framework for the 15th Five-Year Plan period, the "six networks" project total investment exceeds 25 trillion yuan, with approximately 7 trillion planned this year. Domestic computing power networks are particularly critical. Currently, domestic GPUs and other key components face higher unit computing costs due to manufacturing process constraints, limiting international competitiveness. Meanwhile, Western semiconductor export controls continue tightening, making chips a critical bottleneck. Building domestic computing networks ahead of demand would revitalize China's AI ecosystem: supporting domestic GPUs, storage, and other core hardware addresses "bottleneck" deficiencies while providing substantial high-quality domestic computing capacity would significantly reduce development costs for Chinese model and application companies. Increased fiscal support and faster implementation for domestic computing networks are recommended.

China's related investment lags significantly behind the US in scale. US four major cloud companies' 2025 capital expenditure approximates $410 billion, growing over 65%; China's seven major companies (Tencent, Alibaba, Baidu, ByteDance, and three telecom operators) have comparable investments of 400-500 billion yuan, only one-sixth of US levels. The National Development and Reform Commission indicates new direct investment in computing networks during the 15th Five-Year Plan period could reach 4 trillion yuan. With enterprises as primary investors, fiscal funds can leverage substantial social capital with high multipliers; further enhancement is recommended.

On the consumption side, increased "investing in people" support across childcare, elderly care, and service consumption is advised. Strengthened childcare subsidies would release household consumption potential. Despite continuous fertility policy liberalization, China's birth rate remains low, with childcare burden (childcare cost per capita GDP) second only to South Korea among major economies, squeezing current consumption while raising precautionary savings, constraining aggregate demand. International experience demonstrates childcare subsidies effectively stabilize fertility rates; Nordic economies with higher subsidy levels generally show smaller birth rate declines. China's current annual childcare subsidies of approximately 100 billion yuan represent merely 0.07% of GDP, leaving substantial room for policy expansion. Additionally, stabilized fertility expectations would release demand for maternal and infant products, children's education, and housing.

Improved elderly-oriented supply would unlock silver economy consumption potential. The elderly demographic possesses both financial resources and leisure time: households headed by those aged 56-64 hold average assets of 3.554 million yuan, double that of 18-25 age group households. Extended life expectancy further lengthens retirement consumption cycles, with significant potential in cultural, entertainment, and wellness spending such as performances and shows. Enhanced investment in accessible public infrastructure and commercial tourism facilities, along with subsidies for age-friendly tourism products and services, would convert potential demand into actual consumption.

Expanding government subsidies toward service consumption is essential. Over the past three years, consumer goods trade-in programs totaling 700 billion yuan primarily covered durable goods including automobiles, home appliances, and mobile phones, with increasingly evident "overhang" effects. Service consumption offers high frequency, substantial headroom, sustained domestic demand stimulation, and employment expansion. China's resident service consumption expenditure share of approximately 46.1% in 2025 remains well below developed economies' roughly 60% level. Except during the pandemic period, Chinese service consumption growth has consistently outpaced goods consumption, demonstrating substantial potential. Recommendations include expanding the undelivered fourth batch of 62.5 billion yuan in government subsidies to cover service consumption with appropriately increased subsidy intensity.

Risk Warning: Economic transformation progress may fall short of expectations.