1 China stuck trying to play catch-up on the AI front despite new DeepSeek model
DeepSeek releases new AI model
On April 24, the Chinese artificial intelligence company DeepSeek officially released its latest flagship large model, DeepSeek V4. DeepSeek V4 introduced three variants: Flash, Pro, and the high-performance Pro Max preview. Its core technical specifications and market positioning are as follows:
- According to official data, DeepSeek V4 Pro Max achieves performance in standard reasoning benchmarks that approaches — or in certain domains surpasses — some frontier models. However, it still lags behind OpenAI’s GPT-5.5, Google’s Gemini 3.1 Pro, and Antrophic’s Claude Opus.
- The V4 model expands its context window from 128,000 tokens to 1 million tokens. It also emphasizes enhanced “agentic capabilities,” enabling the execution of more complex automated workflows.
- DeepSeek V4 is deeply optimized for Huawei’s Ascend chips, significantly reducing reliance on high-end NVIDIA chips. Observers widely interpreted this optimization as a strategic response to export restrictions.
***
The launch of DeepSeek V4 has reignited controversy surrounding model distillation and the broader trajectory of Sino-U.S. technological decoupling.
In early 2025, DeepSeek’s initial release of the R1 model sent shockwaves around the globe with its performance and cost. The company claimed a training cost of approximately $5.6 million, while achieving reasoning and coding performance comparable to OpenAI’s o1 series flagship models. This level of computational efficiency challenged Silicon Valley’s prevailing paradigm of scaling through massive hardware investment, contributing to significant volatility in NVIDIA’s stock price and prompting a reassessment of AI development economics.
As DeepSeek’s rapid rise continued, Western technology firms and the U.S. government raised serious concerns regarding the origins of its capabilities, focusing on the issue of model distillation:
- Model distillation is a technique in which outputs generated by a more powerful “teacher model” (such as GPT-4 or Claude 3) are used to train a smaller, more efficient “student model.” While common in the industry, large-scale unauthorized distillation is often viewed as a form of intellectual property extraction.
- In February 2026, Anthropic reported that DeepSeek and other Chinese laboratories had systematically extracted reasoning data from its Claude models using tens of thousands of fraudulent accounts at an “industrial scale.”
- On the day of DeepSeek’s V4 release, the Trump administration issued a diplomatic cable via the State Department warning that Chinese AI firms were using illicit “distillation” techniques to replicate the capabilities of advanced U.S. systems. U.S. scientific advisers further noted that such models may lack the full safety architecture and contextual robustness of their original counterparts despite strong benchmark performance.
State media makes excuses for DeepSeek V4’s delayed release
In the period around the official release of DeepSeek V4, the state-affiliated Chinese media account Yuyuan Tantian (玉淵譚天) published multiple commentaries that reframed repeated delays to the models release into a narrative of “technological self-reliance and strategic breakthrough under external pressure.”
In these commentaries, Yuyuan Tantian characterized the launch of DeepSeek V4 as a “silent counteroffensive by Chinese algorithms against Silicon Valley dominance.” The underlying narrative rests on three principal themes:
- “Algorithmic sovereignty”: Emphasis is placed on the claim that DeepSeek has reduced dependence on U.S.-based H100/B200 chip clusters, instead achieving efficiency gains through domestic architectural optimization. Yuyuan Tantian portrayed this as a case of DeepSeek achieving disproportionate results with limited resources.
- “Breaking the American narrative”: The chip export restrictions imposed by the Trump administration on China are framed not as a constraint, but as a catalyst that forced the emergence of a model optimized for domestic computing infrastructure.
- Moral positioning: In contrast to the perceived “closed” and “high-cost” approaches of Western firms such as OpenAI, Yuyuan Tantian highlights DeepSeek’s adoption of open-weight distribution as an embodiment of a “shared future for humanity.”
DeepSeek V4 had originally been scheduled for release around the 2026 Chinese New Year period in February, but was delayed by nearly two months due to hardware adaptation challenges and computing constraints. In response to external concerns regarding the stability of training on Huawei’s Ascend 910B processors, Yuyuan Tantian adopted several framing strategies:
- Narrative of strategic patience: Rather than directly addressing reports of hardware failures or cluster instability, Yuyuan Tantian stressed the importance of long-term technological discipline (“Chinese technology requires the resolve of ‘polishing a sword for ten years’”), arguing that the delay reflected a transition from reliance on the CUDA ecosystem to a fully autonomous domestic stack.
- Reframing as strategic choice: The postponement was presented as a deliberate decision to align with national supercomputing infrastructure and to conduct stress testing under conditions of intensified trade tensions (“anti-interference” and “extreme computing environment”).
- Downplaying hardware limitations: Reports suggesting that Huawei chips operate at about 91 percent of the efficiency of NVIDIA hardware were countered with claims that system-level architectural optimization (“Chinese-style architecture”) compensates for any single-chip performance gap.
Our take
Behind the touted “algorithm-driven breakthrough” represented by DeepSeek’s V4 series lies a set of deeper structural challenges related to technology sourcing, hardware performance, and capital allocation. China’s AI sector appears to currently be pursuing a form of constrained optimization under resource limitations. As the U.S. continues to rely on large-scale hardware accumulation and security-based restrictions, a structural gap in AI development trajectories between the two countries may be emerging and could potentially last for decades.
1. DeepSeek previously attracted global attention by challenging the conventional “compute-scaling” paradigm dominant in Silicon Valley with its extremely low model training costs. Western technology firms and policymakers, however, have argued that DeepSeek could have relied on forms of intellectual property extraction to achieve its “efficiency.” At the same time, DeepSeek’s hardware constraints impose a ceiling on the achievable performance of these models.
One of the core mechanisms currently sustaining China’s AI competitiveness is model distillation. According to a 2026 investigation by Anthropic, Chinese laboratories conducted over 16 million distillation queries using more than 24,000 accounts to extract outputs from the Claude system. But according to the State Department’s recent global advisory, while these distilled models perform excellently in benchmarks, they are essentially “second-class products.”
Although DeepSeek’s latest V4-Pro introduces a “hybrid attention” mechanism that compresses KV cache usage by approximately 90 percent, this approach can also be interpreted as a response to hardware constraints rather than a purely scientific breakthrough. In the absence of high-bandwidth memory and advanced semiconductor nodes, such compression techniques enable model deployment on domestic hardware platforms, including Huawei’s Ascend 950. This divergence reflects a broader structural contrast between Chinese and U.S. models; the U.S. emphasizes higher computational ceilings and original model capabilities, while China focuses on targeted optimization under constrained hardware conditions.
2. Chinese state-affiliated media, including Yuyuan Tantian, have framed the delayed release of DeepSeek V4 as “strategic resilience and autonomous breakthrough.” The narrative, however, obscures underlying development challenges.
Originally scheduled for release in February 2026, DeepSeek V4 was delayed by nearly two months. The delay is widely attributed to the complexity of migrating trillion-parameter models from the CUDA ecosystem to Huawei’s CANN framework, as well as insufficient training stability in large-scale domestic chip clusters. Yuyuan Tantian characterized this delay as a “deliberate strategic choice” by DeepSeek in transitioning toward a fully self-reliant technological stack. In practical terms, however, high communication overhead in large-scale domestic clusters likely reduced convergence efficiency, prolonging development timelines.
Despite claims that “Chinese architectural innovation” offsets single-chip performance gaps with NVIDIA hardware, cost indicators suggest that the V4-Pro model is significantly more expensive than its predecessor. If domestic alternatives were as efficient and cost-effective as presented, pricing for API access would be expected to decline rather than remain elevated. Current pricing dynamics indicate that achieving performance comparable to models such as OpenAI’s GPT-5 still depends heavily on scarce high-end compute resources, including inventory-constrained chips such as NVIDIA’s H20.
3. The dimension of the U.S.–China AI competition at present has shifted from a narrow focus on “chip performance” to the broader foundation of “energy infrastructure.” At this level, China demonstrates notable structural stability and path independence, while the U.S. is undergoing substantial reconfiguration under an “energy-first” strategy.
i) China’s core advantage lies in its capacity to execute large-scale, state-led system engineering. By the end of 2025, China had completed 46 ultra-high-voltage (UHV) transmission projects, with total line length exceeding 62,000 kilometers. This enables data centers to be co-located with low-cost renewable energy bases in western regions (such as Inner Mongolia, where green power accounts for 84.57 percent), thereby mitigating the spatial mismatch and grid-connection bottlenecks commonly observed in other economies.
AI infrastructure projects in China typically progress from planning to grid connection within months, whereas comparable timelines in the U.S. or Germany often extend to 2–7 years due to regulatory and infrastructure constraints.
In 2025, China added nearly 550 GW of new power generation capacity, more than ten times the 53 GW added by the United States. Both Elon Musk and NVIDIA’s Jensen Huang have warned that this “power supply advantage” could become a decisive factor enabling China to offset gaps in semiconductor capability.
2) The U.S. power grid faces significant challenges, including aging infrastructure and fragmented governance. Following Donald Trump’s return to the White House in 2025, Washington has advanced an “energy-first” strategy aimed at deregulation to sustain technological leadership in AI.
However, approximately 70 percent of U.S. grid infrastructure is approaching the end of its operational lifespan and the country lacks a national high-voltage transmission network. As a result, major technology firms — including Microsoft and OpenAI — have been compelled to invest heavily in dedicated power generation facilities to meet the gigawatt-scale energy demands of million-GPU clusters.
3) Nuclear energy is increasingly emerging as a decisive factor in the AI competition, with China currently holding a leading position.
China has achieved commercialization in small modular reactors (SMRs), such as Linglong One, and in fourth-generation nuclear technologies, including the HTR-PM. These are being integrated into “nuclear + computing” industrial parks designed to directly power large-scale data centers.
In contrast, the U.S. has relied in part on restarting previously decommissioned plants (such as Three Mile Island Unit 1) to address near-term energy gaps. While the “energy-first” strategy seeks to accelerate SMR deployment, first grid connections are not expected until after 2030.
4. The CCP’s pursuit of technological self-sufficiency amid structural bottlenecks has seen it impose strict regulatory controls over the AI sector. Such moves, however, are likely to be detrimental to retaining talent and capital in the long run.
i) On April 27, Beijing blocked Meta Platforms’ $2 billion acquisition of agentic AI start-up Manus. While Manus is incorporated in Singapore, its co-founders Xiao Hong and Ji Yichao are PRC citizens; the Financial Times reported in March that both have been placed under exit restrictions in China.
Beijing’s action signals clearly to Chinese AI entrepreneurs that even if headquarters are relocated abroad (e.g., to Singapore) and capital structures internationalized, any technology or talent originating from China remains subject to extraterritorial political control. Such uncertainty will discourage Chinese founders from pursuing a “develop domestically, expand globally” strategy. Going forward, top-tier Chinese talent could increasingly opt to set up shop in fully decoupled, native R&D ecosystems in Silicon Valley or Singapore. This in turn could lead to sustained talent outflows from China’s AI sector.
ii) Chinese regulators have reportedly instructed firms such as Moonshot AI and StepFun to reject U.S. capital, while requiring ByteDance, the parent company of TikTok, to restrict secondary share transfers to American investors. While framed as a reciprocal response to U.S. investment restrictions, these measures effectively confine Chinese AI unicorns within a domestically insulated capital ecosystem that are reliant on funding from large local platforms such as Alibaba and Tencent. The absence of global dollar-based capital participation risks creating a valuation disconnect between market pricing and underlying technological capability.
5. DeepSeek V4 has narrowed performance gaps in benchmark testing through algorithmic efficiency and extreme optimization and could see increased global adoption. However, structural divergence between the U.S. and China in AI development is likely to widen over the long term.
First, China’s AI ecosystem remains heavily dependent on model distillation from leading U.S. systems. Should the U.S. impose a comprehensive shutdown of API access, Chinese developers would effectively lose their “teacher models,” impairing iteration speed and widening the technological gap.
Second, the PRC possesses systemic advantages in power generation and grid infrastructure, enabling lower marginal inference costs. But in terms of “frontier intelligence” and physical compute efficiency, the U.S. — leveraging nuclear energy deployment and hyperscale clusters with millions of GPUs — is establishing a performance ceiling that cannot be bridged through engineering optimization alone.
Third, investors and researchers who have lived through the “unfinished projects” of the Xiong’an New Area and the semiconductor “Big Fund” have seen their trust in state-led “grand narratives” weakened significantly. The endogenous flight of talent and capital from the mainland now constitutes a more severe challenge to the CCP regime than the external technology embargoes.
Systemic differences in the U.S. and China are bifurcating the global AI revolution. One system (the U.S.) concentrates the most advanced and capital-intensive “cognitive infrastructure,” while the other (the PRC) emphasizes scalable, cost-efficient “computational deployment.” DeepSeek V4 illustrates the upper bounds of engineering optimization under resource constraints, yet also underscores its limitations when confronted with structural asymmetries in capital, energy, and foundational innovation. Over the next five years, China risks becoming a technologically constrained ecosystem — locked in by both physical constraints and its political system, with a technological gap permanently stabilized at 18 to 24 months behind the global frontier.
2 April Politburo meeting reveals contradictory orientation of Beijing’s economic policy
April Politburo meeting
On April 28, the CCP Politburo held a meeting to analyze the PRC’s current economic situation and set policy priorities. The meeting repeated recent propaganda rhetoric about the economy, including how China’s economy has made a “strong start” under the “centralized and unified leadership” of Party Central with Comrade Xi Jinping at the core, with key indicators “exceeding expectations” as well as demonstrating “robust resilience and vitality.” At the same time, the meeting acknowledged that the PRC was facing “certain difficulties and challenges.”
The meeting also issued several policies and measures:
Overall policy framework for economic work
- Maintain the general principle of “seeking progress while maintaining stability” (穩中求進).
- Better coordinate domestic and international priorities, as well as development and security.
- Advance technological self-reliance and strengthen independent control over industrial and supply chains.
- Implement a more proactive fiscal policy and a moderately accommodative monetary policy.
- Continue expanding domestic demand and optimizing supply structures.
- Stabilize employment, enterprises, markets, and expectations.
Fiscal and monetary policy stance
- Fully utilize macroeconomic policy tools.
- Ensure the “three guarantees” (basic livelihoods, wages, and government operations) at the grassroots level.
- Enhance the forward-looking, flexible, and targeted nature of monetary policy while maintaining ample liquidity.
- Keep the renminbi exchange rate broadly stable at a reasonable and balanced level.
- Strengthen consistency assessments across macro policy orientations.
Measures to expand domestic demand
- Promote consumption upgrading.
- Deepen initiatives to expand capacity and improve quality in the services sector.
- Accelerate planning and construction of infrastructure, including water networks, next-generation power grids, computing networks, advanced communications networks, urban underground utilities, and logistics systems.
- Advance the commencement of major projects where conditions are mature.
Industrial policy priorities
- Accelerate the development of a modern industrial system and maintain a reasonable share of manufacturing.
- Deepen the construction of a unified national market and address “involution-style” competition.
- Fully implement the “AI+” initiative to foster new forms of the intelligent economy.
- Further deepen reforms of state-owned enterprises.
- Systematically respond to external shocks and enhance energy and resource security.
Key risk areas identified
- Effectively prevent and mitigate risks in critical sectors.
- Stabilize the real estate market and steadily advance urban renewal.
- Orderly resolve local government debt risks and address arrears owed to enterprises.
- Promote reform of small and medium-sized financial institutions.
Social and livelihood priorities
- Strengthen the employment-first policy orientation.
- Ensure agricultural production and stabilize prices of key products such as pork.
- Improve regularized assistance mechanisms to prevent large-scale relapse into poverty.
- Enhance safety measures in production, disaster prevention and mitigation, and food and drug safety.
China’s Q1 fiscal data
On April 24, the PRC Ministry of Finance released China’s first-quarter fiscal data.
National general public budget
- Total revenue: Up 2.4 percent year-on-year to 6.1613 trillion yuan.
- Tax revenue: Up 2.2 percent year-on-year to 4.8505 trillion yuan.
- Non-tax revenue: Up 2.9 percent year-on-year to 1.3108 trillion yuan.
- Major tax categories:
- Domestic VAT: Up 4.9 percent to 2.1473 trillion yuan.
- Domestic consumption tax: Down 4.5 percent to 499.0 billion yuan.
- Corporate income tax: Down 5.6 percent to 1.0369 trillion yuan.
- Personal income tax: Down 10.5 percent to 501.8 billion yuan.
- Import VAT and consumption tax: Up 12.9 percent to 458.8 billion yuan.
- Tariffs: Up 14.1 percent to 55.1 billion yuan.
- Export VAT and consumption tax rebates: Up 4.8 percent to 791.1 billion yuan.
- Securities transaction stamp duty: Up 78.1 percent to 73.3 billion yuan.
- Total expenditure: Up 2.6 percent to 7.4706 trillion yuan.
- Major expenditure categories:
- Science and technology: Down 3.7 percent to 192.7 billion yuan.
- Social security and employment: Up 8.6 percent to 1.4785 trillion yuan.
- Healthcare: Up 17.3 percent to 655.4 billion yuan.
- Interest payments on debt: Up 12.9 percent to 322.3 billion yuan.
National government fund budget
- Revenue: Down 16.2 percent to 775.1 billion yuan.
- State-owned land-use rights transfer revenue: Down 24.4 percent to 517.6 billion yuan.
- Expenditure: Up 3.1 percent to 2.0387 trillion yuan.
- Expenditure related to land-use rights transfers: Down 12.1 percent to 956.8 billion yuan.
China’s Q1 industrial enterprise profits
On April 27, the PRC National Bureau of Statistics released the profit data for industrial enterprises above designated size.
- Total profits (January–March): Up 15.5 percent year-on-year to 1.69604 trillion yuan (calculated using a comparable caliber).
- By ownership:
- State-controlled enterprises: Up 10.1 percent to 619.61 billion yuan.
- Foreign-invested enterprises (including Hong Kong, Macao, and Taiwan): Up 1.2 percent to 383.73 billion yuan.
- Private enterprises: Up 25.4 percent to 430.53 billion yuan.
Our take
The April Politburo meeting displays the CCP’s current decision dilemma and structural contradictions. Beijing wants to promote “new quality productive forces” and high-tech industrial investment, yet also wants to simultaneously crack down on “involution-style” competition triggered by overinvestment. Beijing also wants to maintain the intensity of fiscal spending to support infrastructure investment, yet must contend with a sharp decline in revenues caused by the erosion of land-based financing. Finally, Beijing aims to contain systemic debt risks, yet is reliant on large-scale sovereign bond issuance to fill local fiscal gaps.
This policy orientation reflects the CCP leadership’s attempt to maintain economic growth while being forced to address market distortions rooted in institutional constraints. The current economic landscape thus exhibits a duality characterized by “apparent prosperity in headline data” coexisting with “deep contraction at the micro level.”
1. The most striking contradiction among China’s macroeconomic indicators for the first quarter of 2026 is the sharp rebound in industrial profits alongside only marginal growth in tax revenues. This divergence is not accidental, but rather, the combined result of current industrial policy priorities, the structure of the tax system, and micro-level operating conditions.
i) Industrial enterprises above designated size recorded a year-on-year profit increase of 15.5 percent in Q1 2026. While the aggregate figure is substantial, sectoral divergence has reached an extreme “K-shaped” pattern.
This “K-shaped” divergence indicates that profit growth is heavily concentrated in policy-supported “hard technology” sectors and select upstream industries. For example, profits in computer, communications, and electronic equipment manufacturing rose by 120 percent year-on-year, largely driven by global AI-related demand for servers and storage components. By comparison, core industrial sectors such as automobiles, specialized equipment, and construction materials — the latter of which is closely tied to real estate — remain in contraction. The uneven distribution of profits suggests that the benefits of economic recovery are increasingly narrow, with most traditional enterprises not experiencing a meaningful rebound.
ii) In Q1 2026, industrial enterprises-generated operating revenue increased 5.0 percent to 33.19 trillion yuan, compared with a 15.5 percent increase in profits. Profit growth was thus three times faster than revenue growth, pushing the profit margin to 5.11 percent. Under normal economic conditions, such a pattern would imply improved operational efficiency. But in the current deflationary environment, this development more likely reflects aggressive cost compression.
Beijing’s official data show that costs per 100 yuan of revenue declined to 84.93 yuan, a reduction of 0.40 yuan from the previous year. Amid overcapacity and frequent price competition, this cost reduction is less attributable to technological gains and more to labor cost compression, extended payment cycles to suppliers, and softer input prices. As of end-March, the average accounts receivable collection period reached 72.6 days, indicating that much of the reported profit remains “on paper” while actual cash flow conditions are strained. By comparison, the same metric stood at 67.3 days in Q1 2024 and 55.5 days in Q1 2019.
A portion of these receivables reflects delayed payments by local governments. This is partly confirmed by the April Political meeting, which again emphasized resolving arrears owed to enterprises. According to mainland media reports, local governments had issued 182.6 billion yuan in new special-purpose bonds as of April 28, part of which is allocated to clearing such payment backlogs.
iii) China’s total tax revenue grew by only 2.2 percent in Q1 2026 despite a 15.5 percent increase in industrial profits. This non-linear relationship highlights structural fragility within the fiscal system.
First, Beijing has implemented extensive tax incentives for high-tech manufacturing to promote “new quality productive forces,” including enhanced R&D expense deductions and VAT refund credits. Since the fastest-growing profit sectors (e.g., AI and semiconductors) are also the primary beneficiaries of these incentives, profit growth has not translated proportionally into tax revenue.
Second, firms are currently permitted under tax law to offset current profits against accumulated losses following widespread industrial losses in previous years. As a result, even returning to profitability does not immediately generate corporate income tax liabilities.
Third, value-added tax is levied on turnover rather than profits. With operating revenue rising by only 5.0 percent and domestic VAT increasing by 4.9 percent, the close alignment indicates that the tax base remains dependent on scale expansion rather than profitability. This suggests that improvements in economic “quality” have yet to be reflected in fiscal contributions.
2. If industrial profit data offer even a limited source of reassurance for Beijing, the deterioration in the government fund budget directly undermines the operational foundation of local governments. Indeed, the prolonged downturn in the real estate sector is triggering a profound restructuring of fiscal and investment dynamics at the local level.
i) In Q1 2026, revenue from the transfer of state-owned land-use rights by local governments totaled 517.6 billion yuan, representing a year-on-year decline of 24.4 percent. This contraction significantly exceeds the full-year decline in the previous year (negative 14.7 percent), indicating that the land market’s stabilizing mechanism is breaking down.
The sharp drop in land-related revenues has directly constrained government fund expenditures. Although total government fund budget expenditure rose by 3.1 percent in Q1 2026, spending linked to land transfers declined by 12.1 percent. This suggests that local governments are no longer able to rely on land development to drive urban auxiliary investment, marking the effective collapse of the traditional “land finance-driven growth” model.
ii) While revenues are contracting, debt accumulated over the past decade has entered a peak interest repayment phase. Interest payments within the general public budget reached 322.3 billion yuan in Q1 2026, up 12.9 percent year-on-year. This stands in stark contrast to declines in science and technology spending (down 3.7 percent) and agriculture, forestry, and water-related expenditures (down 6.8 percent).
This rigid expenditure structure implies that a significant portion of every 100 yuan in fiscal revenue must be allocated to debt servicing rather than public services or social welfare. Fiscal policy space is being severely constrained by legacy debt burdens, which explains why Beijing has been compelled to maintain a deficit ratio of around 4 percent and issue ultra-long-term special government bonds.
iii) Faced with the vacuum left by the collapse of land financing, the April Politburo meeting emphasized strengthening the construction of “six major networks” (including power, communications, and transportation infrastructure). The National Development and Reform Commission estimates that these projects could generate over 7 trillion yuan in total investment in 2026.
However, this strategy of attempting to use state-led infrastructure investment to offset the real estate downturn carries two major risks:
- A shift from reliance on “land transfer revenues” to dependence on “central government debt,” effectively transferring debt from local to central authorities rather than resolving it.
- Traditional infrastructure sectors are approaching saturation, while emerging “six major networks” projects are unlikely in the near term to generate sufficient cash flow to cover financing costs. This could potentially increase fiscal burdens in the long run.
3. The April Politburo meeting’s call to “rectify involution-style competition” is not new. This issue was first explicitly raised at the July 30, 2024 Political Bureau meeting. That Beijing had to repeat the call two years later underscores the limited effectiveness of policy measures in addressing structural overcapacity.
China’s “involution” is not purely market-driven competition, but rather the result of distorted resource allocation caused by government intervention. Despite repeated central directives, local governments continue to provide subsidized land and targeted funding to emerging industries in order to meet GDP growth and investment targets, artificially lowering operating costs for firms. Under the policy emphasis on “new quality productive forces,” subsidy-driven expansion has led to blind capacity growth. Subsidy competition has also resulted in disorderly capacity expansion and severe oversupply in sectors such as photovoltaics.
Beijing has oscillated between promoting innovation through subsidies and attempting to curb overcapacity from 2024 to 2026. On one hand, it calls for rectifying involution; on the other, it injects capital into selected high-tech sectors through instruments such as ultra-long-term special bonds. This sustained external “support” allows inefficient capacity that would otherwise exit the market to persist. Current policy tools remain primarily administrative (such as entry controls and forced consolidation) rather than restoring market-based competitive selection. As long as local fiscal incentive structures remain unchanged, the root causes of destructive competition will persist.
The export of low-priced goods generated by this “involution” has become a central source of global trade tensions. In Q1 2026, the United States, Europe, Japan, and India initiated anti-dumping investigations into Chinese products including steel, electric vehicles, and photovoltaic goods. Beijing’s emphasis on addressing involution is therefore also partly aimed at alleviating international pressure related to overcapacity concerns.
4. Beijing’s attempt to “have it both ways” is also evident in the domain of social governance. Under severe fiscal constraints, the central government is forced to strike a delicate balance between sustaining growth and safeguarding basic livelihoods.
The April Politburo meeting reiterated the need to “resolutely prevent the risk of large-scale return to poverty,” a significant warning. Notably, the income sources of tens of millions of migrant workers have already been adversely affected as activity in the construction sector weakens (closely tied to the downturn in real estate). In the first quarter, profits in construction and related industries declined sharply, directly undermining wage income. In response, Beijing has sought to artificially sustain rural income growth through measures such as “employment assistance workshops” and household-level industrial subsidies. This model, which relies on fiscal redistribution rather than market-based employment, faces serious sustainability challenges, particularly in a context where fiscal revenue has grown by only 2.4 percent.
The April Politburo meeting also called for accelerating the clearance of arrears owed to enterprises. With land transfer revenues plunging by 24 percent, however, local governments are facing acute repayment pressures, which may further exacerbate liquidity strains in the private sector.
5. All in all, China’s economy in the first quarter of 2026 exhibits a form of “asymmetrical prosperity” shaped by policy intervention. A close reading of the April Politburo meeting and related macroeconomic data suggests that Beijing’s current policy stance is marked by deep internal contradictions:
- Beijing pursues 15.5 percent industrial profit growth, yet overlooks the extreme divergence underlying this expansion and the fragility of corporate cash flows. Advances in high-tech sectors cannot offset the contraction in traditional industries such as automobiles and the real estate supply chain.
- Beijing aggressively promotes investment in “new quality productive forces,” while simultaneously confronting overcapacity, “involution,” and price wars driven by duplicated investment. This ultimately erodes the innovation capacity of enterprises over the long term.
- The central government requires local authorities to stabilize growth, prevent a return to poverty, and settle arrears. Yet the primary revenue source of local governments — land sales — is collapsing at a rate of 24 percent. This severe mismatch between fiscal capacity and administrative responsibilities often leads to distortions at the grassroots level, including data falsification or coercive revenue extraction (e.g., expansion of non-tax income).
At present, Beijing is attempting to address structural economic challenges through technological subsidies and administrative intervention, without undertaking fundamental institutional reforms (such as constraining government power or establishing a level playing field for competition). The strategy of seeking to sustain both capacity and order, while managing both debt and stability is, in essence, a race against time, with the ultimate boundary defined by the limits of China’s long-term growth potential.
