China’s Two-Speed PMI: High-Tech Manufacturing Reads 53.5, the Rest Reads 48

China’s National Bureau of Statistics released June manufacturing data this morning, and the headline number looks reassuring: the official PMI climbed to 50.3 from 50.0 in May, beating consensus expectations of 50.1 and marking a third consecutive month above the expansion threshold. Output growth accelerated to 51.4 from 51.2. New orders returned to expansion at 51.2, up sharply from 49.9. If you stopped reading at the top line, you might conclude that China’s manufacturing sector is on a steady, if modest, recovery path.

You would be wrong. Or rather, you would be right about one China and dangerously wrong about the other. Because inside today’s PMI release are two economies operating on parallel tracks with almost no intersection — and the gap between them is widening every month.

The 53.5 Economy: AI Chips, Ascend Orders, and Global Demand

High-tech manufacturing posted a PMI of 53.5 in June, up 0.6 points from May and comfortably in expansion territory. This subsector — semiconductors, electronic components, AI hardware, electric vehicle batteries, precision instruments — is not merely growing. It is booming at a rate that makes the overall 50.3 headline look like an averaging artefact, which is precisely what it is.

The driver is the global semiconductor supercycle. Chinese integrated circuit exports rose 110% year-over-year in the first five months of 2026. In April alone, monthly IC exports hit approximately $31 billion — double the year-ago level. Goldman Sachs and Nomura estimate that semiconductors and AI-related products accounted for roughly half of China’s total export growth.

This is partly a story of import substitution forced by US export controls. Huawei’s Ascend AI chip line is projected to generate $12 billion in revenue in 2026, up 60% from 2025. ByteDance committed $5.6 billion in Ascend 950PR orders — the largest single disclosed domestic AI chip procurement in Chinese history. ChangXin Memory Technologies (CXMT) saw revenue surge 130% to over $8 billion. NVIDIA, which once commanded more than 90% of the Chinese AI chip market, has seen that share decline to approximately 50% as domestic alternatives gain traction. SMIC plans to double its production of 7nm chips in 2026.

In new export orders, the June PMI told the same story: the sub-index returned to expansion at 50.1, up from 48.6 in May. But this recovery was concentrated in high-tech and AI-linked categories. Traditional manufactured goods — textiles, furniture, non-electronic machinery — showed little improvement. The Caixin Manufacturing PMI, which surveys smaller, more export-oriented private firms, has consistently outperformed the official NBS reading, registering 51.8 in May. The divergence is structural: firms plugged into global AI supply chains are seeing order books fill. Everyone else is not.

The 48 Economy: Employment, Property, and a Consumer Who Has Stopped Spending

Now look at the sub-indices that measure the domestic economy’s health. Employment contracted for a fifth consecutive month, falling to 48.4 from 48.6 in May. This is the number that matters most for social stability and it is moving in the wrong direction. High-tech manufacturing is capital-intensive, not labour-intensive. A single SMIC fabrication plant employs perhaps 3,000 workers. A single Foxconn assembly complex employs 200,000. The AI chip boom is generating enormous export revenue but relatively few jobs — and none that absorb the 15.6% of urban Chinese youth currently unemployed.

The consumer is reflecting this reality. Consumer confidence fell to 89 in April, continuing a decline that has accelerated since mid-2024. Surveys report a cultural shift toward “thrift culture,” with secondhand goods markets booming and luxury spending slowing. Retail sales grew only 2.4% year-over-year in Q1 — in an economy that needs 5%+ consumption growth to hit its GDP targets. The consumer collapse that first appeared in late 2025 has not reversed. It has calcified.

Property — still the largest store of household wealth — continues to erode. Second-hand residential prices in Beijing fell 8.3% year-over-year in Q1 2026, according to Global Property Guide data. Investment in the sector declined 16.2% in the first five months of the year, accelerating from 13.7% in January–April. Manufacturing capacity utilisation sank to 73.9%, near a decade low outside the early-2020 lockdowns. These are not the statistics of an economy in cyclical recovery. They are the statistics of an economy undergoing a structural transition where the old growth model is dying faster than the new one can expand.

The Structural Mismatch: Why AI Exports Cannot Save the Labour Market

The most important economic question in China today is not whether GDP will hit the government’s 5% target — Q1 came in at exactly 5.0%, and the combination of AI-driven exports and fiscal spending on infrastructure will likely keep the headline number within reach. The question is whether the new growth model can absorb the workforce that the old model is shedding.

The evidence from today’s PMI says it cannot. High-tech manufacturing is capital-intensive by nature. The semiconductor fabrication plants, battery gigafactories, and precision manufacturing lines that drive the 53.5 PMI reading employ a fraction of the workers that construction, real estate services, and traditional consumer-facing industries once did. China’s property sector at peak employed roughly 30 million workers directly and another 20–30 million in adjacent services. The entire semiconductor industry employs fewer than 2 million.

This creates a paradox visible in today’s data: output is expanding (51.4), new orders are returning (51.2), purchasing activity is rising (51.4), but employment is contracting (48.4). Factories are producing more with fewer workers. The productivity gains that make China’s high-tech sector globally competitive are the same dynamics that prevent it from absorbing displaced workers from the domestic economy. Each SMIC 7nm wafer generates more revenue per worker than an entire floor of a Foxconn assembly line. But the economy needs Foxconn-scale employment, not SMIC-scale revenue per head.

The Civil Engineering Bright Spot — and Why It Is Not Enough

One genuinely positive signal in today’s release: the business activity index for civil engineering construction rose above 55, up more than 3 points from May. This reflects Beijing’s fiscal playbook — when private demand falters, state-directed infrastructure investment fills the gap. Local government special bond issuance has accelerated in H1 2026, and visible activity on roads, rail, and water projects has picked up.

But infrastructure spending is a bridge, not a destination. It can prevent a GDP contraction in the short term — and China has deployed this lever successfully for two decades — but it does not solve the consumption problem. The property wealth effect that once translated construction activity into consumer spending is broken. When homeowners in Beijing are watching 8.3% annual declines in their primary asset, no amount of highway construction in Guizhou will persuade them to open their wallets. The savings rate has risen to 4.0% nationally, and household surveys report a shift toward precautionary behaviour that economists at CKGSB have described as a potential generational change in consumption norms.

Services: The Other 50

The non-manufacturing PMI ticked up to 50.2 from 50.1, and the services business activity index edged to 50.4. Both are above the expansion line but barely — and in a services sector that accounts for over 50% of China’s GDP, readings this close to 50 are effectively stagnation. Compare this with India, where services PMI readings have held above 58 for most of 2026, or even Spain, where services expansion above 55 is driving the eurozone’s internal divergence. China’s services sector is not contracting, but it is not growing fast enough to absorb displaced workers or generate the consumption rebalancing that policymakers have targeted for over a decade.

What the Two Speeds Mean for the Second Half

The first half of 2026 ends today. China’s H1 scorecard is clear: Q1 GDP at 5.0%, powered by exports (+53.2% in May, headline) and state-directed infrastructure, while household consumption, property, and private employment weaken. The PMI data released this morning is the final data point of Q2 and it confirms the pattern rather than interrupting it.

Three dynamics will determine H2. First, whether AI chip demand sustains at current levels. The $725 billion in hyperscaler capital expenditure driving global semiconductor orders is enormous but concentrated in five companies. Any retrenchment would hit China’s high-tech manufacturing PMI directly. Second, whether the Hormuz peace deal and the resulting drop in oil prices from $111 to $83 per barrel feeds through to lower input costs for Chinese manufacturers — the purchasing price sub-index has been elevated for months. Third, whether Beijing deploys further stimulus targeted at consumers rather than infrastructure. The July Politburo economic work conference will be the signal.

The deeper issue is structural and will not be resolved in H2 or in 2027. China is attempting something no major economy has successfully done at this scale: transition from a property-and-infrastructure growth model to a technology-and-innovation model while simultaneously managing a demographic contraction (population peaked in 2022), a structural deflation in asset prices, and an employment crisis among the young. The 53.5 high-tech PMI is evidence that the new model is working. The 48.4 employment reading is evidence that the transition is leaving most of the workforce behind.

The headline PMI says 50.3 — expansion. The labour market says 48.4 — contraction. The right interpretation is not one or the other. It is both, simultaneously. China in mid-2026 is running two economies on one landmass, and which one you see depends entirely on which number you read first.

Compare China’s economic trajectory with other major economies on our country comparison tool, or explore the full dataset on our global rankings page. For more on the semiconductor dynamics driving this bifurcation, see our analysis of one-legged economies in the chip supercycle.