China's August purchasing managers' data improved in the place markets notice first: the manufacturing headline. The official index rose 0.6 percentage points to 49.8, close to the 50 line that separates expansion from contraction in the survey. Reuters described stronger demand alongside continued weakness in services and construction.
Close is not the same as broad. The National Bureau of Statistics data show a recovery concentrated in large manufacturers and new orders, while smaller businesses, employment, inventories and the economy-wide composite remained in contraction. That combination is encouraging for industrial momentum but insufficient evidence of an economy-wide turn.
Large factories crossed the line alone
The sharpest division runs through company size. Large manufacturers rose 1.1 points to 50.6, moving into expansion. Medium-sized firms slipped 0.3 points to 49.4. Small firms improved by half a point but remained much weaker at 47.9. The aggregate therefore sits just below 50 even though its largest component by company size crossed the threshold.
This split matters because a rise led by large factories can support industrial production, exporters and major supply chains without spreading evenly through private-company cash flow, local hiring or supplier demand. The survey does not identify why the gap exists, so it would be speculation to attribute it to credit access, ownership or one policy. It does establish that the average improvement was not shared equally.
For investors, breadth is a quality test. Earnings exposure to large industrial and technology exporters may improve before businesses serving smaller factories or local demand. Yet one month of diffusion indices cannot reveal output volumes or listed-company profits. The defensible conclusion is narrower: large manufacturers reported expansion; the other size groups did not.
Orders improved before payrolls and inventories
The internal sequence is more constructive than the headline alone. Production rose to 50.4 and new orders jumped 2.1 points to 50.6. New export orders also edged into expansion at 50.1. The official statistical interpretation therefore has evidence for saying both production and demand improved.
The next links in the chain have not followed. Manufacturing employment fell to 48.7. Raw-material inventories were 48.1, finished-goods inventories 48.4 and backlogs 46.7. Imports improved to 48.6 but remained below 50. Companies reported more orders and slightly more production without collectively signalling hiring, restocking or a fuller order book.
That could be a normal lag. If managers believe the order rebound will last, they may first use spare capacity, then rebuild inputs and hire. Business expectations at 53.8 support that counterargument, even after easing 0.3 points. But the same pattern can fade if orders were temporary or if existing capacity can absorb them. September's employment, inventory and backlog readings will distinguish those paths better than another small move in the headline.
Services kept the whole economy below 50
Manufacturing is only one side of China's rebalancing challenge. The official non-manufacturing business-activity index stayed at 49.0. Services remained at 49.3 and construction slipped to 46.9. More concerning for near-term momentum, non-manufacturing new orders fell to 44.1, including 44.5 in services and 42.4 in construction. The composite output index improved only 0.2 points to 49.5.
Those readings fit a longer-running imbalance without proving its cause. The World Bank's July update described early-2026 resilience supported by exports and high-tech investment alongside weaker domestic demand and cautious consumers. The IMF has similarly argued that property weakness and limited household safety nets constrain consumption and leave growth too dependent on external demand.
The August survey is consistent with that diagnosis: goods orders strengthened while services did not cross into expansion. Consistency is not causality. PMI respondents do not explain household saving or property wealth effects, and the survey is not a national-accounts estimate. Its contribution is timelier: the hoped-for shift from industrial supply toward broad domestic activity was not visible across the August diffusion measures.
Factory prices add a second test
Price indices moved faster than activity. The manufacturing input-purchase-price index climbed 3.4 points to 56.6, while the producer-price index rose 2.6 points to 50.4. More firms therefore reported rising input costs, and output prices finally moved just above the neutral line.
That can be read two ways. If stronger orders let manufacturers pass through costs, the move may help end damaging price declines and improve nominal revenue. If input prices outrun what firms can charge, margins can tighten even as volumes recover. PMI diffusion indices show the breadth of reported direction, not the size of each price change, so subtracting one index from the other would not measure a profit margin.
The useful test is whether producer prices and company earnings confirm that pricing power is returning, especially outside large firms. A price rebound accompanied by weak small-company activity and falling employment would be less healthy than one paired with broader orders, payrolls and services demand.
September must broaden the rebound
The August report gives optimists real evidence: new orders, export orders, production and large-company activity all crossed 50, and expectations remained comfortably above it. It also sets a demanding confirmation test. Medium and small manufacturers need to improve, employment and inventories need to follow orders, and services new orders need to recover from the mid-40s.
A reversal in manufacturing orders would weaken the rebound thesis immediately. Conversely, a second month of order expansion accompanied by better hiring, restocking and service activity would show that the initial strength is diffusing through the economy. Until then, 49.8 is best read as a concentrated industrial improvement—not a broad recovery hidden by a decimal point.