China industry analysis 2026: FDI is down, entrants are up

Two numbers that disagree, and why that is the story

MOFCOM’s full-year 2025 release, published in January 2026, contains a genuine contradiction if you read only one line of it. Actual foreign investment used fell 9.5% to ¥747.69 billion, following a 24.7% collapse in 2024. On that basis the “foreign capital is fleeing China” headline writes itself.

The next line says 70,392 new foreign-invested enterprises were set up, a 19.1% increase. That is not the behaviour of investors leaving. Both numbers come from the same statistical release, measuring different things: one counts money, the other counts decisions. Money is dominated by a handful of enormous projects — a chemical plant, a battery gigafactory, an auto joint venture — so a single deferred megaproject can move the national total by billions. Decisions are dominated by small companies. And small companies were, on this evidence, distinctly bullish.

The average entrant is shrinking, fast

Divide one series by the other and the shift is stark. On 2025’s figures, average foreign investment per newly established enterprise works out at roughly ¥10.6 million. Back out 2024 from the reported growth rates and the equivalent figure was around ¥14 million. The typical foreign entrant to China got about a quarter smaller in a single year.

2024 figures are derived from the reported year-on-year growth rates rather than restated official totals, so treat them as close approximations rather than precise numbers.

Two forces produce that. Large industrial investors are pausing on tariffs, geopolitics and overcapacity in the sectors they would have built into. Meanwhile the cost of starting a Chinese company has collapsed: registration is essentially free in many districts, the 2024 negative list removed the last manufacturing restrictions, and a small consultancy or trading entity can run compliantly on a few thousand renminbi a year. When the entry ticket falls, more people buy one.

Where the money actually went

China’s inbound investment is now overwhelmingly a services story:

• Services: ¥545.12 billion — about 73% of the total.
• Manufacturing: ¥185.51 billion — about 25%.
• High-technology industries: ¥241.77 billion — roughly a third of all inbound investment, cutting across both.

The growth is concentrated in a few identifiable places. E-commerce services attracted 75% more foreign investment than the year before. Medical instruments and equipment manufacturing rose 42.1%. Aerospace equipment rose 22.9%. Those are not the sectors of a country being written off; they are sectors where foreign firms judge that Chinese demand, Chinese engineering talent or the Chinese supply base still cannot be replicated elsewhere at the price.

Who is still investing

The source-economy data cuts against the simple decoupling narrative. Investment from Switzerland rose 66.8%, from the UAE 27.3% and from the UK 15.9% (on figures that include investment routed through free ports). European industrials, Gulf capital and financial services are not following the same trajectory as US-China trade politics, and the countries most exposed to tariff conflict are not the whole picture.

China+1, correctly read

China+1 is real, and it is routinely misread as China-minus. What actually relocates is usually final assembly — the last, most tariff-visible, most labour-intensive step. What tends not to relocate is everything upstream: components, sub-assemblies, tooling, moulds, specialised machinery and industrial chemicals. A Vietnamese or Mexican plant assembling goods from Chinese inputs has diversified its country-of-origin exposure, not its supply-chain dependency.

Much of the capital building that alternative capacity is itself Chinese, as manufacturers follow their customers offshore to preserve market access. For a buyer, the practical consequence is that “we moved to Vietnam” frequently means dealing with a Chinese-owned factory, staffed by Chinese managers, buying Chinese components — at a higher landed cost, with a shallower local supplier ecosystem, and with the same due-diligence questions you would have asked in Guangdong. Diversification is often the right call. It is rarely the simplification it is sold as.

What this means if you are small

• You are now the typical entrant, not the exception. A ¥10 million average entity is a small business. Advice and cost benchmarks written for a multinational subsidiary no longer describe the median case — see what it really costs to set up and run a company in China.
• Services entry is the well-trodden path. Three-quarters of the money and most of the new entities are in services, where entry is cheap, reversible and does not require fixed assets.
• Manufacturing is fully open, and quieter. Restrictions on foreign investment in manufacturing hit zero under the 2024 negative list, just as large industrial investors stepped back. Fewer competitors is a real feature of the current moment.
• Test before you commit. Falling total capital and rising entity counts describe a market being sampled rather than bet on. Sampling — a distributor, a sourcing relationship, a small entity — is a strategy, not a failure of nerve. Many businesses need no Chinese entity at all to begin.

Quick FAQ

Is a third year of falling FDI a warning sign?

It is a real signal about large capital-intensive commitments, and it should be read as one. It says considerably less about whether a small services or sourcing business can operate profitably in China, which is governed by entirely different economics.

How reliable are these figures?

They are MOFCOM’s official series and the standard reference, but “actual foreign investment used” has known quirks: it includes investment routed through free ports such as Hong Kong and the Cayman Islands, which obscures the ultimate source, and it does not net off disinvestment. Use it for direction and composition, not precision.

When is the next data point?

MOFCOM publishes cumulative FDI figures monthly, with the full-year release in January. Sector and source-economy detail appears in the year-end release rather than the monthly ones.