From lender to debt collector: China's Global South reset

Opening

Welcome to today's China report. Today it's Belt and Road, development finance, and what Beijing is actually doing in the Global South now that the lending boom is over. Five reports. Two of them, a Carnegie paper on India, a Brookings piece on the Hormuz oil shock, aren't about development finance at all, and I'll tell you why they earn their place when we get to them. Two of the five landed within a day of each other in late July, and they disagree about whether China's pullback is a retreat or a re-tooling.

Lowy on the Pacific aid map

First, the Lowy Institute's 2026 Pacific Aid Map, published on the twenty-eighth of July by Riley Duke, Roland Rajah and Jack Xu.

The claim: Australia has taken China's old job. Total development finance to the Pacific was four point one billion US dollars in twenty twenty-four. Australia disbursed one point five billion of that, thirty-seven percent, more than three times the next bilateral partner. China's new loan commitments fell from about three hundred and sixty million dollars a year in the twenty-tens to about ninety million a year since twenty twenty. But China hasn't left. It's swapped instruments. Grant commitments in twenty twenty-four hit close to twice the pre-pandemic average, in smaller and smaller packages, school upgrades, police equipment, scholarships. Meanwhile the old loans bite. Samoa and Vanuatu each owe Beijing more than one percent of GDP, and Vanuatu's repayments stay elevated out to twenty thirty-two.

The mechanism is that grants buy the same diplomatic access loans used to, more cheaply. And the report shows aid tracking diplomacy pretty directly, Nauru after it switched recognition, Vanuatu during security talks.

What's it built on? Nine editions of a hand-built database. Fifty thousand projects, sixty billion dollars, two thousand and eight to twenty twenty-four, cross-checked with recipient governments. For Australia and the multilateral banks that's about as good as this gets.

Here's my problem. Lowy says plainly that the China numbers are the weakest part, scraped from budget documents, press releases and embassy websites, and projects will be missing. So a fall in Chinese lending is partly a fall in announced Chinese lending. And the report's own procurement data cuts against its headline: Chinese state firms took fifty-two percent of Asian Development Bank infrastructure contract value in the Pacific between twenty sixteen and twenty twenty-five, and fifty-six percent of the World Bank's. Australia may be writing the cheques. Chinese companies are still pouring the concrete.

CGD on China's shift from lender to collector

Second, "Beyond Belt and Road," a note from the Center for Global Development by Charles Kenny, published the twenty-seventh of July.

Kenny's line is that China has gone from financier to debt collector, and he means it literally. Annual sovereign lending commitments are down to roughly seven billion dollars from around twenty-eight billion in the twenty-tens. Net flows in twenty twenty-four were minus thirty-four billion, more money came back to China than went out. Infrastructure fell from sixty-four percent of commitments in the two-thousands to twenty-five percent in the twenty-twenties. In Africa, public infrastructure support went from over fourteen billion a year to under five.

The chain is: China filled a gap traditional donors had abandoned, on expensive terms, interest rates roughly double the World Bank's, and now it's unwinding the position, leaving a hole in the poorest countries. The recommendation is that rich donors could fill it cheaply, because a sovereign loan guarantee only scores its insurance cost against the budget.

The evidence is AidData's tracking of Chinese lending, an outside research dataset built from open sources, because China doesn't report to the OECD. It's the best there is. It is still estimates.

And here's where I'd push. Kenny concedes the line between public and private Chinese deals is blurred, with state commercial banks lending to recipient state firms under confidentiality clauses. If that's true, then part of the "collapse" is money moving into a category the dataset counts differently. His own figures point that way: seventy-one and a half billion dollars of oil and gas agreements in twenty twenty-five, thirty-two and a half billion in metals and mining, typically collateralised against the resource. That's not a retreat. It's a new business model. And a loan guarantee does not compete with a deal secured on oil.

Carnegie on America and China in MENA

Third, Amr Hamzawy and Kathryn Selfe for Carnegie, published the first of July, on the Middle East and North Africa.

Their argument is convergence, not collision. Both powers want stability and open shipping. Their strengths don't substitute for each other: the US supplied sixty-four percent of the region's arms imports in twenty twenty-five, China zero percent that year and an average of one point three percent since twenty twenty-one. China's trade with the Gulf states was two hundred and eighty-eight billion dollars in twenty twenty-four against America's eighty-six billion. So neither can do the other's job, and regional states are hedging on purpose, sixteen of nineteen are in the Asian Infrastructure Investment Bank, and all but Israel are in Belt and Road.

The evidence here is genuinely public and checkable: SIPRI's arms database, UN trade statistics, Arab Barometer polling, the State Department treaty registry. And they're honest about its edges, they append a correction saying twenty twenty-five Gulf trade data was unreliable, so they pulled a chart, and they say outright there's no comparable data on China's treaty commitments.

That's a solid paper. My challenge is that convergence is inferred from the absence of collision so far, and their own evidence undercuts it. Half of China's crude goes through Hormuz, and when Hormuz closed, Beijing stayed out of the reopening effort. That isn't a complementary division of labour. That's free-riding, and it holds only while someone else pays. What would settle it is simple: does China ever contribute to securing the lane it depends on?

Carnegie on India's China calculation

Fourth, Saheb Singh Chadha for Carnegie, the second of July, "Threading the Needle." Not a finance paper. It's here because India is the largest Global South power pricing China, and the order it prices in is the interesting part.

Four imperatives, and Chadha insists on the ranking: border stability first, economic necessity second, political understanding third, great-power positioning last. The argument is sequencing, disengagement at the border friction points came first, that unlocked the Modi–Xi meeting in October twenty twenty-four, and everything else followed. If economics were driving it, you'd have seen movement earlier, because the pressure was already there: fifty to sixty thousand troops a side, a record trade deficit above one hundred and sixteen billion dollars, ninety-three percent of India's rare earth magnets coming from China.

It rests on official statements from both governments, Chinese state media, and interviews with serving and retired Indian officials. And the author states his own limitation: no Chinese policymakers were interviewed.

That matters, because his headline forecast, that Beijing will offer tactical crumbs without a strategic shift, is a claim about Chinese intent assembled entirely from the Indian side plus Beijing's public messaging. The sequencing argument I buy; the sequence is documented. The mind-reading I don't. Watch whether China moves troops before political trust exists, not after.

Brookings on the Hormuz energy shock

Fifth, Scott Moore for Brookings, the tenth of August. Also not a finance paper, it's here because cleantech exports are becoming China's economic footprint in poorer countries.

Moore's point is deflationary. China rode out the twenty twenty-six Hormuz closure on coal, oil stockpiles and Russian barrels, not on clean energy. Twelve hundred gigawatts of coal, over half the world's. Roughly one point four billion barrels stockpiled, about seven months of imports. So "electrostate" is overstated. But manufacturing share, over ninety percent of solar modules, over eighty percent of batteries, means Beijing profits from the panic and from the transition after it. Solar exports hit a record sixty-eight gigawatts in March. Pakistan imported about fifty gigawatts of Chinese solar and insulated its power sector.

The data is trade and generation statistics from the IEA, Ember and others, reliable. And he flags that the reserve number is a state secret, so his figure may be low.

The weakness is time. Records in March, records in May, read as structure. Price spikes make export spikes. And gross sales aren't leverage, the report never shows a government changing a position because of solar imports.

What to watch

So: Lowy says China changed instrument, Kenny says China left. That's a real disagreement, but notice Kenny cites Lowy, and both lean on the same open-source lineage for Chinese numbers. Two reports agreeing here is closer to one and a half.

Three things would move me. Recipient-country debt registries showing whether state-firm borrowing rose as sovereign lending fell. Whether Chinese solar exports hold once oil prices come back down. And whether Beijing ever pays to keep a shipping lane open.