A Hidden Reality

Sizing the Losses inside the Chinese Banking System

LETTER TO READERS

A few times a year we hand these pages to someone else. We invite a guest author whose work we read closely and whose judgment we trust, and we give them the floor. This month that author is Michael Nicoletos of DeFi Advisors.

I met Michael about ten years ago at an investment conference. I knew the name before I knew the man. The format that week required every attendee to put a thesis on the table and let the room take its shots. One of the speakers, a brilliant and widely respected investor, laid out his idea. When he finished, Michael raised his hand and said, as politely as anyone has ever said it, that he would like to politely disagree, and here was why.

What followed was the most gracious demolition I have ever watched. He took the thesis apart piece by piece, with the charm of a diplomat and the precision of a surgeon, and he did it without raising his voice or scoring a single cheap point. I have never forgotten it. That was the afternoon I decided Michael was someone whose opinion I would always seek and always value. In full disclosure, in the years since he has become not only a voice I value but also a close personal friend.

He is not planting a flag on a brand new idea, and he would be the first to tell you so. The notion that China’s reported numbers understate the damage inside its financial system is not his invention. What he has done is come at it from a different angle. He issues no large, rigid proclamations. He sets down one careful observation after another, then asks the simple questions the official numbers cannot answer.

How does the largest property correction in modern history sit inside the largest banking system on earth and leave a bad loan ratio of 1.52%?

How does an economy whose largest single component is shrinking still grow at close to 5%?

He lays the facts out, asks the question any honest reader would ask, and then invites you to answer it for yourself. That restraint is the whole point, and it is vintage Michael.

We are sharing this for a straightforward reason. His work sits directly on top of ground we have already covered. Readers of Who Gets the House and Turning Japanese will know the terrain: a debt-financed property complex, a banking system that will not mark its losses, and a state that chooses the slow workout over the sudden reckoning. Michael carries that framework to its largest and least transparent application, the Chinese banking system, and sizes the hole.

This is a subject we believe every investor should understand, and not only the ones holding Chinese assets. Most readers will tell themselves they have no position here. The knock-on effects say otherwise.

A banking system this size, spending a decade absorbing losses, cannot buy commodities, foreign assets, or finished goods at the pace the world has come to expect, and a currency that pays the bill in real terms pulls on everything priced against it. What happens inside China’s loan book does not stay inside China’s loan book.

Read it with an open mind. Michael has earned that much, and the questions he raises deserve a real answer rather than a reflexive one.

P.S. You can find and follow Michael’s work directly at his Substack, substack.com/@mnicoletos, and through DeFi Advisors at defi-advisors.com. We highly recommend you do so. 

EXECUTIVE SUMMARY

This research note is a thought experiment with consequences. It asks one simple question. If the largest property correction in modern history is unfolding inside the largest banking system on the planet, why do the official numbers show almost nothing?

China’s real residential property prices ended the fourth quarter of 2025 at their lowest level in 20 years, down roughly 23% in real terms from the September 2021 peak. As of June 2026, new home prices have now fallen for 35 consecutive months, the longest streak on record. Independent estimates put the number of empty or unsold homes at somewhere between 65 and 80 million, enough to house the population of a large country. And yet Chinese commercial banks report a nonperforming loan ratio of 1.52%, essentially unchanged from before the property bust began. That number has no historical precedent. No comparable property crash, anywhere in the world, has ever produced an outcome remotely like it.

The purpose of this report is not to predict a collapse. The Chinese state has tools that other economies lack, and the most likely path is a slow, Japanese-style workout rather than a Lehman-style rupture. The purpose is to measure the gap between what the official numbers say and what global precedent says should already be visible. That gap is the story.

Three findings stand out. First, on conservative assumptions consistent with the United States after 2008, Chinese banks are sitting on roughly $2.53 tn of unrecognized credit losses. On a Spain- or Japan-style outcome, the figure is about $4.89 tn. On an Ireland-style outcome, it reaches roughly $7.65 tn. Second, these are losses on the banks alone. Once the damage that sits just outside the banking system is added, in trust and wealth products, in insurance, in overseas lending, and in the fiscal cost the central government would have to absorb, the full economic loss is larger still. Section 9 sets it out in detail. Third, the buffers that exist inside the banking system today, the full stock of loan loss provisions plus a full year of profits, total roughly $1.28 tn, which is less than even the most optimistic estimate of the banks’ own losses.

Pulling every piece together into a single consolidated estimate, the property complex, local government debt, the core loan book, regional banks, wealth and trust products, insurance, the overseas Belt and Road lending, and the unrealized losses buried in bond books, the triangulated total economic loss in the Chinese financial system lands at roughly $3.86 tn in the best case, $7.68 tn in the base case, and $12.20 tn in the worst case. Set against the $520 bn in bad loans and $1.08 tn in provisions that China actually discloses, this leaves an unreserved gap of between $2.78 tn and $11.12 tn, equal to roughly 4% to 16% of all banking assets and 13% to 54% of the entire 2025 economy. The disclosed provisions appear to provide more than two times the coverage of the official bad loans. However, when measured against historical precedents, the real coverage is closer to one-seventh.

The reported 1.52% nonperforming loan ratio is not an observation about asset quality. It is an accounting choice, made systematically across a banking system that now controls roughly $70.20 tn of assets. That choice has limits. This report explains where those limits sit and what it would mean if they were ever tested.

The case does not rest on the banking numbers alone. It rests on the gap between those numbers and the real economy behind them. Across 2025 and into 2026, every major engine of the Chinese economy has been deteriorating at once. Property sales have more than halved from their peak, manufacturing has spent most of the year in contraction under more than three years of factory deflation, and consumption has retreated as households save rather than spend. The economy as a whole has slipped into deflation, which quietly raises the real burden of every existing debt. An economy under this much simultaneous strain does not produce a banking system with almost no bad loans. The charts in this report show the deterioration directly and trace the line from each cracking pillar of the economy into the loan book that is not yet showing the damage.

None of this requires believing that Chinese officials are simply inventing numbers. The mechanism is subtler and better documented than that. The International Monetary Fund, in its 2025 Article IV review published in 2026, wrote plainly that expanded restraint measures covering property, local government vehicles, medium-sized enterprises, and retail borrowers may veil underlying asset quality and delay loss recognition. The same report noted that China’s national asset managers absorbed about RMB 16 tn ($2.37 tn), nearly 18% of GDP, of nonperforming assets between 2012 and 2024, often at potentially inflated prices with limited pricing visibility. When the system’s supervisor says the disposal prices may not be accurate, an outside analyst is entitled to ask what the actual losses are.

The IMF’s broader read on China points the same way. The April 2026 World Economic Outlook, an edition dominated by the energy shock from the war in the Middle East, projects Chinese growth slowing from 5.0% in 2025 to 4.4% in 2026 and 4.0% in 2027, and to an average of just 3.1% a year over 2028 to 2031, well below the pre-pandemic norm. The Fund urges Beijing to use fiscal policy to stabilize the property sector and confront the local government debt overhang. The supervisor is describing a managed deceleration that still needs active intervention on exactly the two fault lines this report measures. That is corroboration, not coincidence.

There is one more thread, developed in Section 4, that makes the whole picture worse rather than better. The losses estimated here are measured against the size of the Chinese economy, and that denominator is itself almost certainly overstated. Investment is roughly 40% of Chinese GDP, the highest share of any major economy in history. That investment is now shrinking, with fixed-asset investment falling and private and property investment down by double digits. An economy whose largest component is contracting cannot honestly grow at close to 5%, yet that is what the official numbers claim. The same habit of never writing down bad debt that produces an impossible 1.52% bad-loan ratio also inflates GDP because investment that costs more than it is worth is carried at full value rather than written off. The result is that the true economy is smaller than reported, which means every loss figure in this report, expressed as a share of GDP, is, if anything, understated.