What's behind this round of A-share plunge?

[[**A-shares fell on heavy volume on October 22, with the ChiNext Index down more than 40% from its summer high; the external trigger was the Fed restarting rate hikes and the 10-year Treasury yield breaking above 5%, weighing on valuations of global risk assets.**

**Key points**

**1. The Fed restarting rate hikes is the core external trigger**

On September 17 it raised rates by 25 basis points to 3.75%-4.00%, October CPI again beat expectations, and the 10-year Treasury yield broke above 5% in September; the market fears rates "staying high" more than a single hike.

**2. A-shares fell on heavy volume, and even the high-growth AI chain was sold off**

On October 22 the Shanghai Composite fell 1.48% and the ChiNext Index fell 4.37%, with turnover of 1.53 trillion yuan; Zhongji Innolight's orders are booked through 2027 and Yuanjie Semiconductor's net profit was forecast to rise more than fivefold, yet both fell, and the ChiNext Index is down more than 40% from its summer high.

**3. The trading logic has shifted from inflation to credit, and highly leveraged new cloud vendors are the most fragile**

On October 22 London gold fell 2.79% and Brent crude rose 4.0%; stocks and gold falling together shows funds are moving into cash; CoreWeave's debt-to-equity ratio is about 739 times, its 5-year CDS spread has broken 1,000 basis points, and its refinancing window is drawing attention.

**4. A-share valuations are already low, but rate-hike expectations have not stopped; the key is the December FOMC meeting**

The ChiNext Index's price-earnings ratio has fallen to 30 times, the 8th percentile since 2010, yet it still fell; on October 27-28 the Fed is expected to hold steady, and it may raise another 25 basis points in December; the debate is whether this is the 1996 productivity boom or a repeat of the 1999 overheating. 2026-10-25 20:09 What is behind this A-share selloff? Future Charging Station quick take

This article is from the WeChat public account: Future Charging Station, author: Future Charging Station

At the close on October 22, the Shanghai Composite stood at 3,658.03 points, down 1.48%, while the Shenzhen Component and ChiNext Index fell 2.96% and 4.37% respectively. Turnover in the two markets was 1.53 trillion yuan, 99.4 billion more than the previous day; volume expanded on a down day, showing that sellers are anxious and buyers are starting to bargain over price.

Among the hardest hit, a large share were the companies with the fullest order books. Zhongji Innolight's orders are booked through 2027, and Yuanjie Semiconductor's third-quarter net profit was forecast to rise more than fivefold, yet their shares still fell; the ChiNext Index is already down more than 40% from its summer high.

One important reason for this selloff cannot be avoided: the Fed is raising rates again.

It raised rates by 25 basis points on September 17, CPI again beat expectations on October 14, and the 10-year Treasury yield broke above 5% in September. Why should a US interest rate affect A-shares?

To understand this, one must start with why the Fed cut rates in the previous two years.

After the final hike in July 2023, the Fed held rates steady, began cutting in 2024, and by September of this year, before the rate hike, the policy rate had already fallen to 3.5%–3.75%.

There were two reasons for the rate cuts. Inflation was trending down, and for most of 2025 core PCE was moving toward 2%. Interest on US debt was close to crushing the fiscal position; much of what the Treasury raised each month by issuing new debt went to paying interest on old debt.

In fiscal year 2026, the US federal deficit was $1.992 trillion; net interest expense exceeded $1 trillion for the first time, interest paid on public debt surpassed $1.3 trillion, and total debt had long since passed $40 trillion. Interest became the third-largest expenditure after Social Security and Medicare; raising rates further would be more than the fiscal position could bear.

Interest is rigid, and Social Security and Medicare cannot be cut either, so only other spending can be cut. So for those two years, the market's consensus expectation was that rates would keep falling.

That logic was broken this year. In the first half, the Fed held steady, wanting to see whether inflation would fall on its own, but that did not happen.

August CPI was 3.4%, and in September it returned to 3.7%, still far from the 2% target. Core CPI rose 0.3% month on month, strong for two consecutive months; energy prices were up 20.4% year on year, and airfares up 22.4% for the year.

Price increases spread from energy to services. Rents, services, and medical care were all rising, and with tariffs and geopolitics, the supply side was not cooperating either.

In its September statement, the Fed deleted wording blaming high inflation on "supply shocks" and instead said "domestic consumer spending remains resilient." Translated, that means: price pressure is not all imported; domestic demand is also pushing it, so it cannot simply wait for the market to change on its own.

So on September 17, the Fed raised rates by 25 basis points in a 12-0 vote, to 3.75%–4.00%. It was the first hike since 2023 and the first since Warsh took office.

A 25-basis-point hike actually has very little direct effect on inflation.

What really matters is the market's expectation about it. Put simply, the Fed controls short-term rates, while long-term rates are priced by thousands of buyers in the market. If the market believes it can suppress inflation, long-term money will cheapen on its own; once it does not believe that, long-term yields will instead spike.

If it raises 25 basis points and long-term yields spike 50, the hike is effectively pointless. That is why the Fed preferred to move first in September rather than wait until all the data were in.

At the September 17 press conference, Warsh's exact words were: it is hard to say overall financial conditions are restrictive. Translated, he still thinks they are not tight enough.

So what it fears is not inflation itself, but "everyone not believing it can suppress inflation." Once trust collapses, every subsequent punch must be thrown with double force, and may still not work. Once credibility collapses, restoring it will cost far more than 25 basis points in interest.

Expectations are invisible, yet the trillions of dollars in long-end bonds are priced on them. On October 16, Warsh said only one sentence in Bangkok — "I won't prejudge the outcome of the meeting" — and the market's expectation of an October rate hike fell from over 25% to under 15%; that day South Korea's KOSPI rose 3.6%. One sentence can send trillions of dollars swinging up and down.

Expectations govern the long end, but what truly bears down on companies is how long rates stay high.

Companies can ride out a brief spike in rates with the low-interest debt they issued in earlier years. But as time drags on, old debt matures and must be refinanced at new rates, and cash flow breaks. The rate hike itself is not what's frightening; what's frightening is "staying at a high level."

The market's reaction to this is far faster and harsher than to inflation data. On October 22, London gold fell 2.79%, Brent crude rose 4.0%, and stocks and gold fell together. Gold is supposed to be a safe haven in turbulent times, yet that day it slid along with everything else.

When safe-haven assets don't provide safety, it shows people aren't seeking safety — they're swapping into cash. The trading logic has shifted from "inflation" to "credit."

There is another side to the rate hike story.

The better demand is, the more money is borrowed to meet it — that's the old law of the capital cycle. AI demand is real, and orders booked through 2027 are real, but the fuller the order book, the more debt the new cloud vendors take on.

Hyperscale cloud vendors can still rely on cash flow; new cloud vendors can only rely on debt. CoreWeave is the most typical case: its debt-to-equity ratio is as high as about 739 times, and by some measures its total debt is $51.6 billion against net assets of only a few hundred million. Its business itself is fine; the problem is the money. With leverage like that, every one-point rise in rates is a step on a knife's edge — and there aren't many steps left.

Its 5-year CDS spread has breached 1,000 basis points, and its bond yield is around 15%. The longer rates stay high, the sooner this kind of company runs into trouble. The most fragile link is precisely the new players most sensitive to high rates; their trouble isn't in orders, it's in the refinancing window.

One market view holds that by 2030, 80% of new cloud service providers will be unable to operate independently. This chain is like a street where every shop relies on the same money house for credit; when the money house pulls its loans, the first to close is the one with the most credit on the books.

At this point, the Fed itself is arguing — over which year's script to use.

One is 1996, when the tech revolution was just beginning, growth was strong, employment was good, yet inflation was mild. The other is 1999, when demand overheated, the Fed began raising rates in June and by May 2000 had hiked 175 basis points, finally pricking the dot-com bubble; the Nasdaq fell from 5,048 to 1,114, a drop of 78%.

Warsh himself would rather believe it is 1996. He sees AI as a "productivity boom": supply will catch up with demand, inflation will fall back on its own, and raising rates now would be tantamount to killing growth.

What critics worry about is 1999. The productivity dividend has not yet materialized, demand overheats first, and by the time the data confirms it, it will be too late. Warsh himself has said that if inflation does not fall back quickly enough, we are prepared to act again.

The divergence between these two histories ultimately comes down to the same question: is the productivity dividend coming fast enough?

The September 1996 meeting deserves separate mention.

After the 1994-1995 period had just seen 300 basis points of hikes, taking rates from 3% to 6%, the economy made a soft landing, and rates were cut several times in 1995. By 1996, growth was strong and unemployment low, yet inflation would not rise. In those two years, the economics profession called this the new economy, saying information technology had flattened the Phillips curve.

At the September meeting, Greenspan persuaded his colleagues to forgo a rate hike. More than two months later, he delivered his "irrational exuberance" speech, stating plainly that stock prices had detached from fundamentals; the phrase he used was irrational exuberance.

He saw the bubble but believed the cost of pricking it early was too high, preferring to let the market run its own course. This is the core of the 1996 script: bet that the productivity brought by the technological revolution can absorb overheated demand. Was he right? He was right in the first half; after that meeting, the Fed waited another half year, not raising rates until March 1997. But the bubble kept inflating and burst in 2000.

The problem is that the productivity dividend is never broadly shared in its early stages.

McKinsey did the math in 2001: from 1995 to 2000, U.S. labor productivity grew 2.5% a year on average, versus only 1.4% in the previous two decades — it looked like a miracle. But this jump was almost entirely explained by six industries: retail, wholesale, securities, telecommunications, semiconductors, and computer manufacturing.

These six industries account for about 35% of the nonfarm private economy; the remaining 69% of the economy, 53 industries, contributed only 0.3% to productivity growth. Even more counterintuitive, 62% of IT investment flowed to those 53 sectors that did not improve efficiency.

Technology does not equal productivity; the dividend is highly concentrated — this is a "local phenomenon." The dividend is real, but it falls only on a minority of industries; the rest pay in money and effort and get little in return. In other words, technology spreads fast, but efficiency gains spread slowly.

Back to A-shares: look at the afternoon of October 23, when six companies including Eoptolink, Zhongji Innolight, and TFC Optical jointly announced orders, and the ChiNext Index surged 4% in a dozen minutes or so.

Orders can lift a stock for a day, but they have not actually changed current valuations. The ChiNext Index's price-to-earnings ratio has already fallen to 30 times, the 8th percentile since 2010. It is still falling even at such a cheap level because the denominator — that is, expectations of rate hikes themselves — has not stopped changing.

The next observation point is the interest rate meeting on October 27-28. The market expects no move this time and another 25 basis point hike in December; CITIC Securities and Huatai Securities judge the same. The real direction will most likely have to wait for the December meeting, or even later. Further out, the dot plot's path is 4.1% in 2027 and 3.9% in 2028.

So part of the answer to the cause of the A-share plunge at the beginning lies right here in rate hikes.

At 3,600 points, it was dragged down by risks from across the ocean. Channel: Finance and Economics. This content comes from the internet; original link. The views represent only the author and do not represent Huxiu's position. For copyright issues, please contact hezuo@huxiu.com, and we will verify and handle it promptly. Everyone who is changing the world and wants to change it is on the Huxiu APP [ ]