Where is this round of AI cycle heading after global liquidity peaks?

In mid-September, the world's three major central banks completed a shift in the same direction within the same week.

On September 18, the Bank of Japan announced a 25-basis-point interest rate hike, pushing the rate to its highest level in 31 years, marking its second rate hike within the year. The day before (September 16 US time), the Federal Reserve's first interest rate hike in more than three years officially took effect, raising the interest rate range to 3.75%–4.00%; one week earlier, on September 10, the European Central Bank also raised the three key interest rates of the euro area by 25 basis points simultaneously, which was also the second hike within the year.

At this point, the United States, Europe and Japan have all shifted to raising interest rates, and the total valve of global liquidity is being tightened step by step.

Interest rate hikes sound far away from us, but their echoes are right around us.

Over the past year, the topics that went viral on social media feeds have quietly changed from "what new features the new model has added" to "how much a certain stock rose or fell today" and "how many times the valuation of a certain company has multiplied"; in large Silicon Valley tech firms and startups, what engineers are keen to discuss is no longer technical parameters, but whether the valuation of a certain company is inflated, and "the options I hold, should I go to Company A, Company O, or Company X".

These subtle daily changes are signals in themselves: a dollar asset cycle driven by massive liquidity has reached its tail, and the AI industrial cycle is entering its second half.

This is exactly what this article intends to discuss. The materials mainly come from the "Macro Chat" column of the "High Energy" podcast and Feng's recent sharing, and focus on three questions: At what point in the cycle is the global capital market standing now? What is the coordinate of the AI industry at this position? Where will opportunities grow after the inflection point?

01

How did this cycle begin?

The answer lies not in technology, but in currency

To judge where a cycle has gone, let's first briefly review where it came from.

The starting point of this AI boom at the financial level is not ChatGPT in November 2022, but the global liquidity injection in 2020.

In 2019, the ratio of the total market value of global stocks to global GDP was still around 100%, which was in the reasonable range of the "Buffett Indicator"; in 2020, in response to the epidemic, the United States launched unlimited quantitative easing, and the central banks of major global economies printed more than 10 trillion US dollars of base money within a year or so. After this money was multiplied 3 to 4 times through the banking system, it was equivalent to an extra dozen percent of the total global currency in circulation. This has almost no precedent in financial history. Global stock markets and the primary market then generally rose from the second half of 2020 to 2021.

This sum of money was concentrated in dollar assets in 2022: the Russia-Ukraine conflict made Europe a difficult allocation target, China faced uncertainties such as the epidemic, and two of the three major economic regions that account for nearly two-thirds of global GDP were inaccessible, so massive funds poured into dollar assets.

At the same time, the US dollar interest rate hike pushed up the appreciation expectation of the US dollar and dollar-denominated assets: in March 2022, the Federal Reserve raised interest rates by 25 basis points for the first time, and then raised interest rates by 200 basis points continuously from May to July. In November of the same year, the emergence of ChatGPT provided an even better reason for the rise of asset prices. From 2023 to 2025, global funds continued to increase their allocation to dollar assets, driving a rising market value of increasing scale.

Let's look at a set of data: the global stock market value rose from 89 trillion US dollars in 2019 to 148 trillion US dollars in 2025, and the proportion of US stock market value in the global stock market value rose from 38% in 2019 to 47% in 2025. (For a detailed account of the origin of this financial cycle, please read "Frees Fund Li Feng's 2025 Year-end Sharing: Logic and Prospect of AI Investment")

02

Why do we say we are at the tail end of the cycle?

Signs of stock game

This capital cycle can be divided into three stages: before the end of 2024, it was the first stage, when the total market value of the global capital market rose generally; starting from mid-2025, it entered the second stage, that is, the stock game — inflation has become a common phenomenon around the world, no one can print money on a large scale anymore, so when the Nasdaq rises, the Dow Jones does not rise, when the Dow Jones rises, the Nasdaq does not rise, and the A-share market is more or less the same; the third stage is the reallocation of funds.

To judge where this dollar-dominated financial cycle has gone and whether liquidity has peaked, we follow a simple logic: for any major asset to keep rising in price, more and more incremental funds need to flow in, because the higher the price, the more money is needed to push it up one level; once there is not enough new money, the market can no longer rise generally, and will turn into a rotation of "A rises while B falls, A falls while B rises".

At present, global liquidity should be close to its peak, and new funds are starting to run out.

At the same time, the number of people who have already made money and want to take profits and leave is also increasing. Several signs in the last month point to the same direction: the bond yields of major developed economies such as Europe, the United States, Japan, South Korea and the United Kingdom have soared at the same time (rising yields mean that bonds are being sold off), the United States has intervened in the yen exchange rate, and no matter how good the financial reports US tech giants deliver, their stock prices can hardly rise anymore.

A set of figures can illustrate how much money this round of rise has consumed.

The total market value of US stocks rose from 34 trillion US dollars before the 2019 epidemic to about 75 trillion US dollars today. Considering that not all chips are flowing at all times, the extra 40 trillion US dollars of market value may require a net inflow of five to eight trillion US dollars to support. Bonds are different from stocks, every bond has a real sum of money buying behind it: during the same period, the global debt scale grew from about 250 trillion US dollars to about 350 trillion US dollars, an increase of about 100 trillion US dollars; among them, US national debt grew from about 23 trillion to about 40 trillion US dollars, corresponding to a cash inflow of 17 trillion US dollars.

When the total amount of money no longer grows, two mentalities will appear in the market at the same time: fear of high prices, and chasing hot spots. Then the same amount of money starts to "jump from this pit to that pit", the indices no longer rise generally, and the speed and amplitude of rotation between sectors are increasing. This kind of rotation even occurs within the semiconductor industry: CPU rises for a while, storage rises for a while, and GPU rises for a while. The intensification of rotation indicates that the stock game has been fully launched; and the depletion of incremental funds to be discussed next indicates that this stage is coming to an end.

03

No new money to "draw":

America's ledger and the "Fragile Triangle"

The root of the stock game is that the source of incremental funds for dollar assets is being exhausted.

Let's first look at America's own ledger. The federal fiscal revenue is about 5 trillion US dollars. Calculated based on the current interest rate level, the net interest expenditure on national debt for the whole year is about 1 trillion US dollars. Coupled with the national defense expenditure of about 0.9 trillion US dollars, these two items alone account for nearly 40% of the fiscal revenue. The deficit is expected to continue to expand, so US national debt can only be issued more and more — the US Treasury Department expects that more than 700 billion US dollars of national debt will be issued from June to September 2026.

But the number of buyers is decreasing. In the past two years, the central banks of many countries including China have continued to increase their gold holdings and reduce the proportion of US national debt in their reserves. Japan, the remaining major buyer, is too preoccupied with the yen falling to its lowest point in 40 years. The United States claims to help Japan stabilize the exchange rate, but the exchange rate stabilization position that the Treasury can use without congressional approval is only a few hundred billion dollars, which is essentially managing expectations and has limited actual effect.

With more supply and fewer buyers, the interest rate required by the market is naturally high: currently, the 10-year yield of US national debt has risen to around 5%, and the 30-year yield has broken through 5.2%, approaching or reaching the highest level since 2007.

Thus, a "sensitive and fragile triangle" has formed: the United States needs to issue national debt on a large scale; the winning bid rate of US national debt cannot be too high; at the same time, Japan cannot trigger a liquidity explosion by not buying or selling off.

The hardest problem at the tail end of the cycle is the lack of money. For the Federal Reserve, raising interest rates will increase the debt service cost, while cutting interest rates will expose weakness. The US Treasury Secretary and the Fed Chairman are trying their best to maintain the "moderately strong dollar" path, which is thinner than a steel wire, but there are not many usable means left: one is to make hawkish remarks and manage expectations, and the other is to collect the benefits of safe-haven funds flowing back to the US dollar when global risks rise.

A small cycle from March to June 2026 just verified this mechanism and its limits.

After the outbreak of the Israel-Iran conflict at the end of February, global funds first allocated cash in panic, then allocated US dollars and US national debt. From April to May, the US dollar index rose rapidly, and the US stock storage chip sector rose sharply at the same time. This crisis objectively made global funds concentrate on dollar assets for another round. But this time the amount that can be drawn is not much. After it can no longer be drawn, both US national debt and US stocks have entered a wide range of fluctuations. Even if the leverage of Japan and South Korea is blown out, the funds that can be drawn away are not enough to drive the next round of rise relative to the 75 trillion US dollars of US stocks.

To make matters worse, these limited funds are being pulled by several parties at the same time: US national debt needs liquidity support, which is the first priority; to push the 75 trillion US dollars of US stocks upward again, continuous capital inflow is needed; the financing demand of AI data centers is also expanding rapidly, and the scale of private debt is approaching the trillion-dollar level.

Other major global economies are also preoccupied: the European Central Bank has restarted interest rate hikes, the Bank of Japan is also slowly tightening and yen funds are flowing back to the mainland, and neither can deliver large-scale liquidity to the global capital market as in the past.

04

What will happen at the tail end?

Three black swans and two observation indicators

In the case of high valuation and marginal tightening of liquidity, black swans often do not start from the deterioration of the fundamentals of specific companies, but are more likely to be triggered by external liquidity and geopolitical shocks.

The first black swan is the Japanese yen.

For more than a year, Japan has been the main provider of dollar liquidity. Japan itself is one of the largest overseas holders of US national debt. More critically, there is the "carry trade": Japan has long maintained near-zero interest rates and no capital controls, and global speculative funds borrow yen at low interest rates from Japan, exchange it into US dollars to buy US national debt and US stocks, which has been a major source of liquidity for dollar assets in the past year.

Most of the carry trades are leveraged. As Japan continues to raise interest rates, these trades may be liquidated on a large scale; Japanese insurance institutions and other institutions will also concentrate on selling dollar assets and flow back to the mainland to enjoy the benefits of yen appreciation.

The second black swan is data center private debt.

Companies like Apple and Google that have hundreds of billions of dollars in cash on their books were originally important buyers of US national debt. But in the second quarter of this year, except for Microsoft, which has the most stable cash flow, Internet and tech giants such as Google, Amazon, and Oracle all turned from buyers of US national debt to parties issuing their own bonds to grab liquidity due to excessive capital expenditure, and their free cash flow turned negative. Oracle has issued the most bonds, issuing 43 billion US dollars of bonds last year and expecting to raise another 20 billion US dollars this year, and the trading interest rate of its corporate debt has continued to rise.

Once such private debt blows up, it may trigger a chain sell-off of the entire asset class.

The third black swan is the Strait of Hormuz.

The "partial reopening" window brought by the June memo has been closed, and the strait has now returned to a near-shutdown state, while the stress resistance of the global energy and commodity market today is far weaker than two and a half months ago. If the blockade continues, commodity prices will face greater upward pressure (on September 9, Brent crude oil futures broke through $100 per barrel for the first time since July 24), and the entire market may be dragged down.

Even if it does not directly affect the United States, it will affect the US supply chain spread all over the world. According to industry estimates, China's strategic oil reserve capacity has expanded to four months, and the development of new energy has made energy sources more diverse (for example, data centers mostly use green electricity), which is relatively safe; Japan, South Korea and other countries are hard to say.

There are two trackable indicators to judge whether the cycle has reversed.

The first indicator is "only look at the number one": as long as the stock price of the leading company (such as Cisco in 2000) can still hit a new high, the capital cycle is still upward; once the stock price of the number one can no longer hit a new high, it may have entered a downward cycle.

The second indicator is the seesaw of gold and the US dollar: when the market expects the US dollar to strengthen, funds will prefer to go to the US dollar with interest and high liquidity; when the US dollar shows weakness and risk concerns still exist, funds will turn to gold.

At the tail end of the cycle, the hardest money to earn is the last wave of money.

When major assets enter a stage of high volatility or decline, it is difficult to expect a sub-category to rise rapidly against the trend, just like when real estate generally fell in the past four years, there was almost no commercial real estate that could rise independently for a long time. But peaking does not mean falling immediately. Major assets will continue to fluctuate widely until one day liquidity really reverses.

As for where and when the black swan will fly out, it is really hard to predict.

05

A longer perspective:

One and a half bubbles

Pull the camera further away, this financial cycle has an even larger background.

After the US dollar was decoupled from gold in 1971, there was no longer a hard constraint on central bank money printing, and central banks of various countries have become more and more aggressive in using balance sheet expansion to intervene in crises. There is a book called *The Hand of Currency*, which describes exactly how global central bank governors use their "magic hands" to create liquidity. Whenever there are ebbs and flows, they will take measures.

For the quantitative easing (QE) after the 2008 global financial crisis, the United States divided it into three and a half steps and tested it gradually over three and a half years, basically printing a little more when it was not enough, which was relatively prudent. 2020 is different. It started with cutting interest rates to zero plus unlimited QE, and completed the total QE of the previous three and a half years in seven or eight months.

The money printed in 2008 should have been gradually cleared out in the interest rate hike and balance sheet reduction cycle that began in 2017, but the Trump administration's intervention in the Fed's policy interrupted this contraction process. Then in 2020, the epidemic was superimposed with a round of larger-scale liquidity injection, resulting in at least one and a half bubbles piled up together. The consequences are: global inflation is difficult to suppress except in China; asset prices are high (the 75 trillion US dollars of US stock market value corresponds to only more than 30 trillion US dollars of GDP); and the gap between the rich and the poor has widened