Summary

Global liquidity is peaking: there is not enough new money. Li Feng says bond yields are surging across the major developed economies of the US, Europe, Japan and the UK, while global equities are reaching the point where “no matter how good your earnings report is, the stock probably still cannot rise.” Treasuries (the first priority), US equities—with a market cap above $70T and more than 2.3x GDP—and debt-financed AI data centers are each competing for trillion-dollar pools of capital. “Every additional leg up requires more money than before, and more risk.”
After Nvidia rose 8% on earnings, Li Feng bought a small unlevered short, calling it purely a “liquidity experiment.” His five reasons: peak liquidity is pressuring high-risk assets; whether the leading companies can make new highs is a left- or right-side indicator of the capital cycle, as Cisco showed in 2000; data-center delays, price increases and other headwinds are converging in Q3; Nvidia is aggressively pushing a rent-rather-than-buy model used by Cisco and Nortel near the end of the bubble; and Barron’s pointed to another sequential halving in free cash flow alongside elevated receivables. He stressed that none of this means he is bearish on AI technology itself: “This is all personal speculation and has nothing to do with investment advice.”
The capital-markets regime for AI capex has reversed: from rewarding whoever was most aggressive with the highest valuation to punishing whoever was most aggressive. The US tech megacaps and other large companies announced aggressive capex in Q1, while the entire supply chain spent most aggressively in a FOMO-driven Q2. By earnings season, companies that had burned through their free cash flow were being discounted—“even Google could not escape this fate,” nor Alibaba after its equity issuance. “PE is not very meaningful for highly capital-intensive cyclical industries”; single-digit PE ratios are often a sign that the cycle is near its peak.
The dollar is walking a tightrope: it cannot hike or cut rates, leaving hawkish rhetoric or manufactured risk as the only ways to steer safe-haven flows back. In the short term, gold and the dollar trade like a seesaw: when expectations for the dollar strengthen, money flows into the dollar as an interest-bearing safe haven; when the dollar weakens while risk concerns rise, money moves into gold. Gold pays no interest—it can only wait for prices to rise and protect against declines—which is why Buffett does not regard it as an asset.
China has used the AI and capital cycles to push all five categories of economic actors toward direct financing in barely 4 years. Loans’ share of aggregate social financing has fallen from about 70% 5 years ago to 60%, while direct financing has risen to slightly above 30%; Li Feng expects the structure to move toward “something like 50/40.” Among the companies generating the biggest wealth effects—Muxi, Moore Threads, the two large-model companies, ChangXin, and Yangtze Memory—nearly all of the wealth effect, except at Unitree, has accrued mainly to state capital. That has pulled market-based capital back into the market after it did not invest, did not want to invest or did not dare invest from 2023 through the first half of 2025. High-net-worth individuals are even asking on WeChat, “Do you have an allocation in DeepSeek?”
The essence of the new property policy is to turn consumers back from financing counterparties into pure consumers, and it may dismantle the old chain as home prices approach a bottoming zone. Under the presale model, households were pre-levered and made to serve as the financing link in an investment-driven cycle. The new policy pushes risk back onto the entire development chain and its financing institutions. Reduced new-home supply could accelerate the bottoming process, stabilize the denominator of household balance sheets, and eventually create room for consumption and a shift from housing—which accounts for 70% of assets—toward financial assets.
China’s A-share market is building its ownership structure first and talking about a slow bull market second. Insurance capital is absorbing some of the blue-chip inventory transferred out during the national team’s stabilization operations; nearly all of Kweichow Moutai’s new top-10 shareholders in Q2 were insurers. About RMB5T in equity mutual funds form a semi-stabilization layer, while more than a dozen newly approved actively managed index funds give the stabilization apparatus the ability to target specific exposures. Insurance rules taking effect in 2027 will require “long money in long investments, medium money in medium investments, short money in short investments,” potentially making insurers the financial-structure converter for households shifting their asset allocation.
Pandemic-era liquidity has not been fully cleared, leaving the world with “two swords of Damocles” overhead. Trump interrupted the rate-hiking cycle, and the unpurged bubble was compounded by the full-scale 2020 stimulus, delivered in 7 months and equivalent to roughly 3.5 years of post-financial-crisis quantitative easing—“this time it is at least one and a half bubbles.” Clearing it would require a crisis to “destroy wealth once,” but central banks have learned to print money faster to protect markets. Li Feng sees stagflation persisting globally in some form for a long time unless the AI narrative delivers and massively expands supply, while the populist backlash created by widening inequality has changed social structures in ways that are “almost impossible to reverse.”

Deep dive

1. Hong Kong IPO break rates are a liquidity thermometer: 50/50 is healthy

Li Xiang opened with the previous day’s Hong Kong IPOs: one tied to industrial robots and one to cross-border e-commerce had both broken issue price, leaving a friend with losses from IPO subscriptions. Li Feng’s framework: when Hong Kong had “not enough bells to ring” in the first half and 3 companies were listing in a single day, the first-day break rate was only in the teens. That meant huge amounts of short-term speculative and arbitrage capital were concentrated in IPOs—“if you can make several dozen percent by trading for a month or 2 weeks, nobody wants to wait a year to make 10%.”
He has been tracking the metric for 1.5 months: only when the break rate returns to a normal “at least 50/50” does it indicate money is moving from the short term back to the long term. Under a fully market-based registration system, IPO gains and losses should be close to a coin toss. “You cannot guarantee that IPO subscriptions will have a high probability of making money, because if you could, wouldn’t all the money go buy new shares?”
Hong Kong listings have fallen sharply over the past month, with only these 2 companies listing during a roughly 2-week gap. The supply constraint may help the market recover, though Li Feng acknowledged it could also reflect weak sentiment. Combined with the trouble at a roughly RMB1B fund that speculated on IPO cornerstone allocations—investors could not get their money out—he suspects this may be a correction signal, with possible gray arrangements around allocations, locked-up shares and the trading range. But the fact that both companies still broke issue price despite being the only 2 listings shows that “there is still a liquidity problem.”

2. Global liquidity is peaking: there is not enough new money

Before recording, Li Feng described the mechanism as “stormy”: every market rally requires new money, while profit-taking positions continue to pile up. “Every additional leg up requires more money than before, and more risk.”
Global signals have pointed in the same direction over the past month: bonds in the major developed economies of the US, Europe, Japan and the UK have been sold off and yields have surged. Global equities—whether core themes, fringe themes or unrelated companies—are reaching the point where “no matter how good your earnings report is, the stock probably still cannot rise.”

3. Nvidia rose 8% after earnings; Li Feng bought an unlevered short as an experiment

Nvidia rose 8% after earnings last Thursday, and Li Feng bought a very small short position without leverage. Unlike the daily-reset leveraged products linked to SK Hynix and Samsung, where volatility creates decay, an unlevered short has roughly symmetrical gains and losses. “It was purely a liquidity experiment.”
The first logic is that peak liquidity puts pressure on high-risk assets, with US equities having been among the biggest absorbers of the previous liquidity wave; “this has nothing to do with whether you are bearish on AI itself.” The second is historical: look only at the leader, as with Cisco in 2000. Whether the leading company can make a new high indicates whether the capital cycle remains on the left side; it cannot accurately identify the top, so he bought only a tiny amount “as a social experiment.”

4. The AI-capex trilogy: announced in Q1, FOMO spending in Q2, punished in earnings season

Before Q1, the market rewarded whoever was aggressive with the highest valuation. The US tech megacaps, major global companies and mid-sized internet companies therefore announced aggressive AI capex in Q1. Q2 was “the most FOMO-driven quarter”: the entire supply chain believed prices would inevitably rise sharply and rushed to stockpile inventory and spend. After extreme volatility in July, the tone reversed during earnings season. Companies whose capex exceeded free cash flow and burned through that cash flow were discounted—“even Google could not escape this fate,” nor Alibaba after its recent equity issuance. “By the time everyone reported their half-year results, the rule had become: punish whoever was aggressive.”
Nvidia’s gross margin being least affected by price increases in Q2 was a coincidence: it was selling inventory that had been stocked earlier. The consumer market offers a parallel. Apple first announced a RMB1,000 price increase in June and July, after which domestic Android brands followed. The price increase only began to flow through after the inventory was consumed.

5. Q3 headwinds are converging: data centers cannot be built, prices are rising, and the market is no longer rewarding spending

The US is less capable than China at building infrastructure: environmental reviews are stricter, state approval procedures are more involved, the power grid is inadequate, developers must build supporting generation, and nearby residents may need to approve projects or vote on them. The many data centers announced at the start of the year have faced varying degrees of disruption in speed, progress and scale since mid-Q2.
Li Feng’s simplified version: “At the start of the year they said they would build 100 data centers. In Q2, fearing price increases and driven by FOMO, they spent the money aggressively. Then they found that many projects would be delayed or could not be built immediately. In Q3, they found that capital markets did not reward aggressive spending, and the impact of price increases was beginning to show.” He repeatedly stressed that this does not mean growth is over; “I am only saying that all the adverse factors are starting to be priced in.”

6. Rent rather than buy, and Barron’s question: where did the cash flow go?

Nvidia has recently been aggressively pushing a rent-rather-than-buy model: helping customers obtain credit enhancement and loans, or using off-balance-sheet arrangements and guaranteeing loans in its own name to build infrastructure, while recognizing the revenue first. Li Feng’s historical comparison is that Cisco and Nortel used similar methods around the collapse of the internet bubble.
A Barron’s article asked where the cash flow behind Nvidia’s strong earnings had gone. Free cash flow had fallen by another half sequentially, while receivables were far above historical levels. Li Feng suspects the two may be related to the rent-rather-than-buy model: “Perhaps Nvidia has reached a point where it needs to use rent rather than buy to drive revenue growth and meet capital-market expectations.”

7. Three parties are competing for trillions: Treasuries, equities and AI private credit

The competitors for liquidity are US Treasuries, the first priority because failure to issue debt would be most dangerous for the US; US equities, whose market cap has surpassed $70T and 2.3x GDP, requiring substantial capital to move higher; and AI data centers, which are financed mainly with debt and are also a trillion-dollar market. Every economy is trying to attract capital, while most countries other than China and Japan maintain high rates to prevent a dollar drain. “Everyone cannot provide significantly more money, needs money not to leave, and preferably wants to attract other people’s money.”
A small country cannot conjure $1T of liquidity out of nowhere; in China, the same problem would be measured in the tens of trillions of yuan. “I also do not know what comes after the storm.”

8. The dollar’s tightrope: it cannot hike or cut, leaving rhetoric and manufactured risk

From the G20 finance ministers’ and central bankers’ meeting to Jackson Hole, the dollar must remain appropriately strong: once it weakens, money starts fleeing dollar assets. But rates cannot rise without breaking countless existing balances, and they certainly cannot be cut because that would reveal obvious weakness. Only 2 options remain: hawkish rhetoric, or “manufacturing risk—once panic is created, that money will flow back for safety.”
Gold and the dollar trade like a seesaw in the short term. When expectations for the dollar strengthen, safe-haven money first goes to the dollar because it is an “interest-bearing safe haven” with high liquidity. If the dollar shows weakness while risk concerns rise, money may move into gold. Rising Treasury yields—especially if short-dated yields rise as well—may signal greater uncertainty about the long term. But gold pays no interest; it can only wait for prices to rise and protect against declines. That is why Buffett does not regard gold as an asset.

9. Do not use PE for cyclical stocks: single-digit multiples signal a top

Li Xiang asked whether Nvidia’s PE in the teens is high. Li Feng cited a colleague’s research on multiple storage cycles since the 1980s: highly capital-intensive cyclical industries are mainly valued on PB, while “PE is not very meaningful.” When PE reaches the single digits, especially the low single digits, “it is generally at the top.” In cyclical, high-capex industries, “when they look most profitable, that is basically when they are about to run into trouble.” The judgment applies mainly to highly capital-intensive cyclical industries; Li Feng said storage is more representative than chips.

10. Weakness in secondary markets is cooling primary markets: wealth stories shrink after lockups expire

Both primary and secondary markets are driven by incremental capital, but the secondary market “provides the wealth effect for the primary market.” Li Feng says AI, chips and robotics are beginning to cool. State capital has applied the brakes to varying degrees because of Document No. 54; the several-hundred-billion-yuan venture-capital fund of funds financed by ultra-long special government bonds still requires investment at the early-stage and small-company end. US-dollar funds invested aggressively in the first half; after catching up on their targets, they may also reconsider whether to remain so aggressive.
In the second half of 2026, some Hong Kong companies that listed in 2025 will enter their lockup-release periods. “At listing, every story was worth hundreds of billions of Hong Kong dollars; once the lockup expires, the market caps become tens of billions or barely over RMB10B.” The well-known industrial-robot company that listed yesterday was worth just over HK$10B after breaking issue price. After subtracting its listing proceeds—roughly RMB8B—the wealth effect was “not that obvious” relative to its last-round valuation of more than RMB6B. As this keeps happening, the message will continue to flow back to primary markets.

11. China’s financial structure is undergoing a major rotation: from 70/20 to 60/30, moving toward 50/40

July financial data showed bank loans at 60% of new aggregate social financing, direct financing—including government bonds, corporate bonds, and listed-company offerings—at slightly above 30%, and bills and trusts in the single digits. Five years ago, the structure was roughly 70/20. In the US, more than a century of capital markets came before the central bank, and direct financing accounts for more than 80%. Li Feng expects China to become “something like 50/40.”
The significance is that loans most readily support capital-intensive, investment-led sectors—Infrastructure, property and manufacturing capacity. The model of using indirect financing to drive investment and GDP growth “should already be undergoing a complete transformation.”

12. State capital was the source of the wealth effect that brought market-based money back

China’s biggest wealth stories over the past 2 years were the 2 large-model companies, the 2 major chip companies Muxi and Moore Threads, ChangXin, the soon-to-list Yangtze Memory, and, to a lesser extent, Unitree. Excluding Unitree, the wealth effect at the other 6 companies “was almost entirely in state capital,” because state capital happened to invest in 2023, 2024 and the first half of 2025 when market-based capital did not like the opportunities, did not invest or did not dare invest.
Li Feng described the psychological mechanism directly: “A lot of people thought they were pretty capable, but then you made the most money—well, if you can make money, surely I can too.” Capital that had stayed out, did not want to invest or did not dare invest for various reasons therefore came back and copied the state-capital playbook. In response to Li Xiang’s question, he said most of this money had been raised in 2020-2022 and typically had a 4-7-year investment period. There were essentially no exits in 2022, and more precisely, the situation continued through 2025.

13. All five categories of economic actors are shifting: from the government to “do you have an allocation in DeepSeek?”

Li Feng reviewed the 5 layers of market participants one by one. The government is moving from land finance toward equity finance. Financial institutions, including banks, are shifting into direct investment. Many listed companies whose core businesses have nothing to do with technology are setting up funds and investing directly in hot sectors. Small and micro businesses are themselves entrepreneurs and participants in the economic transition. The individual side is the most vivid.
Among middle- and high-net-worth individuals, dedicated funds proliferated when hot sectors raised more than RMB10B in the first half. Over the past 1.5 months, people have been contacting him on WeChat to ask, “Do you have an allocation in DeepSeek?” Li Feng stopped establishing dedicated funds after 2022. He had previously launched several and exited profitably, but “with only one project, there is no asset-allocation logic: if it works, it works; if it does not, you lose.” The Asset Management Association now restricts dedicated funds during registration because the strategy resembles retail investors chasing gains and selling losses and “could very easily end in a mess.” Li Xiang questioned the need for ultra-high-net-worth individuals to pay back taxes within 3 months. Li Feng said he had not studied it closely, but had heard that people who placed equity rather than cash into trusts years ago could face problems: the shares may have fallen, taxes are calculated at the original point in time, and there may not be enough cash.

14. The question raised by Qiushi: rebalance and repair household balance sheets

July’s micro data showed household loans and household deposits both declining, without a corresponding full increase in deposits at non-bank financial institutions. Li Feng agreed with a Goldman Sachs analysis that China faces an asset-allocation shift among households lasting more than 10 years and involving tens of trillions of yuan. “That is completely consistent with the shift in financial structure.”
The decline in household loans is not only a sign of weak consumption. The larger factor is that when 2- or 3-year term deposits matured, households first prepaid commercial loans carrying rates of 4%, 5% or even above 6%. This is consistent with the PBOC governor’s statement in June that incremental loans in aggregate social financing would no longer maintain the high growth rates of the past, as well as Qiushi’s first explicit reference to “the rebalancing and repair of household balance sheets.”

15. Debt-resolution-style repair: eliminate high-interest lending and use subsidies to refinance mismatches

The first move is to cut off high-interest lending such as internet-assisted loans and multiple borrowing, removing in one stroke the latent risks that could be delayed inside banks’ personal and consumer-loan books. Li Feng compared it with the 1999-2000 use of AMCs to write off banks’ bad debts in one go, because banks are about to become the main financial support for a new phase of construction.
The second step is for roughly RMB100B of targeted Ministry of Finance funds to be ultimately applied as interest subsidies to consumer loans, small-business technology loans and technology bonds. Two episodes earlier, they had speculated about whether the funds would subsidize mortgages; they did not. The logic is the same as government debt resolution: “First cut out the problematic part; then replace the duration mismatch and interest-rate mismatch.” Borrowing at a market rate of 10% to fund a project that takes 30 years to break even and produces only a 2%-3% annualized return is a classic mismatch.

16. The real meaning of the property policy: turn consumers back into pure consumers

Li Feng’s reading of last week’s property notice is that under the presale system, households leveraged up during the forward-home stage and “became the main link that was pre-levered and pre-positioned in the circulation of funds,” serving as the financing counterparty in an investment-driven GDP cycle. Once that cycle broke, unfinished projects had to be rescued by the government’s guaranteed-delivery program. The new policy pushes the entire chain back toward market mechanisms: after buying a completed home, borrowing is simply a consumer loan, while risk is reset across the full property-development chain, including financing institutions.
Timing is critical. Li Feng believes this kind of clearing and complete separation of responsibilities is possible only if property prices are near a bottom; otherwise it could trigger panic. Lower new-home supply may also accelerate the stabilization process and drive second-hand sales, stabilizing the denominator of household balance sheets. That would avoid the negative spiral in which leverage falls but asset prices spiral lower, preventing the asset-liability ratio from improving.

17. Li Feng does not think developers will ultimately be crushed: consolidation will follow the coal-and-steel model

On research showing property developers’ IRR falling to 4%, Li Feng’s interpretation is that after a market-based rebalancing, developers that remain viable “will have somewhat higher gross margins.” At the same time, financing capacity and all-in costs will matter more, potentially giving state-owned property companies a larger advantage.
The path resembles coal and steel. When safety regulation tightened during a downcycle in commodities, private owners exited, state capital took over, and the industries were restructured and consolidated until they survived the cycle. “Exactly the same thing happened in Japan after the 1990s.” Once household balance sheets are repaired, 2 long-term pools of demand need to open up: consumption, and an asset-allocation shift from housing—which accounts for 70%—toward financial assets. “It may not ultimately exceed housing in China’s culture, but it will certainly rise materially from just above 10% today.”

18. Li Xiang’s question: what if investable companies have weak earnings? The answer is ownership structure

Li Xiang asked whether households should be directed to the secondary market once property is no longer the engine of wealth growth, given that the companies available to buy have relatively weak earnings power. Li Feng’s immediate response was: “How could that be?” China is building a tiered ownership structure. From 2023 to 2024, equity mutual funds grew from RMB1T-RMB2T to roughly RMB5T, forming one stabilization layer; China Securities Finance and Central Huijin form another. In Q1 and Q2, the national team did not sell all its holdings while stabilizing the index. Some of the blue-chip inventory it transferred out was absorbed by insurers. Among Kweichow Moutai’s top-10 shareholders in Q2, the national team exited and nearly all the new entrants were insurers.
Insurers do not participate in the monthly and weekly ranking competition of public and private funds, so they are willing to buy blue chips with dividend yields above 4%, limited growth and stable earnings. The ideal 4-layer structure is: insurance long-duration capital as the least frequently traded “base inventory”; public index funds as the second, semi-stabilization layer; China Securities Finance and Central Huijin as the third layer dedicated to stabilization; and retail investors, private funds and actively managed public funds at the top, trading most frequently. Daily A-share turnover has grown from a few hundred billion yuan to the point where even RMB2T is considered relatively light; the market’s capacity is entirely different.
Li Feng’s summary: “A slow bull market is not slow and bullish by nature. It first needs to build an ownership structure on which a slow bull market can rest.” Policies urging listed companies to return capital to shareholders and increase dividends are part of the same orderly, gradual process.

19. Why were more than a dozen active ETFs approved so quickly? To put a scope on stabilization

Li Feng’s speculation—explicitly his own—is that the difficulty of stabilizing the market through broad-based funds became clear in Q2. The index was rapidly lifted by a very small number of companies that were not major constituents of the broad-based funds; selling the broad-based funds could therefore put greater pressure on existing investors. After the Lujiazui Forum announced the push for actively managed index funds, more than a dozen were approved almost immediately, allowing stabilization operations to target specific exposures.
His hypothetical example was a fund focused solely on AI hardware: it could buy during a downside cycle after a bubble burst and provide support, then sell only that exposure to stabilize the market during the next overheating phase. The example was intended only to illustrate the allocation method that active ETFs might provide, not to express a view on any specific product.

20. Insurance is the “financial-structure converter” for household balance sheets

Insurance rules released 2 weeks ago and taking effect on January 1, 2027 require duration, return and liquidity matching between liabilities and assets—“roughly similar to the 2018 asset-management rules for banks.” Li Feng translates the requirement as: “long money in long investments, medium money in medium investments, short money in short investments,” while maintaining highly liquid stable assets to meet redemptions. Funds cannot all be allocated to equities, or forced selling could hit the stock market when liquidity is needed.
Why must insurers allocate to equities, currently up to 30%? Dividend insurance products that sold heavily in the first half typically promise returns in the 2%-plus range, while bonds alone may guarantee only around 1.8%, requiring higher-return assets such as equities to fill the gap. Investment returns were an important contributor to revenue and profit growth in the first-half reports of major insurers, mirroring how brokerages made money through STAR Market co-investments and direct investments in major chips, large models, ChangXin and Yangtze Memory. There is also an existing interest-rate spread loss: high-yield policies sold when term-deposit rates were above 3% in 2022-2023 are gradually maturing, and current investment returns must cover the gap between the promised returns and today’s asset yields.
From 2025 through the first half of 2027, RMB50T of high-interest term deposits will mature. Insurers may be able to absorb and convert more of those household deposits, taking on the role of a “financial-structure converter” for household balance sheets—just as banks and trusts once converted savings into indirect financing for property and infrastructure. In the US, insurance plus 401(k) and similar personal accounts may account for more than 10% of household balance sheets.

21. Is the problem risk appetite or a lack of investable assets? Li Feng: assets will emerge

Li Xiang asked whether households avoid financial assets because of low risk appetite or because there are no assets to buy. Li Feng first acknowledged the current conservatism: when term deposits mature, households repay commercial loans first—household leverage has fallen by nearly 2 percentage points—then roll deposits, buy bank wealth-management products yielding around 2%, and move into dividend insurance. “Risk appetite is still gradually shifting up from a very low level. It will certainly not happen quickly.”
On the asset side, he gave 3 examples of things that went from impossible to real. An LP in Fujian’s footwear and apparel industry recalled that at China’s WTO accession, “fear was greater than opportunity” and it seemed they would “certainly be wiped out”; the sector ultimately moved from small workshops to proprietary brands. Three years after China’s pharmaceutical industry began licensing products overseas, it reached roughly two-thirds of global drug-R&D licensing value—like the stage Lixun Precision occupied before proprietary brands emerged. Li Bin went from “the most miserable man in China” in 2019 to the luckiest man in China, with only 2 years in between. China’s annual NEV sales rose from just over 1M to more than 10M.
The other side of all 3 stories is domestic demand. The development of phones and NEVs benefited substantially from the domestic market; the missing final link for pharmaceuticals is also domestic demand, which brings the discussion back to financial-structure conversion. “Basic coverage comes from public medical insurance; consumption upgrades require commercial insurance.” Access to world-class treatments may require commercial coverage, providing another reason for households to buy it.

22. Where did the pandemic money go? One and a half bubbles and two swords of Damocles

Li Xiang’s final question was where all the money printed during the pandemic had gone and why liquidity was peaking again. Li Feng began with 1971, when the dollar left gold: fiat money “no longer had an anchor,” and central banks became increasingly comfortable intervening directly in the economy and financial crises. Volcker used a policy rate in the low double digits to bring inflation back to 2%; Li Feng calls him “the last Fed chair who strictly observed discipline.” After that, most crises were handled with quantitative easing. After 2008, China used RMB4T as the emblematic example and took roughly 3.5 years to execute 3.5 rounds of easing, which Li Feng considers relatively cautious. After the March 2020 crisis, rates were cut to zero and unlimited QE was launched; in only 7 months, the US completed the equivalent of the roughly 3.5 years of QE deployed after the previous financial crisis. He also recommended a book by a Belgian finance minister describing how central bankers use a “mysterious and highly technical” hand to create liquidity.
Central banks developed a path dependence: “At first everyone was feeling their way across the river; later, whenever there was a rise or fall, they immediately reached for the tools; eventually, they used the tools without restraint,” because “the previous problem had never been fully eliminated.” Liquidity can be cleared only by destroying wealth once—destroying money once and waiting for the next crisis to make a large portion of it disappear—but central banks learned to print more when crises arrived. After Trump’s first term brought a dispute with Powell and interrupted the rate-hiking cycle, the unpurged bubble was compounded by the full-scale 2020 stimulus. “This time it is at least one and a half bubbles.” Outside China, inflation has become difficult to suppress globally; asset prices remain high, inequality has intensified, and Li Feng links this to the populism visible today.
If the bubble bursts, theory says the market should lose 1.5 bubbles’ worth of value, but neither central banks nor presidents can allow that. They may respond with even more QE: “Stagflation will exist in some form around the world for a long time, until you eliminate these bubbles.” The potential escape route is a massive technological narrative such as AI; Musk’s logic is that once supply becomes abundant enough, “money will no longer matter.” In the earlier era of gold constraints, major wars could sometimes eliminate bubbles, but “that now seems very unlikely.” Li Feng’s closing image is “two swords of Damocles”: one is the bubble accumulation of at least 1.5 bubbles; the other is the social division and inequality that are almost irreversible. “I do not know where this will ultimately take the entire world.”

Inside the Factory Building America’s Autonomous Navy Town's JD on the AI assistant race: agent moats, $75K/engineer