Global Liquidity Peaks: The More Aggressively You Poured into AI in H1, the Sooner You Get Discounted
A stock market rally needs fresh money to keep buying. Global liquidity is already showing signs of peaking, and the companies that most aggressively invested in AI in the first half are shifting from being rewarded to being discounted.
The video won't play here. Listen to the audio instead:
The argument · tap a timestamp to hear it
Low IPO break rate is not a good sign
Don't take the low break rate of Hong Kong IPOs as a sign of market strength. In the first half of this year, the break rate was only around 10-20%, far below the roughly 50-50 normal level under a market-driven mechanism. That means the participants are mostly short-term speculative arbitrage funds, not long-term investors. Recently, the supply of new listings has decreased, and some IPO-related funds have had problems, but the two newly listed companies still broke their issue price, showing that temporary supply-side controls haven't solved the root problem: Hong Kong also faces peaking liquidity.
The more aggressive the AI investment, the sooner the punishment
Market sentiment reversed within a quarter. Before Q1, the capital market rewarded aggression: the louder the AI capex talk, the higher the valuation. By the end of July, when half-year reports were released, any major internet company that had made aggressive capital expenditures in H1—especially those that burned through free cash flow—was discounted, and even Google was not exempt. Q2 was the quarter of maximum FOMO across the supply chain; everyone rushed to spend before prices rose, and inventories could still support gross margins. By Q3, inventories are depleted and the impact of price increases returns, so the negative factors will show up together.
Nvidia is closer to Cisco in 2000
Nvidia's earnings look good, but its cash flow is changing. Before the dot-com bubble burst in 2000, Cisco launched a 'rent-to-own' program to help customers finance purchases, recognizing future revenue as sales upfront. Nvidia is now also pushing rent-to-own. In its latest report, free cash flow fell by half quarter-over-quarter, and accounts receivable are far above historical levels—which is speculated to be partly related to the credit enhancement and payment structure from rent-to-own. So Nvidia may have reached a point where it needs more aggressive financial tactics to sustain market expectations, similar to Cisco's situation back then.
Primary market money comes from bullets fired three years ago
The answer lies in the timing of fund raises. The money that will be deployed aggressively in AI and robotics from H2 2025 to H1 2026 mainly comes from funds raised at the market peak in 2020-2022. From 2023 to 2025, they barely invested, until they saw state capital placing bets when no one dared and making the most money—the stories of CXMT and YMTC, the two big model companies, and the two big chip companies stimulated market-based funds. The mentality is: copy the state's playbook. So this is not a rational long-term increment but more like a delayed emotional release.
New property policy removes households from the financing chain
The key of recent property policy is not to rescue developers but to detach households from the old financing loop. In the past, buying off-plan housing meant households played the role of financing counterpart in the investment-driven chain: paying first, receiving the property later. The policy pushes the chain toward marketization, so buyers now get completed homes and only act as consumers, with risk reset to developers and financing institutions. The policy was introduced when housing prices were near the bottom, aiming to stabilize the denominator on the asset side first, avoiding panic during a chain dismantling amid a decline. Japan after the 1990s and China's coal and mining industry consolidation followed similar paths.
A slow bull is not a slogan but a chip structure
A slow bull doesn't come from shouting; it comes from layering market chips by duration: at the bottom are low-valuation, high-dividend blue chips held by insurers as long-term money that rarely trades; above that are public funds and index funds, functioning close to a price-stabilizing role; China Securities Finance and Central Huijin sit at a higher level to adjust the index; and at the top are retail investors, private funds, and active management money. New insurance rules require liabilities and assets to match in duration, yield, and liquidity, institutionally ensuring that long money invests long, medium money invests medium, and short money invests short—not allowing short-term household money to be put into ten-year illiquid assets.
Crises grow bigger with each bailout; at least one and a half bubbles
Since the dollar left gold in 1971, global central banks have become increasingly accustomed to directly intervening in crises with money printing, and each action is faster. The quantitative easing after 2008 was supposed to be cleared during the rate hike cycle, but it was interrupted by political intervention, and then the pandemic hit in 2020, so another larger wave of easing was piled on before the previous bubble had dissipated—today we have at least one and a half bubbles accumulated. Central bank governors have learned to print even more money the moment a crisis appears, and the endogenous result is that inflation, financial asset appreciation, and wealth inequality all intensify simultaneously. As long as bubbles are not allowed to clear, the world may face some form of prolonged stagflation.
In their own words · checked verbatim
The reason I say it's a bit shaky is that today the whole world has a problem of peaking liquidity.
我说有点风雨飘摇的原因 就是今天全球都有一点流动性见顶的问题了
The capital market still thinks whoever spends aggressively is impressive, but when half-year reports come out, it flips to punishing whoever is aggressive.
资本市场还是谁花钱激进 我就觉得谁厉害 然后到大家公布半年报的时候 就变成了谁激进惩罚谁
Perhaps Nvidia has also reached a point where, in some sense, it needs to rely on rent-to-own to drive revenue growth and meet capital market expectations.
也许英伟达也到了某种意义上的 需要借助以租代买来推动收入增长 达到资本市场预期
Everyone sees that you can pull off that story, so they think, 'I can definitely pull off that story too'—that's roughly the psychology.
大家一看说 你都能把这个故事这么搞一遍 那我也肯定能把这个故事这么搞一遍 从心理上大概就是这样
The so-called slow bull we always talk about is not because it is inherently slow and bullish; it needs to first build a chip structure on the basis of a slow bull.
就所谓我们老讲的慢牛 不是由于他本身既慢且牛 他要在慢牛的基础上 先做一个筹码结构出来
Before the bubble piled up last time had dissipated, an even bigger one was added on top, and it was added without hesitation.
上一次堆积起来的泡沫还没有消散的时候给他加上了一次更大的 而且是没有犹豫就加上去了
Figures
| Hong Kong IPO first-day break rate in H1 2025 | 10-20% | 0:33 |
| Total US stock market cap | over $70 trillion, about 2.3 times GDP | 19:10 |
| Bank loans as share of current social financing | 60% | 30:19 |
| Direct financing as share of current social financing | just over 30% | 30:19 |
| US direct financing as share of total financing | over 80%, indirect financing in the teens | 30:19 |
| Equity public fund scale (original text refers only to stocks) | from about 1-2 trillion to about 5 trillion | 57:33 |
| Upper limit for insurance funds' equity allocation ratio | already can be allocated up to 30% | 1:03:34 |
| Household time deposits maturing from 2025 to H1 2027 | 50 trillion | 1:07:36 |
| China's household sector debt ratio decline | down nearly 2 percentage points | 1:11:38 |
| Time taken to complete the previous three-and-a-half-year QE after March 2020 | seven months | 1:19:47 |
Glossary
- Capex
- Capital expenditure: money a company spends on long-term assets; here mainly refers to capital investment in data centers and computing power.
- FOMO
- Fear of missing out: the fear of missing profits others are making, prompting the entire supply chain to spend before prices rise.
- Mega 7
- The seven largest US tech companies by market cap; the show uses them to gauge whether the capital cycle is still turning upward.
How to listen
For investors tracking both US AI and China/Hong Kong asset allocation, as well as primary market practitioners judging the source and pace of hot money in 2025.
You can skip the case-by-case discussion of Hong Kong IPO deals; the main thread resumes at 8:11 with the Nvidia liquidity experiment.