Summary

Two rounds of risk-off failed to leave investors more cautious; instead, they pushed capital back into the most crowded, highest-valued AI and robotics trades in China and the US. In Q4 last year, money pulled out of AI over bubble concerns; in late February this year, the US-Israel-Iran war and Strait of Hormuz risk drove it into dollar cash. But when risk-on returned from late March to early April, both risks were ignored: Nvidia and other leaders made new highs, and Nvidia’s market cap rose to $5.5T. 李峰峰’s summary was that the market had “wiped out all the risks of the past.”
This growth outlook rests heavily on the same foundational block: AI capex must be delivered—and can only rise, never fall. Citing internal research, 李峰峰 said roughly half of expected US GDP growth and S&P 500 constituent growth comes from AI capex committed by large companies, transmitted layer by layer through data centers, chips and the extended supply chain. The Q1 beat also rested on low-cost legacy capacity, corporate FOMO, boards demanding an AI narrative and pre-buying ahead of chip price increases; any cutback at one link could propagate upward.
Capital markets are rewarding the companies moving fastest toward negative free cash flow, while AI’s commercial loop still has not been proven by token revenue. If current guidance is delivered, Amazon may already have capex above free cash flow, while Microsoft and other slower spenders may reach the same point by Q1 or Q2 next year at the latest. Current token pricing still does not cover the full cost of models and data centers; priced at true cost, AI might not be cheaper than human labor. 李翔 invoked “this time is truly different,” while 李峰峰’s view was that “in fact, every time is the same”—and he expects the answer may emerge within a year.
Hong Kong’s IPO boom, 1.6T of unlocking market value and tighter cross-border investing suggest regulators may be targeting not only capital outflows but also short-term profit effects. Hong Kong indexes have been broadly weak since Q4 last year, but IPO subscriptions and pre-unlock speculation can generate multiples of returns within 1 to several months—far ahead of the “hard-earned 15% a year” delivered by long-term capital—and drain existing liquidity. 李峰峰 speculated that restrictions related to Futu and Tiger Brokers, together with Hong Kong’s tougher account-opening requirements, may jointly target new and hard-to-trace hot money. 李翔’s more direct explanation was that onshore and offshore returns are at different levels, and regulators do not want capital to leave.
Trump’s China visit signaled a medium-term equilibrium in US-China trade, while the sequence of high-level visits to China led 李峰峰 to raise his estimate of a US-Iran ceasefire. Disclosed items include roughly $30B in measures, agricultural and Boeing purchases, guaranteed supply of engine parts and reciprocal cuts to some tariffs. If a visit to the US by China’s leader this autumn is ultimately confirmed, the two sides would need at least a framework lasting 1-2 years of Trump’s term. The visits by Iran’s foreign minister, Trump, Putin and the Pakistani side led 李峰峰 to speculate that China is mediating, but Iran will not give up its “most important card”—the Strait of Hormuz.
The contest over the petrodollar is expanding from energy settlement into clearing infrastructure and the US government’s funding costs. The episode linked Iraq, Libya, Venezuela, Iran and Russia’s attempts to move away from dollar settlement. 李峰峰 estimates that CIPS could process more than RMB300T this year; combining CIPS and SWIFT figures roughly, the RMB’s share of global transactions may already be close to 10%, ranking 3rd. Meanwhile, the 30-year Treasury auction yield has broken above 5%; together with roughly $38T of debt, $30T of GDP and an annual deficit near $2T, that puts America’s inability to live within its means and AI companies’ future high-interest debt issuance on the same risk chain.
China’s economy looks more like weak stabilization and weak recovery, with structural change more important to track than the rebound in aggregate activity. Secondhand home prices in major cities may have stabilized, but household medium- and long-term loans continued to fall even as transactions improved, suggesting a mix of move-up purchases, deleveraging and substitution by housing provident-fund loans. CPI turned positive while PPI rose faster, leaving midstream profits under pressure. Weak consumption is concentrated in autos, appliances and jewelry, while incremental imports and exports are coming more from AI capex-related electromechanical goods, digital products, chip equipment and automobiles. As the economy shifts toward services and higher-value industries, loans and aggregate social financing may become structurally less predictive of growth.

Deep dive

1. Hong Kong IPO Records Meet a 1.6T Unlocking Wall

李峰峰 said several early-stage investments made about 10 years ago have recently gone public in succession. Public subscriptions for 易飞科技 came close to 15,000x, while 锡智科技 set that year’s record for the biggest first-day gain in a Hong Kong IPO. Their short-term performance primarily reflects the market’s current sentiment structure.
At its current pace, Hong Kong could still rank 1st globally in IPO count and in the top 3 by proceeds. But if mega-IPOs such as SpaceX come to market, the US could retake 1st place in proceeds; AI model companies are also actively preparing to list.
The excitement sits alongside roughly 1.6T of market value coming off lockups in a concentrated wave from this month through year-end. New IPO supply, unlocking shares and short-term capital are all expanding at once, creating the main liquidity overhang for Hong Kong equities.

2. After Two Risk-Offs, Money Returned to the Riskiest Trades

The first risk-off wave came in Q3 and Q4 last year. Excessive valuations across the AI Magnificent Seven triggered a rotation from high to low; Nvidia fell a little more than 10%, while many other names dropped 20% or more and capital moved into traditional sectors such as Walmart.
The second wave came from the US-Israel-Iran war in late February this year. Risks around the Strait of Hormuz raised fears that the war would not end quickly and that supply chains and energy would face a medium-term shock. The dollar became the only major winner, with USD/CNY returning to 6.8 or above; even Treasuries sold off.
From late March to early April, investors’ “money” suddenly turned risk-on again and crowded back into high-valuation AI and robotics themes in China and the US. What most puzzled 李峰峰 was that the market had “wiped out all the risks of the past.”

3. AI Capex Has Become the Common Foundation of US Growth Expectations

李峰峰 cited an internal macro-oriented study estimating that roughly 50% of the impact on US GDP growth and expected S&P 500 constituent growth comes from AI capex committed by large companies, spreading layer by layer through data centers and the upstream and downstream supply chain.
That means different companies’ revenue and profit expectations are not independent of one another; they are “built on the same block.” The capex at the base must be delivered and can only keep increasing, not be cut.
李翔 pressed on whether revenue and profits can support high valuations. 李峰峰’s conditional answer was that as long as the AI capex block at the bottom holds, the growth above it can work; if it fails, the shock will amplify along the chain.

4. The Q1 Beat Relied on Capacity Built Before the Price Hikes

Most data-center capacity able to contribute revenue in Q1 had been built and brought online before Q3 last year, so its construction costs were largely insulated from the subsequent broad-based increase in memory prices.
Revenue from that low-cost legacy capacity helped hyperscalers come in slightly above consensus and became one reason leading AI assets made new highs again from late March onward.
That cannot be extrapolated directly to returns on the next investment cycle. New data centers face higher memory and chip costs, along with potential financing pressure, so the returns shown by old projects may not represent the economics of incremental projects.

5. Corporate FOMO, Market Narratives and Hoarding Are Amplifying Investment Together

After SaaS companies sold off sharply last year, many technology companies began fearing they would be displaced by AI and adopted the attitude of “regardless of how much we ultimately use, at least start using it.” This is corporate FOMO, not proven and stable demand.
Boards are adding pressure. 李峰峰 observed that directors advise or even pressure management teams to show more AI impact in the business, because “if it isn’t AI enough, the company seems not good enough.” The application narrative then feeds back into capex.
The third force is inventory hoarding. Companies fear chips and memory will be in short supply and more expensive in the second half, so they are pre-ordering data-center capacity and related chips. AI spending is simultaneously a business decision, a capital-markets signal and a hoarding response.

6. The Companies Being Rewarded Are Moving Fastest Toward Negative Free Cash Flow

The internal research placed Microsoft, Meta and Tencent on the relatively less aggressive—and therefore lower-valued—side of the market. “Aggressive” here refers not to absolute spending, but to the market’s horizontal comparison of companies at the same level. 李峰峰 also noted that Meta may simply talk more; it is not necessarily more aggressive in reality.
If current capex guidance is delivered, Amazon may already have capex exceeding free cash flow. Slower spenders such as Microsoft could enter negative free cash flow as early as Q1 or Q2 next year.
李峰峰 described an inversion in valuation logic. Under value investing, the companies that turn cash-flow negative fastest carry the highest risk; under today’s upside narrative, they receive the highest valuations because their AI investment is the most aggressive.

7. Real Token Demand Does Not Mean the Business Model Has Closed the Loop

李翔 offered the strongest case for the bulls: unlike the 1996-2000 internet cycle, AI demand has already passed from the consumer side through models, chips and storage. Model companies are also working to build revenue models, and the answers so far “look reasonably good.”
李峰峰’s rebuttal was that current token pricing definitely cannot cover the full investment in foundation models and data centers. The “savings” users feel correspond to losses still being borne by upstream investors.
The sharper question is whether AI would still be cheaper than human labor if all investment were allocated back into a rational token price. Whether scale will ultimately lower costs and generate profits remains a linear extrapolation, not an already realized fact.

8. Between Cisco and Nvidia, History’s Bet Turns on One Phrase: “This Time Is Different”

In 2000, the market viewed Cisco as the company least likely to have problems; its peak market cap was roughly $550B. Today, AI chips are seen as the “water, electricity and coal” of AI, and Nvidia’s market cap has reached $5.5T—exactly 10x Cisco’s peak.
李峰峰 acknowledged that the monetary backdrop is different: US M2 is about 5x its level 25 years ago, and dollars circulate globally, making the actual total impossible to calculate precisely. But that alone does not prove the valuation logic has changed.
李翔 quoted Marc Andreessen: “The five most dangerous words in investing are ‘this time is different’—but this time really is different.” 李峰峰’s personal judgment is the opposite, and he expects validation within a year.

9. Cross-Border Restrictions May Target Outflows, Bubbles and Short-Term Capital at Once

Futu and Tiger Brokers stopped accepting new mainland users roughly 4 years ago. This time, the policy moved from deferred restrictions to enforcement, requiring net selling to begin and the exit to be completed within 2 years. Chinese ADRs showed little overall reaction on Friday; the direct impact fell mainly on the brokers themselves.
李峰峰 cautiously speculated that regulators may also fear mainland capital being caught in the late stage of an offshore bubble. Historically, the final wave of trapped investors has often included large amounts of European and Middle Eastern capital, not only US investors.
李翔’s inference was more direct: this may not be about protecting a small group of high-net-worth individuals, but about the difference in returns onshore and offshore and a desire to prevent more mainland capital from moving overseas. 李峰峰 acknowledged that this explanation could entirely be correct.
Hong Kong raised its account-opening requirements on the same day, leading 李峰峰 to suspect a common cause. Tax scrutiny, hot money that regulators do not want to make traceable, and risky assets and short-term speculation in Hong Kong and US stocks may all be in the regulatory crosshairs.

10. Buffett’s Cash Can Identify the Regime, Not the Timing

Berkshire held more than $300B in cash and cash equivalents; more importantly, cash had risen above 50% of total assets. 李峰峰 asked colleagues to review the historical record, with the conclusion that it can “identify the regime, not quantify the timing.”
Crises have generally followed periods when Buffett’s cash was high, but they often arrived after cash had already fallen from its peak. The interval has ranged from more than half a year to a year and a half; he can “forecast the phase, but not the event or its timing.”
The larger the portfolio, the longer the execution lead time. If Buffett is reducing a huge Apple position, he cannot wait until the market is close to falling before selling. The scale of his capital means he can be neither fully out nor fully invested.
李翔 offered another explanation: “As long as the music keeps playing, we all have to keep dancing.” 李峰峰 instead found that when the market restarted after risk-off, risk appetite may actually have become stronger.

11. Global Liquidity Is Not Loose, Yet Risk Assets Still Need Incremental Capital

No major economy is engaged in large-scale easing, leaving the market looking more like a contest over existing liquidity. Japan’s low-yield yen once supplied carry-trade funding, but yen depreciation and possible tightening are weakening that source of liquidity.
If the S&P 500 and Dow break to new highs, that implies incremental money may be entering the market. 李峰峰’s puzzle is that if some capital has indeed recently returned to the US, it does not appear to have prioritized long-dated Treasuries yielding close to 5%; it may instead have bought more volatile AI and equity assets.
李翔 suggested the market may no longer define these AI assets as high risk. 李峰峰 acknowledged that this could be the answer, but admitted that current capital flows have moved beyond his own knowledge and explanatory toolkit.

12. Trump’s China Visit Pushed US-China Relations Toward a Medium-Term, Not 90-Day, Equilibrium

The US disclosed the terms that tilted toward its own gains first. China then supplied the other side of the ledger: Boeing purchases paired with guaranteed engine-parts supply, agricultural purchases paired with reciprocal cuts to some tariffs, and a temporarily disclosed amount of roughly $30B.
The more important signal was China’s confirmation that Trump had invited China’s leader to visit the US this autumn. 李峰峰 reiterated his earlier view: a formal visit usually means tariff issues cannot be managed through repeated 90-day extensions and instead require a medium-term arrangement.
Because Trump cannot bind his successor, the most realistic duration for such an arrangement is 1-2 years covering his term. The two sides may currently have agreed only on direction and principles; details still need to be settled before the visit, which 李峰峰 guesses could happen in July or August.

13. The Sequence of Visits to China Raised the Probability of a US-Iran Ceasefire

Before Trump’s visit to China, Iran’s foreign minister arrived first. Putin came on the 4th day after Trump left, and the Pakistani side followed on the 3rd day after Putin departed. 李峰峰 inferred that China may have coordinated compromise terms with each side.
He therefore moved from his previous view that a ceasefire would be difficult to the judgment that “a ceasefire is probably coming.” But a ceasefire and end to hostilities is not the same as a completed agreement; the parties may continue to haggle between a principled balance and specific terms.

14. Uranium Enrichment Is the Stated Focus; Hormuz Is the Real Bargaining Chip

李翔 asked whether Iran has a red line that is effectively non-negotiable. 李峰峰 said uranium enrichment is the US’s official rationale, Israel’s urgent security concern and the visible indicator by which outsiders judge whether Iran has compromised.
The Strait of Hormuz is more important in practice. Iran has demonstrated that its ability to control the strait can shift the balance of a war, so it cannot agree to restore Hormuz fully to its previous status, much less accept joint management by the US, an international organization or multiple parties.
Restoring navigation is also an outcome both China and the US need. Any negotiation therefore has to do two things at once: restore the flow of energy and supply chains while allowing Iran to retain its “most important card.”

15. Countries Challenging the Petrodollar Keep Facing the Same Kind of Pressure

李峰峰 linked together Iraq, which tried to settle oil in euros after the euro launched in 1999; Libya, which advocated an African common currency for energy settlement; and Venezuela, Iran and Russia.
As of the end of last year, the episode cited an estimate that roughly 85% of Venezuela’s oil settlements used RMB. Iran and Russia have also worked to move out of the dollar system under sanctions, making the RMB a factor that has gained traction only over the past 10 years.
The discussion also touched on reports that the UAE was leaving OPEC, but 李峰峰 said the reasons may be more complicated and the same framework cannot be applied directly. The core judgment remains that the currency used for energy settlement touches the dollar’s influence and pricing power.

16. The Rationale for Controlling Oil Keeps Changing; the Underlying Interest Does Not

李峰峰 divided the official narratives for US intervention abroad into 3 phases: ideological mobilization during the Cold War; autocracy versus democracy after the Cold War; and counterterrorism and security after 9/11.
His question is why some Gulf states friendly to the US are no closer to American values in governance and electoral systems than hostile states. The public rationale alone may not explain the policy choices.
Iran fits the logic of “no crime except possessing a treasure” particularly well. From 1951 to 1953, Mossadegh pushed oil nationalization and clashed with Britain and oil companies; similar conflicts over oil access and control emerged during the later Pahlavi period.
李峰峰 cited William Engdahl’s The Wars of Oil as supplementary reading rather than a formal recommendation. The book attributes many events to a small number of political and oil-interest groups, and 李峰峰 acknowledged that much of its content lies beyond his own knowledge.

17. If the US Keeps Pulling Back, the Middle East Will Seek New Security and Capital Partners

After the oil crisis of the 20th century in the 70s, the US strengthened control through the petrodollar system. After the Cold War, it expanded from the Gulf War and deployments in Afghanistan and Iraq to a broad network of bases, upgrading financial constraints into military influence.
Around 2011, US strategic focus shifted toward Asia. Shale oil and gas also no longer required prices above $80 a barrel to be commercially viable, turning the Middle East from an energy lifeline into a region where the US only needed to preserve pricing and order.
李峰峰 speculated that Israel may try to use one final opportunity to knock out Iran before the US pulls back further. But if negotiations ultimately remain necessary and produce a new balance, the war could also accelerate the US reduction of its military presence.
If that trend holds, the Middle East could reconfigure security relationships with Pakistan, Egypt and others, while reallocating financial assets. Around 2023, Middle Eastern capital was still probing China more tentatively; after this episode, it may allocate to China more explicitly. 李峰峰 repeatedly stressed that this is only speculation.

18. CIPS Is Evolving from a Messaging Channel into the RMB’s Clearing Base

CIPS processed less than RMB200T last year. Based on daily throughput above RMB1T in March, 李峰峰 estimates full-year volume could reach more than RMB300T or higher.
李峰峰 said one important change in 2025 may be that CIPS is no longer merely a messaging system similar to SWIFT. As more foreign banks connect directly, it is beginning to handle actual clearing as well and become an integrated clearing and settlement system.
Combining CIPS and other measures for a rough estimate, 李峰峰 thinks the RMB’s share of global transaction currencies may be close to 10%, above the roughly 4% shown by SWIFT alone and ranked behind only the dollar and euro.
With the capital account still closed, he said reaching 15%-16% over the next 3-5 years would already be “very good,” potentially bringing the RMB close to the euro. RMB usage has already surpassed the yen and pound; on a relative basis, 李峰峰 guesses the dollar is the most likely to lose share.

19. A 5% 30-Year Treasury Yield Links Dollar Credit and AI Financing in One Risk Chain

Over the past 3 weeks, the auction yield on 30-year US Treasuries rose above 5%. This does not mean bonds became more popular; it means the Treasury needed to offer an auction yield of roughly 5% at the time to complete the issuance, reflecting weak demand at the long end.
Possible causes include Fed balance-sheet runoff, long-term inflation, US growth competitiveness and debt sustainability. The episode cited roughly $38T of US government debt, $30T of GDP and an annual deficit close to $2T.
李翔 argued that higher yields would eventually attract capital back and push rates lower. 李峰峰 responded that if investors require yields near 5% even for supposedly risk-free assets, they are demanding compensation for currency, inflation or concerns over the US’s long-term economy and debt. Current inflows are not yet enough to reverse the result.
After turning free-cash-flow negative, AI companies can draw down cash, issue debt or sell equity. Equity is more expensive, so companies generally choose debt. When the risk-free benchmark is already 5%, even a high-quality company paying only a few dozen basis points more faces a materially higher funding cost for long-term capex.

20. Stabilizing Property Prices Have Not Restored Household Long-Term Leverage

李峰峰 predicted early this year that property markets in major cities would stabilize within the year; he now thinks the actual timing may be earlier. Secondhand prices in Beijing, Shanghai, Guangzhou and Shenzhen posted at least slight gains in April, while new-home prices continued to fall in some cities.
More notable is that household medium- and long-term loans declined even when transaction volumes were reasonably healthy, suggesting households are still actively reducing long-term leverage. New transactions may be driven more by chains of selling an old home to buy a new one than by fresh credit expansion.
Local governments have adjusted the ceilings, amounts and ratios for housing provident-fund loans, which may also be diverting demand from commercial mortgages. If this is indeed substitution, it may indicate a higher share of essential housing demand. 李峰峰 cannot separate the contributions from deleveraging, move-up transactions and provident-fund substitution, only that all 3 may be occurring together.
Even if property stabilizes, banks may not be able to return to easy profits from mortgages. State-backed funds investing in advanced chips and large models, while enabling bank AICs to earn profits, amounted to “pulling banks along”; if long-term lending does not grow, banks may be “forced to turn” further.

21. China Is Showing Weak Stabilization as Growth Moves Beyond Traditional Social Financing

CPI turned positive year on year in March and April, but PPI rose faster, meaning producer prices are recovering more quickly than consumer prices and raw-material costs may be squeezing midstream profits. That is consistent with manufacturing and exports holding up while end-consumption remains weak.
April retail sales grew by only a few tenths of a percent, with the drag concentrated in automobiles, home appliances and furnishings, and gold and jewelry, alongside factors such as pork. Autos and appliances faced a high base from last year’s trade-in program, while services, travel and tourism were relatively resilient, making the picture closer to “weak stabilization and weak recovery.” 李峰峰 said he has not specifically studied the hog cycle and has only a limited view of its persistence.
Strong growth in imports and exports is not primarily a function of higher commodity prices. The biggest incremental gains came from AI capex-related electromechanical goods, digital products, chips and equipment, with automobiles also contributing to exports; commodities accounted for roughly 20%-30% of incremental imports.
As growth shifts toward services, automobiles, chips and biopharmaceuticals, each unit of output requires relatively less land, equipment and credit, so the old relationship between aggregate social financing and economic growth will change over the long term. Global shocks also lead 李峰峰 to judge that China’s industrial chains remain stable, and that foreign investors’ willingness to retain capacity in China may be improving versus the past 2 years.

22. “Civilization-State” Explains Why China Resists a Binary Washington Consensus

李峰峰 ended by recommending The Ambition of a Great Power, leaving a detailed discussion for the next episode. Its key framework is that China is not merely a territorial state or a nation-state, but a “civilization-state wearing the clothes of a nation-state,” with continuous cultural, written and historical systems.
The Washington Consensus bundles markets, finance, trade, governance and values into a system treated as correct. If the West accepts that China’s path can work, it must also accept that some of its principles are not the only possible answers.
李峰峰 summarized the Western tradition as more inclined toward black-and-white, either-or binaries, while Confucianism, Buddhism, Taoism, Taiji and “socialism with Chinese characteristics” are more accustomed to recombining opposing elements. 李翔 noted that European social democracy is also a middle path; 李峰峰 reserved judgment: “I haven’t studied that.”

Building an AI Guardian for Enterprise with Onyx Security CEO Maxim Bar Kogan Legendary Investor Dan Loeb on AI, Credit, & Third Point’s $25B Strategy