Vol. 240 Macro Talk 113 | China’s Preemptive Rebalancing and Global Equity Volatility (Recorded 10.30)

Summary

• The US AI trade is shifting from rewarding spending to scrutinizing revenue and returns; growth no longer earns a premium by itself. 李丰 says investors are asking whether aggressive capex will require a significant increase in debt and whether companies can deliver the expected return on every dollar invested. Microsoft was rewarded for taking a cautious approach to spending, then gave back part of its gains after announcing $25B in financing; a potential Anthropic IPO and OpenAI’s massive private fundraising are competing for capital that is already tight.
• 李丰 sees the current US market as marked by violent swings at the tail end of a liquidity cycle, but not yet in a sharp selloff. US equity market capitalization is about 3.5x US M2, leaving fewer sources of capital to push risk assets higher. Sharp selloffs are usually triggered by sudden external shocks, and their timing is unknowable. “The last money in a liquidity cycle is the hardest to make,” and short positions can still face squeezes and decay.
• In 李丰’s view, A-shares are rebalancing ahead of the rest of the world, so the decline cannot be explained simply as spillover from US equities. After July’s sharp volatility, oversight of public funds’ style drift and new rules that begin a trial run on December 1 may have prompted funds to adjust early; insurers, after a volatile third quarter for profits, may also be adding stable, high-dividend stocks. His conditional view: early adjustments still inflict losses, but if global markets suffer a true sharp selloff, the decline “might be somewhat slower and shallower.”
• 李丰 sees active ETFs and limits on index constituent weights as complementary fixes to the market’s stabilization toolkit. In the past, state-backed funds could only sell broad-market holdings to restrain index gains, draining more capital from blue chips in the process; active ETFs could allow for more selective buying and selling. Meanwhile, rising technology stocks entering indexes—and larger weights for companies such as ChangXin—can amplify index swings, making a 15% single-stock weight cap meaningful in practice.
• As 李丰 sees it, housing policy is aimed at supporting first-time buyers at the lower end of the market while leaving developers, local governments and banks to bear their own market risks. He first cited a 1-percentage-point interest subsidy for first-home purchases under RMB1.5M and 120 square meters, but later said the price threshold was RMB2M, leaving his account inconsistent. He expects high-tier cities to tailor policy locally and is not certain the market has stabilized. If home prices stop dragging down household balance sheets, that could ease the biggest constraint on consumption.
• 李丰 links bank marketization, foreign competition and the development of offshore RMB finance in Hong Kong as parts of the same financial shift. Intensifying competition for deposits and loans may initially favor large banks with lower funding costs, stronger brands and wider branch networks; the crackdown on loan facilitators requires banks to manage borrower relationships and credit risk themselves. He sees foreign access, brokerage mergers and a broader range of Hong Kong financial products as preparation for internationalization, but reads only very cautious signals on capital-account opening.
• The long-term value of a technology cycle and a capital-cycle bubble can coexist. 李丰 argues that wealth effects have significantly inflated short-term expectations for technology, with the strong consensus since 2025 turning into “no room for dissent.” In domestic industrial policy, he sees a more pragmatic emphasis: companies articulate their needs, research institutes support them, and technology is brought into existing industries at scale while businesses seek “real-world applications and sustainable business models.”
• 李丰 sees the next phase of venture investing shifting from robotics hype toward real-world applications, while leaving room for technology to branch out. He says some funds had already told him since September that they would make no more embodied-AI investments that year; the next question is “who is actually getting work done,” and valuations will struggle to sustain their previous highs. Since the start of the year, his team has paid more attention to biopharma, consumer AI and edge solutions, as well as data generated by new sensors and architectures beyond the standard Transformer.

Deep dive

1. AI capex is starting to face a return test

• 李翔 opened with Google’s decline and reports that Amazon’s free cash flow could turn negative. 李丰’s explanation is that the market has at least stopped rewarding aggressive spending, particularly when it requires significant additional debt or cloud revenue fails to deliver “the expected return on every dollar invested.”
• 李丰 says Treasuries, data centers and rising share prices all require huge amounts of capital. When money is scarce, companies can keep growing and still trigger a flight to safety if they merely fall short of expectations; in that environment, it is also hard for the whole market to rise together.
• Microsoft was first rewarded for spending cautiously, then gave back some of its gains after announcing $25B in financing. He says Anthropic’s valuation discussions moved from $2T to $1.5T, while OpenAI is also pursuing a trillion-dollar valuation and financing worth tens of billions, accelerating the “drain” on available capital.

2. Tail-end volatility is here, but the timing of a sharp selloff is unknowable

• 李丰 describes the market as “pushing a rock uphill”: it can no longer make progress, but is still being kept in place rather than rolling rapidly back down. Past sharp selloffs have often started with external shocks that trigger a chain reaction of risk aversion; current volatility offers no way to predict when that process will begin.
• He admits, “I’m losing money too.” His position is more than 50%; he rotated from tech into defensive assets relatively early in the second quarter, only to first bear the brunt of capital being pulled from blue chips. Colleagues with lighter positions may be modestly in the money. Kweichow Moutai is still up from January 2.
• His 1x short in Nvidia is “very, very small,” mainly to track the cycle. His average entry was about $223; after roughly 3% of reset decay over a month, he estimates the cost basis is down to about $217, leaving only a small gain.
• Why not short highly leveraged cloud companies such as CoreWeave? 李丰 says shorts there are crowded and vulnerable to a squeeze. He chose Nvidia because he thinks it will be the last leading stock to fall in a sustained downtrend, making it a better gauge of the cycle. Daily-reset products decay through repeated swings and are difficult to hold over the long term.

3. Tech companies’ dollar bonds need to be viewed alongside their listed entities

• 李翔 asks why Chinese tech companies issue dollar bonds offshore when domestic banks need to lend. 李丰 speculates that the key factors are the primary listing venue and financing arrangements; Hong Kong syndicates likely include Chinese banks, which hold at least a substantial share of the offshore dollar-bond market.
• 李丰 does not simply agree that borrowing in renminbi would be cheaper. Drawing on Bilibili’s financing while he was a board member, he says Hong Kong financing costs are also low, and debt may carry conversion features or other preferential rights. Issuing onshore would run into capital-account separation constraints on convertibility.

4. A-shares started rebalancing before global markets

• 李丰 dates the shift to the sharp volatility in July. Enforcement against public funds for style drift and concentrated positions in August and September, along with management rules due to begin a trial run on December 1, are all constraining the concentration of capital in a few market segments.
• Funds are unlikely to wait until November 30 to adjust their positions all at once. 李丰 suspects they have already been repositioning from September through November, using the third-quarter reporting window. As flows shift, the tech stocks with the biggest prior gains naturally come under pressure, with declines in overseas markets potentially adding to the effect.

5. Insurers’ profit swings could push long-term capital toward high dividends

• In 李丰’s explanation, when insurers buy stocks to capture trading gains, quarterly price moves feed into current-period profit. For long-term dividend holdings, price fluctuations do not flow through the same current-profit accounting; realized gains are instead recorded under other income. He sees this distinction as key to understanding changes in insurers’ portfolios.
• Third-quarter profits at some large insurers fell sharply because of swings in investment valuations, which may encourage them to add stable, high-dividend assets. Bond yields are around 2%, while participating insurance policies need to clear yield hurdles such as 2.6% and 2.8%, leaving pressure to allocate to equities.
• He links this to regulators’ preference for “long-term capital and long-term investment.” Buffett can hold positions for years partly because insurance capital backs him—not only because he looks at free cash flow, dividends, buybacks and absolute prices.
• He revisits research showing that Buffett’s cash allocation has historically peaked about 8–10 months before sharp selloffs. 李翔 asks whether Buffett’s retirement could break the pattern; 李丰 says, “It’s possible.” He also recalls an approximately $10B allocation to Google in May, with an entry price, as he remembers it, around $350 versus a current price near $280, acknowledging that even long-term positions carry short-term risk.

6. Active ETFs could make market stabilization more selective

• 李丰 says state-backed funds had largely sold the roughly RMB1.2T of stabilization holdings by May. The problem was that they held mostly broad-market funds when they needed to restrain an index rise led by tech stocks; selling blue chips from those funds, on top of funds selling blue chips to buy tech, meant “blue chips got hit twice.”
• He sees the recent wave of active ETF launches as a repair to the toolkit: these products can selectively hold index constituents, potentially giving China Securities Finance and Huijin more options for selective buying and selling when they intervene to stabilize the market.

7. Indexes add past winners—and amplify their pullbacks

• 李丰 notes that June’s index changes added fast-growing technology companies to the Hang Seng Tech Index, the Hang Seng Index and related domestic indexes. When those stocks later retreated, they weighed more heavily on the indexes.
• He links the 15% single-stock weight cap to ChangXin. As he describes it, ChangXin’s market value is around RMB3T and peaked near RMB4T; its weight would have been too large on inclusion, making a pullback harder to stabilize. South Korea’s top 2 companies, together accounting for around 40%–50% of index weight, are a more extreme example.
• The top 3 US companies account for around 20% of the market and the top 10 around 40%, creating a similar amplification effect. 李丰 stresses that early rebalancing cannot prevent A-shares from falling during a sharp global selloff; it might only slow and soften the decline. In a panic, that distinction could be overwhelmed.

8. Technology breakthroughs and capital bubbles need to be discussed separately

• 李翔 observes that public-market investors, including Ken Griffin, tend to call this a bubble, while venture investors and founders point to a massive technological wave. 李丰’s response: “The technology cycle and the capital cycle need to be viewed separately.” Technology can matter over the long term even as capital leverage pushes short-term expectations much higher.
• He recalls that consensus formed slowly after November 2022. In 2023–2024, domestic large language models still relied mainly on state capital and local-government investment. Asked whether venture funds’ own circumstances also played a role, 李丰 returns to the point that very few people were genuinely involved at the time.
• By 2025, the wealth effect had turned consensus into “no room for dissent.” Investors could only extrapolate in a straight line: “They stopped considering other options.” People with a financial-market lens attribute the outcome to liquidity; those with a technology lens are more likely to credit technology itself. His prescription is to disentangle the two.

9. Finding more capital to keep asset prices rising is getting harder

• Comparing the current market with the dot-com bubble, 李丰 says US M2 is now around 6–7x larger and market capitalization about 10x higher, with global capital also flowing back into dollar assets. Total US equity market capitalization is about 3.5x M2; as he puts it, “All the money that could buy goods has bought them, and all the money that could buy assets has bought them.”
• 李翔 asks whether central banks and governments could create more money to support markets. 李丰 says they will try, but countries are raising rates because of exchange rates or inflation, making another synchronized global easing cycle difficult. Rates were unchanged in October, but the market has started discussing a possible December hike.
• A stronger dollar has limited room to attract more capital, while higher interest costs have lasting effects. He points to US mortgage rates of 7%–8% weighing on home-buying demand. More expensive Treasury financing also raises data-center borrowing costs, which are priced at a spread over Treasury yields.

10. Housing support targets first-time buyers as market risks return to businesses

• 李丰 first says first-home purchases with a total price under RMB1.5M and an area of no more than 120 square meters qualify for a 1-percentage-point government interest subsidy. He later says the price threshold is not RMB1.5M but RMB2M, leaving his account inconsistent. He also notes that homes priced below RMB2M make up half of Beijing transactions, showing that this is not a niche market.
• He thinks the policies have had some effect in higher-tier cities, but cheaper secondhand homes are selling first, and prices may continue to fall. Lower-tier cities face tighter constraints from population and fiscal capacity; the subsidy offers support where supply and demand are relatively weak.
• In his view, developers have lost the benefits of presale funding and high leverage; local governments must rely on city competitiveness to generate land revenue; and banks have to assess the risks tied to the city, land parcel, builder and sales. Every part of the property market is being pushed toward market discipline.
• On whether the market has bottomed, he leans toward it being close to stabilization but leaves room for doubt. He compares Japan’s roughly 25 years of adjustment before Abenomics with the pain after China’s property bubble was pierced in 2021. Stabilizing home prices would be necessary to ease the biggest constraint on consumption: household balance sheets.

11. Deposit competition is reshaping flows across banks and wealth management

• 李丰 interprets signals on deposit and lending rates from the Financial Street Forum as a move toward fuller marketization. He links that to the RMB1.9T drop in non-bank financial deposits in September: time deposits and large-denomination certificates paying close to 2% may have drawn money out of wealth management products and into deposits.
• Volatile wealth-management net asset values, equity-market swings and precautionary saving all make deposits more attractive. In his view, that puts pressure on fixed-income wealth management products as well as new life and property-and-casualty insurance sales.
• Large banks may benefit in the near term from their brands, branch networks and lower funding costs. He sees the RMB100B-scale capital injections into major banks and large insurers in September as a way to “top them up, then push them into the market.”

12. The loan-facilitation crackdown requires banks to own credit risk

• 李丰 links the tightening of loan facilitation since last October—and the rate cap being brought down first to 24%, then to 4x LPR—to the broader move toward bank marketization. As local-government financing vehicles and property financing contract, smaller banks may try to shift toward small businesses and consumers without having the capability, ultimately becoming funding pools for loan facilitators.
• Small banks that pay high rates to attract deposits but do not understand their assets take on added interest-rate, duration-mismatch and bad-debt risks. Regulators want banks to reach borrowers directly, understand their assets and household finances, and manage underwriting themselves: “You can use loan facilitators, but you can’t turn the bank into a funding pool.”

13. Foreign access and brokerage mergers both point to stronger competition

• 李丰 sees Visa’s clearing-and-settlement license, along with easier foreign-exchange use and quotas, as signals of financial opening. On capital-account access, however, he says the country may be “very cautiously testing” whether to open “the tiniest crack.”
• He notes that foreign banks are more actively offering deposits, with three-year large-denomination certificates paying around 1.92%–1.93%. 李翔 points out that they previously had little consumer-facing business. 李丰 agrees that their activity is still mainly corporate, adding, “I don’t know whether this counts as a signal.”
• In 李丰’s view, the mergers of brokerages such as Guotai and Haitong are meant to build scale, product breadth, branch networks and resilience ahead of foreign competition. He sees Citi’s expansion into brokerage business in China as part of the same set of developments.

14. Hong Kong needs to give offshore renminbi more uses

• 李丰 says the long-term goal of this shift is for China to become an international financial center and for Chinese financial institutions to compete globally. Whether that happens still depends on US-China competition, China’s own capabilities and global acceptance.
• Before the capital account is fully open, Hong Kong is positioned as the most important offshore renminbi center. Holders of offshore renminbi need more than a way to make payments: they need bonds, other financial products and derivatives. He also sees pension-fund support for offshore RMB bonds in Hong Kong as support for products and liquidity.

15. Tech policy is putting more emphasis on business demand and industrial applications

• From recent central-government documents and official interpretations, 李丰 sees 2 major changes: companies, especially private companies, are to lead innovation by defining needs and working with research institutes, with results commercialized at the company level; and technology must be put into practical applications.
• He connects this with the direction of August’s SASAC Document No. 54: “Don’t all rush in at once,” and don’t use a single approach to attract investment; instead, build on local industries’ strengths. 李翔 sums it up as “a bit more pragmatic.”
• He sees 3 levels of application: adding technology to scaled manufacturing; improving quality and efficiency in established emerging industries such as new energy, electric vehicles and aerospace; and keeping multiple paths open in future industries such as quantum computing and controlled nuclear fusion while exploring “real-world applications and sustainable business models.”

16. Robotics’ next phase will have to prove its value through real work

• 李丰 mentions circular transactions between robotics companies and data-collection vendors, and says portfolio companies have told him that some funds had already announced from September that they would make no more embodied-AI investments that year. 李翔 thinks existing companies are too expensive; 李丰 agrees, while noting that new companies are still emerging.
• He expects the next question to be “who is actually getting work done.” Whether a robot is humanoid is not the key point: an arm with a brain or legs with a brain can both work; what matters is real-world deployment.
• He expects robotics companies in 6 months to have to balance commercial deployment with research: companies will define what they need robots to do and work with research institutes to develop those capabilities, “one side in the sky, the other on the ground.” Even if they fit the policy direction, valuations are unlikely to regain their previous heat.

17. Venture investors are looking for consumer returns and technological branches

• 李丰 says his team continued to invest in biopharma this year and, from the start of the year, was asked to look more closely at AI applications and edge solutions: software, consumer electronics, chips, or chips paired with algorithms. The key is for AI to do something useful for consumers—and even generate revenue.
• They are also looking at how new sensors create new data and at architectures beyond the standard Transformer. After consensus has become so strong, he would rather watch for “roads that branch off” than bet only on a straight-line extension of existing technology.

Verification Notes

• The transcript gives conflicting thresholds for the housing interest subsidy: first, it says first-home purchases priced below RMB1.5M and no larger than 120 square meters qualify; later, it says “the threshold isn’t RMB1.5M, it’s RMB2M.” The final threshold cannot be determined from the transcript.