Macro Talk 90: Aug–Dec: Stocks, Anti-Involution & Next DeepSeek Moment
Summary
The latest rally in A/H shares is first and foremost a shift in the funding mix—not a full return of foreign capital or a completed fundamental turnaround. A-shares have been driven mainly by increased allocations from insurers and other long-duration capital; banks combine high dividends, blue-chip status and bancassurance synergies, making them the preferred destination. Hong Kong stocks have relied mainly on Southbound flows and passive allocations from MSCI and other indices, with the former more than 2x to 3x the latter. Active foreign equity funds recorded their first weekly net inflow only in the week before recording—far too soon to confirm a trend. Capital is still rotating according to the usual sequence: offshore before onshore, bonds before equities.
Li Feng sees late August through October 2025 as a window in which confidence could bottom, but the trigger would not be a single policy—it would be a stack of consecutive events pointing in the same direction. The upcoming SCO summit, September 3 events, the Fourth Plenum of the 20th Central Committee and the 15th Five-Year Plan account for roughly half of the variables China can control. If external variables—including a final China-US tariff agreement, Fed rate cuts and a phased Russia-Ukraine ceasefire—arrive at the same time, they could produce one or two “DeepSeek moments.” His probability scale is: “A 70% chance of happening would be okay; a 100% chance would be quite good.”
Second-half upside in Hong Kong will depend more on foreign capital rotating back in, while A-shares will depend more on improving insurer fundamentals, household confidence and the conversion of excess savings. Once the A/H premium narrows, Southbound flows will not be as forceful as in the first half and could even see profit-taking. But a long-term strategy of allocating only to the US is unsustainable for foreign investors, leaving substantial room for active funds to repair their underweights. The speed and scale of the return will depend on China’s relative performance, the dollar and external risks. If capital can shift “from fast money to long money,” the rally could broaden from concept stocks into consumption, the full healthcare chain and other parts of the real economy.
The key US risk is not a simple recession, but possible stagflation driven by labor-market mismatches, tariff inflation and slowing growth. High-value-added services and technology companies are cutting jobs, while restaurants, hotels, caregiving and construction face labor shortages as immigrant workers decline. Job demand and net job creation can therefore diverge. Service-sector wages are difficult to bring down, and tariffs on imported goods add another layer of uncertainty to the inflation trend. After employment data were revised, the market briefly pushed the probability of a September rate cut close to 90%. Historically, US easing cycles have often coincided with capital outflows from the US, which would be positive at the margin for the renminbi, Hong Kong stocks and China’s policy room.
The substance of “anti-involution” is to harden resource costs, budget constraints and industrial-chain cash flow—not merely to order companies to stop cutting prices. Local subsidies, cheap land and factories, and soft budget constraints prevent some companies from exiting. Industry leaders can also squeeze upstream gross margins while extending payment terms from “3 months to 6 months to 9 months,” amplifying the damage through China’s long B2B chains. A unified national market, fees for resource use and supply-side adjustment could correct the distortions, but the 2015–2017 capacity-cutting experience shows that the process will hurt. PPI may not begin to improve visibly until September or October.
Whether the food-delivery war qualifies as involution depends on whether the platforms or small merchants bear the subsidies, while the competition itself may be breaking a bottleneck around the RMB5T restaurant market. If the platform pays RMB8.5 of a RMB10 discount and the merchant pays RMB1.5, the promotion looks more like incremental demand. If both sides pay RMB5, the damage to the industry could exceed the lift. Li Feng’s structural case for Taobao’s entry is not about short-term victory, but about how high-engagement apps are better positioned to create a “browse-and-eat” mindset. Over the long term, an enormous industry cannot be monopolized by one link and will continue to split across traffic, fulfillment, delivery and supply chains.
The main AI investment question is shifting from “whose model is bigger” to “who can sell it, get it used and make money,” while open source continues to erode pure-model moats. DeepSeek showed that model competition need not be a race to scale and opened a window for high-quality open source. As Kimi, Qwen, Brock and even GPT-5 stopped generating broad social excitement, capital began moving toward agents, AI hardware, robots, edge chips and real applications. The primary market has also moved from “warming” in 4Q24 to pockets of genuine competition since 2Q25: roughly half of the early-stage deals under consideration now require investors to fight for allocations. The secondary-to-primary-market transmission typically lags by 1–2 quarters.
Deep dive
1. Real growth of 5.3% masks still-weak nominal growth
Li Feng first recalibrated China’s first-half GDP: real growth was 5.3% year on year, better than many expected. But the GDP deflator was negative, meaning that once the effect of falling prices is included, “real GDP growth exceeded nominal GDP growth.” He used 4.5%–4.8% as an illustrative range—not as an official nominal-growth estimate—to explain how deflation is added back into the real-growth measure.
The US second-quarter headline growth rate of roughly 3% was an annualized quarter-on-quarter rate: the change from the first quarter was magnified to an annual pace, while the first quarter itself was negative, creating a low-base effect. Li Feng estimates that, viewed on a smoothed first-half basis, growth was only a little above 1%. Investors should not directly compare the China and US figures, which use different statistical conventions.
2. Half of the previous liquidity call was validated; half needs to be revised
Reviewing global liquidity from 2020 to 2022, Li Feng confirmed that the core direction was right: China’s balance-sheet expansion was relatively modest, while other major central banks created an unprecedented amount of money. He revised his estimate of the global total to roughly $20T, below the previous estimate of about $30T, but still called it “the largest amount of money in history.”
Starting in 2Q22, Europe was hit by the Russia-Ukraine war and international investors grew concerned about China’s policies and outlook, sending large volumes of capital to the US. Even after the rate-hiking cycle began, incoming funds allocated more to US equities than equities’ natural share of US financial assets, creating an overweight.
The earlier view that this flow began reversing in 4Q24 is the part that needs correcting. The statistically safer conclusion is only that “money flowing into US markets stopped growing”—not that capital has systematically left the US and returned to China.
Hong Kong stocks in the first half were driven mainly by renminbi Southbound flows and passive foreign index allocations, with the former more than 2x to 3x the latter. Active foreign equity funds continued to record net outflows from A/H shares until the first weekly net inflow in the week before recording. “A reversal to net inflows for one week does not necessarily mean sustained net inflows have begun.”
3. The more than $40B flowing into China and Hong Kong went mainly into bonds, not equities
Li Xiang asked how this data squared with the program’s earlier conclusion that foreign capital had already begun returning. Li Feng acknowledged that he was “a little surprised” to see active equity funds still flowing out, and stressed that secondary-market funding sources, channels, instruments and shell structures are complex. Even cross-checking cannot claim 100% accuracy.
The Hong Kong Economic Journal’s estimate of more than HK$300B, or a little over $40B, in net inflows from Europe and the US into China and Hong Kong during the roughly 2 months after the tariff war does not mean foreign investors rushed to buy Chinese stocks. Research shows that most of the money was first allocated to bonds, exactly matching the usual sequence: offshore before onshore, bonds before equities.
The long-term outflow curve for active foreign funds that began forming at the end of 2022 flattened only in 2Q25, despite interim fluctuations around “9/24” in 2024 and the start of 2025. The most cautious formulation is therefore that active allocation may have bottomed—not that sustained net inflows have already formed.
4. Global capital is still comparing who has become “less bad relative to the others”
Li Feng summarizes the China-US-Europe allocation framework as a relative race: “Whoever becomes a little better relative to the others becomes better at attracting capital.” If the US deteriorates faster relative to itself, the reallocation into Chinese assets will accelerate; if China deteriorates faster, the reversal will not arrive.
A considerable share of the money flowing into Europe since the start of this year was money that had previously moved from Europe to the US beginning in 2022 and was now partially reallocated back into European assets. The broader rule is not that China merely needs to improve in absolute terms, but that capital continually compares the marginal direction of growth, policy and risk across the 3 major markets.
This also explains the difference between Chinese and US fund managers’ views of US equities. Some Chinese private-fund managers still believe US stocks can hold at high levels, while major US-based managers are more divided and more sensitive to rising risks: “From outside, it looks as though things are still okay; from inside, people are more cautious.”
5. Hong Kong-dollar strength and insurer allocation explain the first-half divergence between Hong Kong and A-shares
Because the Hong Kong dollar is pegged to the US dollar, its relative strength against the dollar acts as a directional short-cycle indicator for Hong Kong stocks. A firm Hong Kong dollar and expectations of appreciation are generally positive for equities; a stronger dollar against the Hong Kong dollar is usually a headwind. By extension, Fed cuts and a weaker dollar would be clear marginal positives for China’s capital markets.
The drivers of A-shares are different. In addition to incremental public-fund flows through ETFs and similar products, insurers have increased equity allocations since 4Q24, prioritizing high-dividend blue chips and infrastructure assets. Banks connect to the bancassurance business, offer high dividends and face low default risk, making them “the first category of stocks” for long-duration capital.
Li Feng notes that bank shares’ dividend yields naturally fall as prices rise, but relative to global bank valuations on a price-to-book basis, they are still not expensive. The first-half pattern of “strong A-share banks, active Hong Kong technology and concept stocks” was driven above all by differences in the duration of incremental capital in the 2 markets.
6. The baseline for late August through October is not a major rebound, but the possibility of consecutive improvement
At the time of recording, Li Feng judged July’s economic performance to be ordinary, affected by the “ban on alcohol” and weak consumer confidence, among other factors. With August only a few days old, his baseline was also low-level stability similar to July—not an immediate sharp move up or down.
The window worth watching begins in late August or early September. If there are no new domestic or external shocks, a string of positive developments could emerge, with the ultimate goal not being a one-day market move but a sustained improvement in consumer confidence. A similar sequence formed after the Spring Festival before being abruptly interrupted by the tariff war in April.
If confidence bottoms in September or October, household investment, consumption and the conversion of excess savings could all improve. CPI, currently near zero or slightly negative, could also turn. Supply-side anti-involution measures need time to feed through to PPI; like trade-in programs, they may not show results in August, with data feedback more likely in September or October.
7. US equities have completed a V-shaped recovery, but tariffs and employment are not fully priced
After the sharp sell-off on April 7, US stocks had largely recovered the decline several weeks ago, forming a fairly typical V-shaped rebound and now trading around the pre-April highs. The market has absorbed roughly 60% of the tariff uncertainty, but may not have correctly priced the remaining risk.
Li Feng estimates that tariff arrangements for most countries have roughly “settled down” by 60%, but the largest deal—between China and the US—remains unresolved. Temporary tariff increases or adjustments on pharmaceuticals, chips and other products add further uncertainty. What the market genuinely does not know is how much tariffs will raise the underlying level of goods inflation.
After employment data were sharply revised, US stocks first wobbled on signs of economic weakness and then rebounded on rate-cut expectations. At the time of recording, the market’s implied probability of a September cut was close to 90%. The same weak employment report contains both growth risk and a liquidity benefit, so short-term pricing keeps reversing.
8. The US is developing a mismatch of “many jobs, weak employment”
Li Feng’s earlier forecast was that after Trump restricted illegal immigration and work permits, kitchens, restaurants, hotels, construction, repairs and caregiving would still have substantial hiring demand but insufficient low-cost labor supply. Job vacancies could therefore remain high even as net employment growth stayed weak.
The opposite pattern is appearing at the top end of services. Even as Microsoft’s market cap reaches a new high, the company is cutting tens of thousands of employees. Full-time jobs are declining in high-value-added services such as banking and technology, but those workers cannot simply be transferred into lower-paid basic services such as restaurants and caregiving.
This mismatch could keep service-sector wages rising slowly. If tariffs also raise the price of imported consumer goods, the US could enter stagflation in the ordinary sense of the word: “economic growth faces some challenge, or stagnation, while inflation also appears.”
US products with a domestic advantage may not rise in price. Li Feng observes that prices for some agricultural products have edged down, and characterizes typical US exports as military equipment, energy and agricultural products. The key question is how tariffs on imported industrial and consumer goods transmit after inventories rushed out before the tariffs are exhausted, beginning in June or July.
9. Roughly half of the US’s 3% second-quarter growth came from a reversal in trade
US GDP was negative in the first quarter and positive in the second. Li Feng estimates that roughly half of the second-quarter improvement came from a sharp swing in imports and exports caused by front-loading before tariffs: an unusually large deficit in the first quarter was followed by a major narrowing in the second, mechanically boosting the annualized quarter-on-quarter rate.
Inventories rushed out before tariffs took effect can support roughly 1 quarter of demand; price pressure will then feed back gradually. Deflation in Chinese industrial goods provides some buffer for US import costs, but cannot fully offset the tariffs themselves.
The disagreement in US capital markets is therefore not simply about whether the economy is strong or weak. No one knows where the net-effect midpoint lies among tariff inflation, weaker employment and easier monetary policy. Li Feng calls this the core variable shaping relative capital allocation between China and the US.
10. China also has a job-matching problem, not merely a shortage of jobs
Li Xiang offered the example of a well-run mid- to high-end restaurant chain in Beijing. Its servers and kitchen staff earn more than Beijing’s average wage, yet the company still struggles to hire. The owner sees unemployment headlines while facing persistent labor shortages—a snapshot of the mismatch between statistical employment and industry demand.
Li Feng adds that services account for 70% to 80% of US GDP. After excluding software, finance and similar sectors, broad services such as restaurants, entertainment and tourism still account for more than half of GDP. China’s gross tertiary-education enrollment rate is already above 60%; the US has a relatively high non-completion rate, while China’s completion rate may be higher for cultural, historical and family reasons. Ultimately, close to half—or even more than half—of workers in China’s broad services sector may have received higher education.
The knot in current employment attitudes has not yet been fully untied. Society is only beginning to change.
11. Anti-involution starts by dismantling companies’ soft budget constraints
Li Xiang relayed Sun Weijian’s explanation: price wars can continue indefinitely because companies obtain resources too cheaply. Even when an auto company is no longer viable, local governments may resist letting it fail because of tax revenue and employment, creating a classic “soft budget constraint.”
Long-form video platforms offer a counterexample. When financing was easy and the cost of capital low, they bid aggressively for content while refusing to charge users adequately. Once financing tightened, they strengthened membership pricing and operating discipline, and iQiyi became profitable. When constraints harden, companies shift from chasing scale to calculating returns.
Li Feng agrees that this is one core reason and points to the post-2015 capacity cuts as a historical reference. Manufacturing capacity utilization was once as low as roughly 56%; after a long adjustment, it returned to above 70% before the pandemic. Without that cleanup, the trade war’s impact on small and micro enterprises in 2019 might have been worse.
The cost cannot be omitted. The 2015–2017 capacity-cutting and destocking process was “quite disruptive.” Anti-involution can improve supply-demand balance and PPI, but it is not painless and will not show up immediately in prices or corporate profits after an August directive.
12. The implicit task of a unified national market is to reprice land, factories and tax incentives
The unified national market, repeatedly reinforced over the past year, is not only about eliminating local protectionism. It also requires resource use, tax incentives and competitive conditions in local investment attraction to become more uniform. Even if “policies from above, countermeasures from below” remain possible, the policy direction is to reduce local governments’ implicit subsidies to companies.
Li Feng uses 10 years of free land and government-built factories as examples. If these resources are provided without charge, they do not enter local government revenue and continue to suppress companies’ true costs. As higher authorities impose pressure for fair competition while local finances need revenue, the 2 forces will push up fees for using state-owned resources.
His explanation of non-tax revenue is also worth retaining. Headlines often equate it with fines, cross-regional enforcement and “distant-water fishing,” but breaking down the components shows that a large contribution comes from higher fees for using state-owned resources. This is not a separate topic from anti-involution; both are part of the same mechanism of repricing resources.
13. The industrial-chain damage from involution comes from squeezed margins and longer payment terms
Li Feng distinguishes industrial involution from ordinary price competition. Industry leaders not only demand lower upstream selling prices; they also demand continuously lower gross margins while extending payment terms from “3 months to 6 months to 9 months,” squeezing profits and cash flow at the same time.
China’s manufacturing chains are long and contain many B2B companies. That is an advantage of an integrated supply chain, but it also magnifies the transmission of pressure. Once leading companies begin squeezing suppliers, the shock worsens at every link, spreading layer by layer like a money multiplier in the banking system.
The result is not just falling PPI, but a contraction in the survival space of “at least more than half of companies.” Anti-involution must address this cross-chain appropriation of working capital and transfer of costs, rather than banning every price cut, promotion and new entrant from competing.
14. To determine whether the food-delivery war is involution, first ask who pays the RMB10 subsidy
Li Feng’s test is specific. If the platform pays RMB8.5 of a RMB10 discount and the merchant pays RMB1.5, the positive effect on merchants and demand may be larger. If the platform and small merchant each pay RMB5 during an already weak July and August for restaurants, the industry damage could exceed the incremental demand.
Couriers and delivery capacity will also be reallocated by the subsidies. Interviews at SF Intra-city Delivery found that some ride-hailing drivers had switched to delivery, possibly because short-term subsidies lifted their income, or possibly because ride-hailing supply had already become saturated.
Li Xiang argues that this instant-retail competition differs from industrial involution. Alibaba, JD.com and other platforms face clear budget constraints, are mainly spending their own money and do not have a low cost of capital. Whether this becomes destructive involution still depends on whether costs are forcibly passed on to merchants and suppliers.
15. A RMB5T restaurant market cannot remain bottlenecked by a single platform over the long term
Meituan previously held roughly 70% or more of the food-delivery market. Li Feng believes that in a roughly RMB5T restaurant industry connecting countless small, micro and midsized businesses, excessive concentration in a single link creates a “bottleneck” that suppresses upstream and downstream profits and innovation. It may look rational in the short term, but competition will inevitably challenge it over the medium to long term.
Ele.me would struggle to challenge Meituan on its own because consumers’ food-delivery habits are already anchored in Meituan. Taobao and Tmall are high-engagement apps: users are already browsing, making it easier to add the idea of “browse-and-eat” and quick commerce. JD.com has shorter usage time; its advantage is not browsing, but its existing ground-delivery network.
Li Feng compares this with QQ. High-engagement apps can keep layering on services such as music, games and QQ Show. Taobao has the same entry-point advantage; the difference is whether it is willing to take on the “tough, dirty and heavy work” of offline fulfillment.
For consumers, Li Xiang closes the discussion with a family complaint about the war: “We’re being forced to drink bubble tea every day.” Seeing a quick-commerce entry point leads people to buy more desserts and eventually gain weight. The joke reflects a real point: subsidies do create usage frequency, but whether that frequency persists depends on consumer habits and fulfillment.
16. Enormous industries ultimately move toward both multi-platform competition and vertical specialization
Instant delivery does not necessarily need to be built in-house by the entry-point platform. SF Intra-city Delivery initially served major brands such as McDonald’s and KFC, with an average price per order roughly RMB0.5–1.5 higher than crowdsourced or platform-operated delivery. When the food-delivery war tightened delivery capacity, lighter platforms began using SF’s network and cutting prices to win volume.
Restaurant supply chains already demonstrate vertical specialization. Early restaurants bought, cut and cooked their own ingredients. Central kitchens then emerged, followed by more industrialized food processing and prepared-food stages. As front-end efficiency improved, restaurants could focus more on seasoning, heating, dining environments, chain operations and service experience.
Formats have therefore diverged. Small menus, single-item hits and high-value-for-money chains are easier to scale. High-end restaurants survive on distinctive products and service. Traditional Chinese restaurants with complex menus and mid-range prices are under the greatest pressure, squeezed between industrialized low costs and high-end differentiation.
Didi and Amap show that disruption can come from either the service side or the traffic side. Amap uses map traffic to support many small and midsized ride-hailing platforms. Li Feng’s long-term view is that “an extremely large industry cannot form a high degree of monopoly in a single link” and will continue breaking the supply chain into smaller pieces.
17. Every additional RMB10T of market cap still requires several trillion renminbi of capital support
Li Feng offers a ratio for intuition, not a precise calculation: every RMB10T increase in capital-market market cap requires roughly RMB3T–4T of incremental capital, because holdings of founders and long-term funds may be tradable but are unlikely to be sold over the medium term.
The ratio changes with the market phase. After taking profits, some investors sell while others add, meaning RMB2T may be enough to push the market up another level. When consensus weakens and selling pressure rises, RMB6T may be needed to support the same market-cap expansion. “It still takes money to push the capital market higher.”
Regulators are also adjusting the market’s asset structure to bring it closer to the direction of economic transformation. Measures include the STAR Market, Hong Kong’s Chapter 18A, technology companies returning to the Shenzhen Stock Exchange, delistings, dividends, industrial M&A and new technology listings. Structural upgrading can improve the investable universe, but it cannot replace incremental capital.
18. A-share gains rely more on long money; Hong Kong’s funding mix contains more short- and medium-term capital
Passive public funds such as ETFs usually allocate across companies above the median size and provide a degree of stabilization. Insurers have longer liability durations, so their allocations to banks and high-dividend blue chips are closer to long money. Together, these flows explain the relatively stable large-cap A-share rally in the first half.
After the expansion of Stock Connect, the number and assets of renminbi public funds able to invest in Hong Kong stocks more than doubled from before “9/24.” Li Feng regards part of this capital as medium-duration money, but says he cannot accurately distinguish how much of total Southbound flow is long money, medium money or arbitrage-driven short money.
The A/H premium provides the clearest trading rationale. If the same company’s Hong Kong shares are roughly 30% cheaper, a fund already holding the A-shares will naturally buy the H-shares to close the gap. As the premium narrows materially, the marginal appeal of the valuation arbitrage declines. Southbound incremental flows will probably continue in the second half, but are unlikely to match the force of the first half.
The joke that “Hong Kong stocks are increasingly becoming A-shares” may also reflect the nature of the money. Repeatedly pushing up an AI, robotics or consumption concept before suddenly reversing suggests a high share of short- and medium-term capital, rather than long money building positions around slow-moving fundamentals.
19. For A-shares to keep rising in the second half, insurers and households must first be doing better themselves
Insurers still have incentives to raise equity allocations, but policy cannot focus only on the investment side; it must also improve the income side. Li Feng’s possible tools include adapting Class C drug catalogs to commercial insurance, developing individual commercial insurance and introducing tax benefits, deferrals and capital-gains arrangements so households are willing to convert part of their excess savings into long-term protection assets.
Household confidence is the harder variable. If a series of positive events changes expectations, both investment and consumption could be released from “reasonable savings” and “excess savings.” If confidence does not move, institutional buying alone will struggle to create a broader market and economic cycle.
Hong Kong stocks, meanwhile, need foreign capital to rotate from bonds into equities. Li Feng believes that global capital allocating only to the US for more than 5 years “cannot be a sustainable strategy,” so a return allocation to China will eventually occur. But he retains the key qualification: its speed, scale and trend will depend on China’s own efforts, US volatility and the direction of the dollar.
20. A true “DeepSeek moment” is a sequence of events, not another company appearing out of nowhere
Li Xiang questioned whether the “right time, right place and right people” behind DeepSeek could ever be replicated, and whether technological innovation could again create the same level of shock. Li Feng agrees that the probability of any single event is low, but stresses that expectations are usually changed by “a stack of consecutive events pointing in the same direction,” not by one policy acting alone.
The sequence in early 2025 included policy improvement, DeepSeek, Ne Zha, robots, the private-enterprise symposium and the future-industries plan from the Two Sessions. Together, these events pushed sentiment to a March peak before the tariff war interrupted the sequence. “DeepSeek moments” is therefore a metaphor for consecutive catalysts; another large model need not do the job alone.
Known domestic events include the SCO summit, the September 3 events and the Fourth Plenum of the 20th Central Committee in October to discuss the 15th Five-Year Plan. Li Feng believes speculation over whether Trump will be invited to the September 3 events has strong headline appeal, but the probability may be low.
One further detail is that constraints on some foreign institutions’ hedged index-futures programs are scheduled to ease in October. Li Feng sees this as a clue that regulators may be reserving an opening window, but explicitly labels it personal speculation. If the policy arrives after existing positive events, it could have an effect like the private-enterprise symposium after the Spring Festival—“twice the result with half the effort.”
21. 3 external variables determine whether the window can move from “50% controllable” to above 70%
China-US tariffs are the largest unresolved variable. Based on the latest extension, Li Xiang estimates that a meeting between the Chinese and US heads of state could take place by the end of October at the latest. Li Feng guesses it could occur sometime between September and mid-October if it does not happen on September 3, with a major agreement potentially signed at the leaders’ meeting. His personal guess for the final average tariff rate is 15%–20%; a more friendly outcome might be closer to 15%, but he stresses that there is no reliable basis for the estimate.
Once settled, the agreement would mean that most of the tariff arrangements proposed by the US in April had been completed. For China, it would mean that the risk attached to foreign trade, one of its economic “engines,” had been resolved. If the leaders’ meeting sends broader signals about bilateral relations beyond trade, the impact could be larger than the tariff number itself.
Fed cuts are the second variable. Li Feng says US easing cycles have historically tended to coincide with capital outflows from the US, while also easing China’s interest-rate differential pressure and creating more room for domestic cuts and other financial policies. If global liquidity also loosens, the renminbi and Hong Kong stocks would benefit.
Russia-Ukraine is the third variable. Trump is applying pressure, Ukraine has proposed a ceasefire, meetings and negotiations, and it remains uncertain whether Putin will attend relevant events. Winter conditions are also unfavorable for ground operations, creating a possibility of a phased ceasefire before October. If one occurs, pressure on China’s trade and relations with Russia, Europe and Central Asia could ease. But like the China-US talks, this is not an event China can control alone.
22. AI attention has shifted from model parameters to applications, hardware and revenue
Li Xiang observes that when Kimi, Qwen and Brock release new models, they no longer generate the kind of society-wide discussion seen earlier. Even the approaching GPT-5 release lacks the same heat: “Everyone has become accustomed to being shocked.” Attention first moved to agents, then to AI combined with hardware, robots and end-device applications.
Li Feng sees this as the normal path of a technology cycle. Competition shifts from “whose technology is the best” to “who can sell it, who can get it used and who can ultimately make money.” Autonomous driving is following the same path. The market no longer focuses only on theoretical leadership in L4 and L5, but on how many vehicles L2 and L3 can be installed in, how many scenarios they can cover and whether they are safe.
China has a particular advantage in AI hardware. Glasses, toys, earphones, robots and even autonomous vehicles are constrained by edge compute, latency and cost. They may not need the most powerful large model, but they do need supply-chain strength, product definition and commercialization capabilities. Events such as Insta360’s listing have further reinforced capital’s interest in “technology consumption.”
23. DeepSeek’s long-term contribution may be to lower the moat around pure-model companies
DeepSeek did 2 things. It rejected the premise that bigger models are always better, reversing the trend toward pure scale competition. It also open-sourced models with real capability and application value, opening a competitive window that prompted several other Chinese model companies to accelerate their own open-source efforts.
Open source benefits the industry but makes commercialization harder for model companies. As accessibility improves and technical barriers fall, “just building models and technology” becomes less likely to produce a lasting advantage. The first open-source company can capture ecosystem benefits, but pure technical scarcity across the industry is weakened.
Li Feng’s historical conclusion is that no internet giant has grown large by selling technology to B2B customers alone. Ultimately, every company must become a product company with a Matthew effect, scale effects or user-habit barriers. Search engines improve as more people use them; office software relies on switching costs. Model companies must also answer what their product loop is.
His China-US comparison is direct: “China is best at using applications to drive technology, rather than using technology to originally drive applications.” As capital moves simultaneously toward top-layer applications and bottom-layer chips, model companies stuck in the middle without products or scale effects will face greater pressure.
24. The primary market has moved from bargain hunting to selective competition for allocations
From 3Q22 through the end of 3Q24, the primary market was “very cold.” It began warming in 4Q24 as early-stage biopharma and some technology projects resumed financing. Technology companies could reach mid-stage valuations of several hundred million renminbi, but valuations above RMB1B remained rare, and pure consumption had not clearly returned.
The shift accelerated after May and June 2025. Technology consumption heated up, while early- and mid-stage projects increased simultaneously. More companies valued above $100M were raising new rounds, and projects valued in the RMB1B-plus to RMB2B range also became active. Li Feng’s team meetings went from having almost no projects during the cold period to needing to be capped at 3.5 hours.
The stronger signal is competition: “There is a 50% chance that a project we intend to invest in will require us to fight for an allocation.” Follow-on rounds for 4 to 8 existing investments also required the team and LPs to compete for capacity. The market is no longer merely searching by prior-round valuation for companies that have been dormant for 1.5 or 2 years but have not yet died.
Li Feng attributes this to the usual 1–2 quarter lag from the secondary market to the primary market. Hong Kong’s wealth effect, policy encouragement, gradually measurable tariff impacts, technology internationalization and listed technology-consumption hardware examples have all improved risk appetite. It is not yet a broad overheating cycle, but it has clearly moved beyond the “very mild” phase.
25. Primary capital is concentrating at the 2 ends of AI and in biopharma; the next question is whether long money can take over
AI projects are migrating toward 2 ends. At the top are applications and end products with validated demand; at the bottom are edge, cloud and inference chips based on architectures other than Nvidia’s. In the middle, sectors such as biomedicine remain where industry applications are clear.
The recovery in biopharma is driven mainly by license-outs. Multinational pharmaceutical companies are buying Chinese R&D and clinical pipelines, lifting Hong Kong biotech stocks. Li Feng cites QuantumPharm’s nominal roughly $6B BD deal as an example. Although the headline amount is long-dated and conditional, the company has already received an upfront payment.
If the consecutive events of August through October materialize, the capital market could gradually expand from imaginative but as-yet-unrealized AI and robotics themes into new consumption, the full healthcare and pharmaceutical chain, and other companies tied to food, clothing, housing and transportation. The essence is not a rotation of themes, but a shift in the nature of capital “from fast money to long money.”
This is the most important conditional statement of the entire episode: for the rally to continue, it needs more than policy and stories. It also needs improving insurer income, household confidence, active foreign reallocation and better earnings expectations for real-economy companies. Li Feng is forecasting a window that may form, not an unconditional rise.