Vol.207 Macro Conversations 100 | Year-End Special: The Logic of AI Investment and the 2026 Macro Outlook
Summary
Li Feng’s overall view is that AI could be a productivity revolution, but today’s mass enthusiasm cannot be explained by technological progress alone. The internet produced only a modest increase in total-factor productivity in the US from 1999 to 2003, while AI has already gone through four waves—big data, computer vision, AI drug discovery and foundation models; even if this time is different, technology still takes years to move from breakthrough to mass adoption. Whether you participate, how enthusiastically you participate and how quickly you move depends on whether you believe AI can genuinely do people’s work. If you do believe it is a productivity revolution, there is no need to panic, but you should start preparing because the evolution and diffusion process still has a long runway.
The foundation of this AI valuation cycle was the unprecedented monetary expansion of 2020–2021 and the forced concentration of global capital in dollar assets in 2022. The central banks of the 8 major economies expanded their balance sheets by roughly $12T in 2020; the Fed expanded by more than $3T in 8 months, equivalent to the total of 4 years and 3.5 rounds of quantitative easing after the 2008 crisis. With a roughly 3x money multiplier, new liquidity over 2 years reached tens of trillions of dollars. The Russia-Ukraine war took Europe out of consideration, China remained constrained by the pandemic, and the Fed rapidly raised its policy rate from 0.25% to 2.25%; ChatGPT appeared in Q4 2022 at exactly the moment it could become the narrative axis for “why should I rise, why can I rise and why should I rise this much?”
If the world neither resumes aggressive liquidity injections nor actively shrinks balance sheets in 2026, capital markets will shift from an incremental bull market to a “stock game.” Global nominal GDP was roughly $114T in 2025, against about $130T in total equity-market capitalization, touching the upper end of the Buffett Indicator at roughly 1.2x. Capital is beginning to rebalance from its extreme concentration in dollar assets, which helps explain Hong Kong becoming the world’s top fundraising market, the DAX outperforming the Nasdaq despite economic stagnation in Germany, and gold’s rise. “Nobody can guess the timing” of the US AI peak, but if OpenAI or Anthropic lists at an extremely high valuation, investors should watch for the pattern seen in 2000, 2007, 2015 and 2021: the listing of the largest representative company marking the end of the boom.
AI investment has entered its second half: the market is no longer rewarding model capability alone, but asking who can turn the technology into applications and make money from them. The first half ran from 2023 through H1 2024, while Agents and embodied intelligence inherited the biggest imagination premium but also became the hardest near-term directions to commercialize by trying to “do everything for you” in the digital and physical worlds. The US may lean toward software in the next phase, while China leans more toward hardware and concrete use cases; the real screening criterion is simple: “Who uses the technology to build an application and, ideally, makes money from it—that’s who gets hot.”
China’s relative AI advantage is not just catching up in models, but combining frontier algorithms, sensors, a complete supply chain and an unusually intense application market. DeepSeek matters first because its open-source approach, convergence and reasoning capabilities are friendlier to startups and application vendors, and second because it demonstrated at a critical moment that China remains technologically competitive; over the longer term, applications and data governance will decide the winners. Facial recognition followed a path from lagging at the start, to mass adoption, to overtaking the frontier in research. Li Feng expects the market by the end of 2026 may begin to form a consensus that Chinese autonomous driving has surpassed Tesla, because China has more sensor-equipped vehicles and greater data dimensionality, quality and volume: “The next breakthrough in technology’s second half comes from applications.”
The central geopolitical theme of 2026 could be a US retrenchment from global military commitments and a renewed focus on its American “front and backyard,” giving China a rare breathing window in a decade. Li Feng does not believe US-China competition will disappear, but facing the November midterm elections, the Trump administration may stop using “the biggest hammer” to apply direct pressure on China. Meanwhile, China is improving relations through visits by European, Korean and Canadian leaders and through global initiatives. If China’s international influence rises over the next 5–20 years, the biggest beta in investment and business will be internationalization: “If you take risks, move just a little faster than China’s diplomatic influence; if you don’t, move just a fraction slower.”
The renminbi is suited neither to a sharp depreciation nor to a sharp appreciation in 2026; the policy objective is to find a narrow balance among internationalization, purchasing power and export competitiveness. China’s trade surplus in the first 11 months has already exceeded $1T, while the export mix is shifting toward roughly 5M vehicles, RMB135.6B in innovative-drug pipeline licensing and roughly RMB1.5T in chips; higher-value-added products are less sensitive to the exchange rate. Li Feng summarizes the required state for renminbi internationalization as: “Maintain a certain expectation of appreciation, but do not fully deliver the appreciation.” The broader reform signals are gradual opening of the capital account, regional pilots such as Hainan and “using opening-up to promote reform.”
The core scarce resources 10 years from now may be energy and data, while the most practical strategy today is to be early to the market, but not early to the cycle. Data can be reused, circulated and priced differentially, but faces challenges around anonymization, governance, pricing and circulation. Shanghai’s public-data pilot and the organization of hospital clinical data may determine whether Chinese applications such as AI drug discovery can move up another level. Li Feng acknowledges that he stopped investing in robotics too early and was too aggressive on AI hardware; his advice to founders is to position ahead of the cycle, and his advice to parents is to “start with the end in mind.” He also cites “do the difficult but correct thing,” while noting that the phrase did not originate because of 左晖—左晖 merely made it more famous.
Deep dive
1. First decide whether AI is a productivity revolution; only then does participation have a reference point
Li Feng opened with Diamond’s explanation of social polarization: the internet and social media continuously reinforce views users already like, so people “only see what they want to see.” This monologue therefore deliberately offers a voice that may differ from mainstream sentiment rather than catering to existing conclusions.
He distinguishes an information revolution from a productivity revolution. Even in the US, where the internet was born, total-factor productivity saw only a modest one-off jump from 1999 to 2003 before broadly flattening out. The internet mainly improved the digitization and transmission of information; unlike cattle, steam engines, internal-combustion engines and electric motors, it did not directly expand human work capacity.
The most important question for the next 5–20 years is therefore not whether a particular model is good, but whether AI can genuinely do people’s work. If the answer is yes, adoption will still take a long time, so “you don’t need to be too anxious”; eventually it will enter every industry, but you need to start thinking about “exactly how to use it,” rather than interpreting “just start” as an unconditional guarantee.
2. AI has already gone through 4 waves, each an intersection of technology and industry
Li Feng began with his founding of Miaozhen in 2006. Big-data infrastructure was hot for the first 6 years; Google then taught machines to identify cats in chaotic images, driving computer vision and facial recognition. Uber’s dispatching vision, combined with Tesla FSD, pushed the wave toward autonomous driving.
After Go-playing, protein-structure prediction was still only a technological breakthrough. The pandemic suddenly made biomedicine a global focus, and the intersection of the 2 produced the 2018–2021 AI drug-discovery boom. Crystal Pharmatech, backed by Fengrui and listed in Hong Kong under Chapter 18C in 2024, is his example for illustrating this cycle.
Foundation models are already the 4th AI wave, not the first. Li Feng stresses that hot themes do not appear out of nowhere: they emerge when technology clears a threshold and collides with another industry undergoing rapid change. Investment opportunities often sit at that intersection.
3. New-energy vehicles show that productivity technologies offer participants a long window from controversy to adoption
New-energy vehicles were not written into the 15th Five-Year Plan as prominently this time. In Li Feng’s view, that does not mean they have fallen out of favor; China already has an end-to-end advantage, with new-energy vehicles approaching 60% of new-car sales, so they no longer need to be treated as a future industry awaiting promotion.
They were given meaningful emphasis in the 2015 “Made in China 2025” plan, and most well-known new-energy automakers were founded in 2014–2015. Public opinion continued to doubt that some of them would survive until around 2021. The clearest and largest-scale payoff did not arrive until 2023, when China became the world’s largest auto exporter.
For hundreds of thousands of auto-parts companies, any year during the 7-year period from 2013 to 2021 was still early enough to seriously pivot toward new-energy and hybrid supply chains. Li Feng uses this to answer AI anxiety: if you believe AI is a productivity revolution, start thinking about how to use it, because even when the direction of a productivity revolution is clear, penetration still takes a sufficiently long time.
4. The scale of liquidity creation in 2020–2021 was far beyond the market’s usual understanding of “liquidity”
In 2019, before the pandemic, global GDP was roughly $86T and total equity-market capitalization roughly $89T, close to a 1:1 ratio. US GDP was just above $20T, yet its stock market already accounted for more than one-third of global capitalization, reflecting its special role as the center of dollar assets and derivatives.
In 2020, the central banks of the world’s 8 major economies expanded their balance sheets by roughly $12T. The Fed alone expanded by more than $3T in just 8 months, equivalent to 4 years and 3.5 rounds of quantitative easing after the 2008 financial crisis, before even counting 2021.
Once central-bank funds enter fiscal accounts and commercial banks, they pass through the “deposit, lend; lend, deposit” money multiplier. Using a rough global multiplier of 3x, the nearly $20T of base expansion in 2020–2021 ultimately corresponded to tens of trillions of dollars in new liquidity: “Human history had never seen so much money printed in such a short time.”
5. The 2022 allocation dead end forced new funds into dollar assets
After the Russia-Ukraine conflict continued, Europe simultaneously exposed risks around energy security, supply-chain stability and national security. China was still under pandemic controls with low mobility. China and Europe each represented about 20% of global GDP, while the US represented about 25%; with both major allocation regions losing appeal, capital “could only go to the US.”
The Fed’s pace reinforced that choice. It raised rates by 25 bps in March 2022, then by 50, 75 and 75 bps in May, June and July, taking the dollar benchmark rate rapidly from 0.25% to 2.25%. Expectations of dollar appreciation and higher yields accelerated global purchases of dollar assets.
Li Feng calls this “extreme allocation” under historical conditions. A normal manager of $100B would not put more than 80% in one region, but when a huge amount of new capital suddenly had nowhere to go in Europe or China, it had to temporarily abandon the principle of balance.
6. ChatGPT supplied the upside narrative; real technology does not mean valuation is explained entirely by technology
Asset prices do not publicly say, “There is too much liquidity, so I have to rise.” They need to answer: “Why should I rise, why can I rise and why should I rise this much?” ChatGPT appeared in Q4 2022 and happened to become the central narrative for carrying global capital.
The market then saw companies with $1T, $2T, $3T, $4T and even $5T in market capitalization in succession for the first time. The “Magnificent 7” briefly had a combined market capitalization larger than the GDP of any country other than China and the US, and close to one-quarter of total US stock-market capitalization.
Private-market valuations were also pulled upward by public markets. OpenAI was discussed at more than $500B and reportedly could exceed $800B if a new financing closed; Meta’s roughly $17B deal related to Scale AI appeared bearable only because a trillion-dollar platform could view tens of billions as 1% of its market capitalization. Li Feng’s qualification remains constant: “There is technological change,” but the shape of valuations also comes from “too much money, and money making an extreme allocation.”
7. “American exceptionalism” can be explained by the transmission of financial and technology bubbles into GDP
Li Feng uses an intentionally rough decomposition to explain US resilience: finance and related legal and consulting services account for more than 20% of GDP. If Shanghai is used as a reference, high-tech manufacturing and services could add another roughly 30%, putting the two together above half.
Global capital first flows into finance, and the AI narrative then channels that money from finance into technology. US services overall account for roughly 87% of GDP; as long as the remaining food, entertainment, housing, transportation and other services continue earning money from finance and technology workers, macroeconomic data can look stronger than in other economies.
8. Without new liquidity in 2026, global markets will enter a stock-based rebalancing
Global nominal GDP was roughly $114T in 2025, while total equity-market capitalization approached $130T, equivalent to a Buffett Indicator of roughly 1.2x and already at the upper end of the so-called reasonable range. If central banks do not expand, the bubble stays broadly fixed; if they shrink, the bubble contracts; if they do neither, only “more A, less B; less A, more B” stock competition remains.
Since Q4 2024, the new US president and policy uncertainty have prompted global capital to reconsider its extreme dollar allocation. Li Feng points to Hong Kong becoming the world’s largest fundraising market in 2025, Germany’s DAX ranking among the top 3 performers and beating the Nasdaq despite economic stagnation, and gold’s rise as evidence that capital has begun rebalancing and reallocating.
Asked whether US AI has already peaked, his answer is “I don’t know.” But if OpenAI or Anthropic completes its final financing round and lists, investors should watch an “unfortunate pattern”: the booms of 2000, 2007, 2015 and 2021 all ended with the prospective listing or actual listing of the largest representative company.
9. The second half of technology investing asks only who can make money with AI
Li Feng divides every technology cycle into 2 halves. The first half asks only “who has the technology,” as in big-data investing from 2008 to 2011; the second asks “who used the technology to build an application and, ideally, make money from it,” with technological identity giving way to revenue and profit.
From 2023 through H1 2024, almost all discussion focused on foundation models. Once model capabilities cleared a threshold and became more linear, capital shifted toward general-purpose Agents and embodied intelligence, which had the greatest imagination premium—one aiming to do everything in the digital world, the other everything in the physical world.
The flip side is that “the greatest imagination” is usually also “the least likely to land in the short term.” The US may lean toward software next, while China leans toward hardware and concrete applications. In either market, the focus will shift from model demos to products that can charge users and, ideally, are already profitable.
10. DeepSeek proved China can compete, but back-end technology naturally favors large companies
DeepSeek has 2 layers of significance. First, its open-source approach, convergence and reasoning capabilities are friendlier to startups and application vendors. Second, at a critical moment when global capital urgently needed an AI narrative, it demonstrated that China at least has the ability to compete head-on in core technology.
Yet when ordinary users talk about Chinese models today, most are already using Doubao or Qianwen; in the US, users are gradually moving from GPT toward Gemini. Li Feng asks why foundation-model applications ultimately become dominated by large companies again: if innovation happens only in the middle and back end, companies large and small can adopt it, but startups struggle to become super-platforms on the strength of a general capability alone.
The historical pattern is that the US usually leads in the first half of technology cycles, the middle is competitive, and China often catches up through applications in the second half. To produce a new giant, back-end technology is not enough; the front-end UI and consumer habits must also undergo a major change at the same time.
11. Super-startups need simultaneous upheaval in technology, interaction and habits
ByteDance’s core technology is big-data recommendation, but Toutiao and Douyin became platforms worth hundreds of billions of dollars because mobile internet also changed human-computer interaction from keyboard input to “swipe left, swipe right, swipe up, swipe down.” Recommendation algorithms and touchscreen habits combined at the same moment to create a new information gateway.
Microsoft’s graphical operating system was not an isolated technology that large companies could not replicate. It was the mouse, which freed users from programming and keyboard commands, that brought graphical interfaces to the mass market. Li Feng therefore stresses that “other changes on the front end were equally profound”; only then can a technology company gain a platform-scale opportunity.
Tesla put AI into cars and autonomous driving, changing how users issue commands to machines and experience machine capabilities. Together, the 3 examples show that technological leadership alone can produce a good company, but may not be enough to produce a super-company of the era.
12. Embodied intelligence’s strength is movement; its weakness remains manipulation
China’s most eye-catching robot demos almost always involve walking, running, backflips, dancing and wrestling—in essence, balance and movement. What is still broadly missing is the hand and fine manipulation. The former is linked to technologies such as reinforcement learning and benefits from China’s already mature motor industry.
Inovance Technology initially captured demand from real-estate elevators by separating and simplifying controllers in foreign equipment that were difficult to install and repair. It then benefited from manufacturing upgrades and demand for new-energy-vehicle motors, including supplying Li Auto. Today’s capabilities in robot joint motors are not the result of 20 years of robot-specific R&D, but of multiple industries “laying eggs along the way.”
The chain shows that China’s hardware advantage is often not a single-point breakthrough. Huge markets in real estate, elevators, manufacturing and new-energy vehicles first drove costs down and performance up; new AI applications can then call on capabilities that already exist.
13. New applications will rewrite the chip landscape; AI will not belong to Nvidia forever
Nvidia originally solved the precision and efficiency requirements of graphics rendering for internet games; its parallel-processing capabilities were only later adapted to AI. After smartphones emerged, Qualcomm and Broadcom grew alongside Intel, showing that “new applications create new infrastructure.”
When AI genuinely enters earphones, smartphones, PCs and different industrial devices, those terminals cannot fit a data-center-grade Nvidia chip. Applications will also impose different requirements for power consumption, size, cost and real-time performance, creating room for new AI-chip companies.
The earliest large-scale users of Agents will often be existing companies. Their upstream, downstream and internal processes must be highly digitized, and delivery must contain extensive natural-language interaction. E-commerce, sales and RPA therefore have a better chance of monetizing early than low-digitization industries.
14. China is taking Japan’s “electronification” one step further into “intelligence”
In the 1980s, Japan electronified large numbers of mechanical products: mechanical watches became Seiko and Casio electronic watches; pianos became electronic keyboards; Kodak film cameras were replaced by electronic cameras. The result was not only product change, but lower prices and mass adoption for watches, instruments and cameras.
China is replacing internal-combustion engines with electric motors while pushing down vehicle and usage costs. Li Feng observes that per-1,000-person car ownership in parts of Central and South America, Eastern Europe, Central Asia and Southeast Asia has exceeded historical levels for economies at the same GDP stage earlier than expected, thanks to the cost advantage of Chinese EVs.
Unlike Japan at the time, China can add sensors, algorithms and software, upgrading electronification into intelligence. The Ministry of Industry and Information Technology’s coordination of 8 ministries to promote “AI + manufacturing” is the policy expression of the AI second half: making strong industrial chains intelligent one by one.
15. Sensors often mature elsewhere before being migrated into new products through the supply chain
Motors, optoelectronic sensors, cameras and lidar in smart hardware were almost never cultivated specifically for that product. They first gained scale, precision and cost improvements in larger markets, then migrated into new forms such as robots, watches and cameras.
When autonomous driving first became hot in 2015, 16-line lidar and the rare 64-line products could cost tens or hundreds of thousands of renminbi. Today, multi-line products have fallen to just over RMB1,000. Two early US lidar companies went bankrupt, while China’s RoboSense and Hesai continued cutting costs through domestic competition.
Smartwatch functions such as heart rate, respiration, blood oxygen and sleep monitoring came from optoelectronic components developed for smartphones and consumer electronics. The compact wide-angle and panoramic cameras needed by Insta360 also came from the spillover of smartphones pushing pixel counts higher and prices lower.
16. Insta360’s leap shows why integrated software and hardware suit China’s environment better than pure software
Insta360’s founder started a company after graduating from university in 2013, initially controlling image-cropping and stitching software. Facebook’s acquisition of Oculus made VR and 360-degree imagery hot, but after spending a year in China, the team found that in 2014 it was difficult to make money by selling software and technology alone.
From late 2014 to early 2015, he shut down the Nanjing company, moved to Shenzhen and combined image-stitching capabilities with the consumer-electronics supply chain. The first-generation panoramic camera still looked crude, but the supply chain enabled rapid iteration. By 2017, the product had entered Apple’s offline stores as the only non-Apple photography-equipment brand.
Insta360 then iterated 4 versions a year, using software to solve stabilization, stitching and distortion while catching up with GoPro at lower cost. It captured GoPro’s market in 2 years and became the global leader in another 3. Li Feng says the process looks simple, but the real difficulty was choosing to abandon the pure-software path.
17. Applications push technology forward; China often gains an advantage in the second half
Before 2017, China’s technology sector broadly believed facial recognition could not catch up with the US. But rapid adoption in hotels, boarding, security checks, identity verification, large payments and business-registration changes created conditions for continuous iteration. After 2019, Chinese papers occupied a significant share of top computer-vision conferences and journals.
Autonomous driving is similar. After China approved L3 in early December 2025, Li Feng believes the field will increasingly become “data-led and algorithm-supported.” By the end of 2026, the market may begin to form a consensus that China has surpassed Tesla, because domestic vehicles equipped with sensors are more numerous and their data has greater dimensionality, quality and volume.
Innovative drugs are entering the same phase. In 2025, the total value of Chinese pharmaceutical pipeline licensing contracts reached roughly RMB135.6B, more than half the global total. As end-to-end R&D efficiency improves, advanced chips, sensors, microfluidics, data, AI and virtual screening will be used more broadly: “The next breakthrough in technology’s second half comes from applications.”
18. China’s manufacturing industry is advancing from assembly to precision manufacturing, technology manufacturing and global brands
China became a major mobile-phone assembly base around 2000. After Apple appeared in 2008, some companies upgraded from low-end assembly to precision manufacturing, giving rise to Luxshare Precision, Goertek, AAC Technologies, Sunny Optical and Lens Technology.
Around 2013, China became the world’s largest country for both industrial-robot use and production, turning precision manufacturing into technology manufacturing. Consumer brands including Huawei, Xiaomi, vivo and OPPO also emerged mainly in 2013–2014. Huawei moved from carrier customization to a consumer brand represented by the Mate series.
China surpassed the US in 2012 to become the world’s largest single smartphone market. Devices that were cheap enough and good enough spread early; the country’s huge, demanding market then forced applications to compete on product quality, ultimately providing the foundation for Douyin to become TikTok.
Li Feng believes any industry can be placed on this path. New-energy vehicles are approaching the global-brand stage, while biomedicine is roughly at the precision-manufacturing stage of the smartphone industry before 2008. Companies need to determine not whether a path exists, but which rung they occupy.
19. US strategic retrenchment puts the Americas “front and backyard” first in 2026
The National Security Strategy issued by the US in November 2024 already clearly proposed reducing its global military presence. Li Feng uses “offshore balancing” to explain what comes next: the US will reduce direct troop deployments while using Japan, the Philippines, the UK, Russia and other forces to prevent a single dominant power from emerging across Eurasia.
Retrenchment will also pressure allies to raise defense spending, creating weapons-sales opportunities for the US. The balancing of China will not disappear; Japan’s recent behavior may even exceed Washington’s original expectations, but the approach will rely more on proxies and regional balancing.
In early January this year, a term appeared that embedded Donald Trump’s “Don” into Monroe’s “Monroe,” calling it the “Donroe Doctrine.” It symbolized a shift of attention toward Venezuela, Greenland, Canada, the Gulf of Mexico, Colombia and Cuba. Beyond oil and resources, the core idea is to “fence off other people’s yards as my yard,” consolidating the US sphere of influence in the Americas as it retrenches globally.
20. China may receive a rare external breathing window in a decade
Since China actively pursued destocking, deleveraging and capacity reduction and adjusted its property market in 2015, it has also gone through the ZTE and Huawei episodes, the 2020–2022 pandemic, pressure on relations with Europe after the Russia-Ukraine conflict, and a new tariff war. Economic restructuring has almost always been accompanied by major external pressure.
Li Feng’s 2026 view is explicitly conditional: US-China competition will not change, but the US may shift its priority to its immediate neighborhood and stop repeatedly hitting China directly with “the biggest hammer.” Trump also faces the November midterm elections; if he loses the House, his execution capacity and personal political risk over the following 2 years would rise sharply.
If previously arranged plans materialize, Trump’s visit to China in April and a visit by China’s leader to the US in May could also provide a buffer. “If possible,” this would give China a rare “bit of space” in a decade, allowing the handoff between old and new growth engines to occur under lighter external pressure.
21. Rising international influence will turn internationalization into a long-term corporate beta
The leaders of France, Germany and Italy, South Korean President Lee Jae-myung and possibly Canadian Prime Minister Carney have visited or may visit China. Li Feng sees this as a signal that China is using friction between the US and its allies to pursue medium- and long-term cooperation: after demonstrating military and technological capabilities, China is further demonstrating that it will handle international relations according to different principles.
China’s proposals for a community with a shared future for humanity, global development, cooperation, support for weaker countries, green development and carbon reduction provide an agenda for its rising international position that differs from traditional Western value narratives.
The investment implication comes from a sequence of questions. If China’s international influence, relationships and status rise over the next 5–20 years, more international business will inevitably emerge. “If you take risks, move just a little faster than China’s diplomatic influence”; if you seek stability, move “a fraction slower,” but long-term direction should not be determined solely by whether overseas business is profitable today.
22. Old and new growth engines may begin to hand off; the renminbi should rise or fall only modestly
Property and infrastructure once together accounted for more than 20% of China’s GDP, while new quality productive forces may have accounted for less than 10% in 2015. Replacing a huge old sector with a smaller new one inevitably creates a phase in which the old declines faster than the new grows, putting pressure on employment and distribution. Li Feng had long said he was still observing; he now leans toward a clearer handoff between the two in 2026.
Last year he opposed a sharp renminbi depreciation because it would damage renminbi internationalization, cross-border settlement, renminbi assets and property. This year he likewise opposes a sharp appreciation because policy must also balance exports, purchasing power, international trade relations and capital flows.
The trade mix is already less exchange-rate-sensitive. China exported roughly 5M vehicles in 2025; at more than $20,000 per vehicle, that exceeded $100B. Innovative-drug pipeline licensing reached RMB135.6B, while chip exports reached roughly RMB1.5T, or more than $200B. Their value added is far higher than that of goods relying purely on low labor costs.
23. The US is breaking the dollar cycle; China’s largest institutional gap is shifting toward finance
Li Feng divides national power into 5 layers: military strength, manufacturing, GDP, finance and global value propositions. He believes China’s problems in the first 3 layers are no longer significant; the largest gap is finance. After World War II, the US at one point accounted for more than half of global manufacturing and GDP, while helping rebuild global markets through the UN, IMF, World Bank and GATT.
After the US shifted from trade surplus to trade deficit, global demand for dollars increased, and the dollar detached from gold in 1971. Petrodollars, liberal market economics and dollar financial products then formed a cycle: the US used an overvalued dollar to buy global goods and export dollars, while surplus countries bought US Treasuries and dollar assets, lending the money back to the US.
The US now wants to retain military and financial advantages without continuing to bear the trade deficit. Through tariffs, subsidies for strategic industries, government stakes in chip companies, and coordination of energy and AI industries, it is rebuilding manufacturing. These “very China-like” state-led measures are actively dismantling the free-market cycle the US once dominated.
24. Financial opening will proceed through small-step experiments; the renminbi needs “expectation first, delivery later”
China did not fully open its capital account through the 3 exchange-rate reforms. At the end of October 2023, the Central Financial Work Conference proposed further opening of insurance, securities and funds to foreign capital. Li Feng sees household registration and population mobility as key clues for domestic reform, and exchange rates and financial opening under the capital account as the external reform track.
Trade creates demand for the renminbi first. Even as the Milei government has clearly tilted toward the US, China remained Argentina’s largest trading partner in 2025, with almost all trade settled in renminbi. Chinese manufacturing also covers roughly 60%–70% of local daily necessities, helping inflation fall from above 110% to the 30%-plus range.
After China’s trade surplus exceeded $1T in the first 11 months, it needs appreciation expectations to attract holders of the renminbi, while also raising purchasing power, expanding imports and improving international relations. Li Feng’s nuanced formulation is: “Maintain a certain expectation of appreciation, but do not fully deliver the appreciation,” meaning the country should not use up all the room at once.
The Hainan island-wide customs closure, free-trade-port tax incentives, exemption from import duties when processing value added exceeds 30%, and Yangpu Port construction could make Hainan an opening-up pilot zone similar to early Shenzhen. Financial pilots are only Li Feng’s speculation—“I have absolutely no idea”—but Shenzhen initially solved land problems for Hong Kong-funded factories and unexpectedly gave rise to land-transfer experiments, showing that opening-up often produces institutions beyond the original design.
25. The next decade’s competition will come down to data and energy; personal decisions should also start with the end in mind
In 2023, China established a new financial regulator and the National Data Administration, then began a Shanghai pilot for the centralization, organization and use of public data in hospitals, education and finance. Unlike money, people and land, data can be reused repeatedly, circulated and used at different prices, but it is also “hard to govern, hard to manage, hard to price and hard to circulate,” and must be anonymized.
Hospital clinical data can improve innovative-drug R&D efficiency, but it is often fragmented, scattered, dirty and disordered. A single data record in the US can cost more than $100 to more than $1,000. Who handles anonymization and structuring, decides the recipient and price of authorization, and combines the data with laboratory and research data will determine whether China’s AI drug discovery can move up another level. Li Feng believes future competition may ultimately condense into “energy and data.”
On investment timing, he advocates “be as early as possible to the market, but not too early.” Fengrui began investing in robotics in 2022 and stopped in August 2024, later acknowledging that it underestimated the extra 1.5 years of heat generated by the Spring Festival Gala and the symposium with private entrepreneurs. It began positioning in AI hardware in early 2025, even though the theme had started heating up around October 2024, and still plans to invest through after the 2026 Spring Festival because many of these companies are already profitable.
On AI for Science, he compares the opportunity with China’s manufacturing industry after joining the WTO: once demand arrives, companies immediately seek to “make things faster, make more of them and make them differently,” naturally adopting chips, sensors, microfluidics, data, AI and virtual screening. For the next generation’s education, he believes the foundational capabilities of mathematics and physics and the higher-level abilities of creativity and expression will be harder to replace, while intermediate processes are easiest for AI to compress. If China is more internationalized 20 years from now, children will also need international knowledge, perspective and experience.