Macro Talk 80: U.S.–China Contest and Large Models in Robotics

Summary

• Front-loaded government bond issuance is the only mildly unusual signal in the January–February data. Government liabilities accounted for an unusually large share of February aggregate social financing and grew sharply year on year—even against last year’s post-holiday economic recovery month, a comparison that should have worked against it. You can translate that as clear evidence that this year is front-loading fiscal support, and it happened before the Two Sessions opened. The rest—weak borrowing, stronger deposits, and a wider M1–M2 scissors gap—largely reflects the Lunar New Year shift; looking at January and February together offsets roughly 60%-70% of the seasonal distortion.
• Consumption is diverging, with lower-tier cities stronger. By 2024, household consumption in cities below Tier 3 had recovered past its 2019 pre-pandemic level, while Tier 1 and Tier 2 cities had not statistically returned to it. Higher-tier cities are being squeezed by income expectations, property-price volatility, and employment; Tier 1 cities were less exuberant in good times, but have proved more resilient.
• Shipbuilding, ports, and sea power form a single chain. China’s share of global order books for large-tonnage vessels ranges from at least 55% to more than 70%, almost completely displacing South Korea, which had held the lead for nearly 20 years. The U.S. has proposed port fees of roughly $1M per Chinese vessel or Chinese-built vessel and placed more than a dozen Chinese shipbuilders on a Defense Department sensitive list. Feng Shu’s speculation: the Li Ka-shing port transaction became sensitive because ports are the economic landing point for this sanctions chain—“I’m just guessing.”
• The Taiwan Strait is back on the table as a bargaining chip, while rare earths are China’s hard leverage. The State Department revised its Taiwan language, Beijing’s tone hardened (“sooner or later we’ll take you” made it onto CCTV), and rumors of an April U.S.-China leaders’ meeting were twice met with “no information to announce.” China’s two rare-earth controls at the start of the year were “stricter each time,” while Washington rushed a Russia-Ukraine minerals deal, moved to open domestic critical-minerals production, and may be seeking Greenland’s resources. “If any deal can be reached, it would probably be something that exceeded almost everyone’s expectations by a little”—and would repair trade expectations torn apart by tariffs.
• In embodied intelligence, the story is the brain, but the progress is in the joints. Capital is buying the “brain of the robot,” but the viral side-flip demos have little to do with the brain; they reflect hard advances in degrees of freedom and joint motors. The financing pattern resembles China’s 2016 EV startups: long lists of investors each putting in a little, with nobody willing to make a large bet. BYD will 100% build robots, Huawei 100%, DJI 70%-80%—the most capable players have yet to enter.
• AI applications come down to three choices. Should large models replace people, become mass-market tools, or serve as assistants to professionals? Feng Shu sees the third as the most realistic: a powerful assistant for capable people; full replacement remains “quite far away.” The most active market today is actually DeepSeek all-in-one machines, driven by people who “want to try it first.” “Whether an agent is a technical term or a business-model term is a very big distinction.”
• FSD has lost its lead in China. When Tesla began rolling off its China line in 2019, “there was simply nobody in China’s new-energy vehicle market who could match it”; now FSD is entering a China that has already gone through several rounds of intelligent-driving competition, and “it absolutely has no obvious lead.”
• The direction of “red-and-expert” distribution policies is worth reading as a policy signal. JD.com providing riders with the full package of five social-insurance programs plus a housing fund, major South China tech companies requiring lights-out times, clocking out, and two-day weekends, and a large Shandong appliance maker—Li Xiang said it was “probably Haier”—being cited for its two-day weekend all align with the call for workers’ incomes to grow faster than GDP. A thicker safety net “makes people willing to spend,” a strategy the Biden administration also used in its second half.

Deep dive

1. January–February financial data: the noise is the Spring Festival; the signal is government debt

• Feng Shu first dissects the emotional readings. During the Lunar New Year month, corporate deposits grow slowly or even fall as bonuses are paid, while household deposits rise accordingly; “you generally don’t buy a home during the Spring Festival month.” Weak lending and strong deposits are mostly calendar effects, while M1’s definition changed on January 1 and the scissors gap was bound to narrow. Combining January and February offsets roughly “60%-70%” of the holiday distortion.
• The only mildly unusual signal is that government liabilities in February aggregate social financing were both a large share and fast-growing, even though the comparison base was last year’s post-holiday economic recovery month and should have been unfavorable. “This year clearly needs to front-load support,” and it had already happened before the Two Sessions. Before August last year, government bond issuance was far below normal as authorities respected debt red lines and focused on resolving hidden debt; this year the front-loading is obvious.
• The sharp month-on-month drop in exports reflects the holiday as well as a high base from last year’s front-loading of shipments ahead of tariffs. If tariffs are feeding through in earnest, “more of it will show up in the second quarter,” with the key date pushed back to April 2.

2. Consumption is diverging: below Tier 3 is above 2019, while Tier 1 and Tier 2 are not back

• National Bureau of Statistics data by city tier corroborate colleagues’ Ne Zha report: by 2024, household consumption in cities below Tier 3 had slowly recovered past 2019 levels; even two years after reopening, Tier 1 and Tier 2 cities had still not statistically returned to their pre-pandemic levels.
• Feng Shu’s explanation is that higher-tier cities have more large tech companies and SOEs, so wages are higher but cyclical swings are amplified. Expectations are being jointly suppressed by “the economy, future income, property-price volatility, and employment,” pushing households to become more conservative. Lower-tier cities have recovered better overall. Tier 1 cities are not as good in good times and do not outperform other cities, but “they are actually more resilient”: volatility is lower, with no extreme swings between optimism and pessimism, which makes recovery easier.

3. JD.com’s food-delivery push cannot move Meituan, but social insurance is “red and expert”

• Li Xiang asks whether JD.com’s entry into food delivery can ultimately affect Meituan. Feng Shu’s answer: “The possibility is relatively small.” Meituan has built a strong user experience, and it is difficult for a new entrant alone to break established habits.
• The real signal is Liu Qiangdong’s social-security post and JD.com’s decision to provide riders with the full package of five social-insurance programs plus a housing fund. The sector has traditionally used outsourcing, dispatch arrangements, minimum contribution bases, and contracts with individual businesses—“many, many methods”—to push away costs and liability. JD.com’s approach “is, directionally, a red-and-expert move,” clearly aligned with the call around the Two Sessions for workers’ incomes to grow faster than GDP.

4. Receiving and the feeling of receiving: the safety net determines whether people spend

• Feng Shu distinguishes between receiving something and feeling that one has received something. Even without a pay raise, people can feel better when work intensity falls and the safety net gets higher. “When the safety net isn’t high, you keep thinking about saving for emergencies,” the same logic behind saving for old age or putting money aside in case of illness.
• The Biden administration offers a comparison: the first half focused on direct cash transfers, while the second half increased spending on education, healthcare, and other programs to thicken the social safety net and “make people willing to spend”—precisely the part Trump is now cutting heavily. Services have a higher wage share than manufacturing, so putting riders on full benefits serves two purposes at once.

5. Anti-involution, Haier’s two-day weekend, and Mexico’s working-hours contrast

• Major South China tech companies are mandating lights-out times and protecting two-day weekends. A large Shandong appliance maker—Li Xiang said it was “probably Haier”—was also cited for adopting a two-day weekend; one explanation is that EU market access requires an absence of forced labor, and some European countries already have three-day weekends. But companies that are still growing rapidly “are no longer just working 996—they are still operating as usual.” Whether a company joins the rat race depends on its growth rate and room to expand.
• Li Xiang describes what he saw in Mexico: Chinese bosses could not adjust to local workers not answering calls after work and explicitly rejecting overtime. A check showed that Mexico legally has a six-day workweek, while China moved to five days long ago; “even so, we still feel their working hours fall short of what Chinese companies expect.”
• The counterexample came from ByteDance during the pandemic, when overtime restrictions prompted employees to send begging memes saying, “Please, I’m a ByteDance employee.” Some workers believed they needed overtime to raise their incomes; policymakers also want to free up white-collar workers’ time to consume. “You’d think the money they earn simply leaves them no time to spend it”—the company even served four meals a day.

6. Port fees, sensitive lists, and Li Ka-shing: the economic landing point of the sanctions chain

• In February, the U.S. proposed special fees on Chinese vessels and even “Chinese-made” vessels docking at U.S.-controlled ports: $1M per vessel or roughly $1,000 per ton of cargo. The bill has not passed, but it has been proposed; the U.S. then added more than a dozen Chinese shipbuilders to the Defense Department’s sensitive list.
• The underlying issue is the shipbuilding landscape. China has “almost completely displaced South Korea” in large-tonnage vessels, after South Korea held the lead for nearly 20 years. Depending on the methodology, China’s share of global order books is at least 55% and more than 70% at the high end. “China’s capability, efficiency, and cost in building large ships have probably reached a level of competitiveness that is hard to imagine.” Civilian leadership translates into rapidly expanding naval strength.
• Feng Shu explicitly labels this a guess: if a large number of ports are under U.S. control, the law’s economic and partial military impact would be magnified. “I don’t know whether this whole chain of events ultimately has an economic landing point in the transfer of control over ports.” That may explain why the transaction involving more than 100 ports linked to CK Hutchison—heard in the original as “Changshi”—became a focal point, rather than simply a story about the politics of a “national entrepreneur.”

7. History’s rhyme: from the Spanish Armada to the tonnage race

• When Britain defeated the Spanish Armada, “the Spanish Armada at that point was no longer built by Spain itself.” An inflow of silver shifted manufacturing, drove inflation, and left services catering only to the wealthy. Between the 2 World Wars, major powers raced in military tonnage and eventually signed agreements imposing limits; by then U.S. shipbuilding had far surpassed Britain’s, alongside the rise of U.S. steel, including Carnegie’s.
• Tonnage still matters today because it determines ammunition load and strike range, but “you can’t rely on thick armor to absorb hits.” The Type 055 large destroyer “has already surpassed the capabilities of some European countries’ aircraft carriers.” After a Chinese fleet exercised near Australia, France sent an aircraft carrier to patrol the South China Sea, prompting Chinese internet users to joke that “in both combat capability and tonnage, it doesn’t match our destroyer.”

8. Large aircraft are replaying the large-ship script

• Boeing is beset by problems involving Air Force One, the 747, and spacecraft docking; Trump gave it a lift with the F-47 contract. Feng Shu says he is a military outsider and only sees “a sixth-generation fighter on a PowerPoint.” C919 is already in service and C929 is coming soon. “On the timeline, this is probably close to where ships were nearly 20 years ago,” raising the possibility of a repeat of the script in which China captured more than half of global orders within 10 years.
• Some of the low-altitude projects they invested in were staffed by former members of the C919 design team. C919 uses imported engines and flight-control systems “mainly so it could achieve airworthiness certification”: certified components shorten the path to approval. China has the ability to develop everything in-house, as its fighter jets demonstrate, but testing a fully self-developed system from scratch through commercial certification “would take a very, very long time.” The military side is already domestic; commercialization speed is the constraint.
• A side recommendation for nervous flyers: colleagues who fear flying should specifically choose C919. To prove safety and economics to the world, “it will do everything possible to become the best-flying, safest, and highest-customer-satisfaction aircraft,” while the cabin crew and flight attendants have been “raised by half a level.”

9. The low-altitude economy: an aerodynamic question with no global consensus

• The low-altitude economy was written into the government work report as a future industry, but low-altitude aircraft face several unresolved problems. Fixed wings offer the best energy efficiency through thrust-to-lift performance—the Coandă effect means 1 kilogram of thrust can produce roughly 3 kilograms of lift—but they require runways and trained pilots, conditions cities cannot provide. Rotors, as on DJI aircraft, are easy to control and can be autonomous, but “their energy consumption is too high” to support routine flights carrying 1, 2, 3, or 4 people per aircraft.
• The industry ultimately needs “a way that has never existed before or was never widely used,” and must rethink its energy source as well. Fuel-powered systems raise safety concerns, while electric aircraft are constrained by battery weight and endurance. The passenger market could be enormous—three-dimensional transport with massive capacity—but everything remains exploratory.

10. The Taiwan Strait is back on the table: bargaining-chip logic and the April summit question

• Two changes came from opposite sides: the U.S. State Department’s website revised its Taiwan language, while China’s Defense Ministry and foreign-affairs statements “became more serious and more forceful.” CCTV deliberately retained the phrase “sooner or later we’ll take you.” At 2 press conferences, questions about an April meeting between the U.S. and Chinese leaders received the same answer: “I don’t have any relevant information to announce at the moment.”
• Feng Shu uses a framework from a longtime Chinese American friend: the U.S. is retreating from international strategy while expanding around its own periphery—Canada, Greenland, Mexico, and the Panama Canal—so domestic consolidation temporarily ranks above external projection. Trump’s pattern in his first term was that “every time he was about to negotiate a deal, he first marked out his bargaining chip.” On Taiwan, “they use the same chip every time, and can keep using it back and forth, because it is your pain point.” Kissinger’s books explicitly identify it as one of China’s red lines.
• The meeting not being made concrete has 2 possible explanations: it is simply a rumor, or “before the meeting, they have to learn from past experience and lessons and first agree on something.” A meeting between top leaders is only used to announce results already reached. Wang Yi’s recent meetings with U.S. visitors to China and events such as the China Development Forum are supporting signals.

11. Rare earths are squeezing the U.S. by the throat; China’s posture has become reciprocal

• China’s 2 rare-earth export controls at the start of the year were “stricter each time.” Against that, the U.S. has 3 relevant moves: pushing hard for a Russia-Ukraine ceasefire while repeatedly emphasizing a minerals agreement, signing something that may loosen restrictions on domestic critical-minerals production, and possibly seeking resources in Greenland. “You can infer that our rare-earth restrictions are constraining them quite severely.”
• The State Council has just introduced legislation related to the Anti-Foreign Sanctions Law. China’s posture this time looks “somewhat tougher, or somewhat more reciprocal,” than during the 2018-19 technology and trade wars; it is no longer entering negotiations from a position of assumed concession. Trump’s chips are Taiwan and tariffs, while the areas where he wants to make money are tariffs, energy, and agriculture. China can also negotiate by buying energy and agricultural products: “what everyone wants is not quite the same, so theoretically the room to negotiate is quite good.”
• The market implication is the hardest line in the episode. The issue has been consistently denied and has not been widely discussed privately. “If any deal can be reached, it would probably be something that exceeded almost everyone’s expectations by a little.” It would repair expectations that Trump’s tariffs have “torn into pieces,” much as DeepSeek and other AI events reset expectations for China’s technology competition.
• The amateur-hour interlude: Musk went to the Pentagon for a meeting and Trump immediately denied it; the editor in chief of The Atlantic was pulled into a group chat discussing strike deployments against the Houthis. Li Xiang: “It confirms that the world is a ramshackle operation”（世界是个草台班子）. Feng Shu, meanwhile, singled out Israel as “truly deserving of condemnation”: after a prisoner exchange it launched an extraordinarily large attack, struck U.N. facilities, and still described it only as an “accidental hit.” The situation is worse than before the first-phase ceasefire.

12. The hottest AI application is the all-in-one machine—whether people can use it, let alone use it well, is unclear

• Li Xiang asks whether AI applications are genuinely active. Feng Shu says they are, but the most active segment is currently “all-in-one machines” that directly package DeepSeek. Demand comes from “people who are mostly unsure whether they can use it or use it well, but want to try it first.” As for actually deploying it and getting good results, “it’s still hard to say.” “It’s not that it can’t happen; it certainly will eventually, but the results and trend are not yet obvious today.”

13. The 3-way AI startup choice: replace people, serve the masses, or assist professionals

• Feng Shu’s core question for application entrepreneurs is where large models should sit today—not in the future: fully replace people, as with industrial robots developed over nearly 50 years, where a production line “doesn’t need many people”; become simple tools for most people, as PCs were in their early days and search is today; or become powerful assistants to people with professional expertise, as with CNC machine tools, which require someone who understands both machining and programming.
• His judgment is that today there are only “very few opportunities” to become a mass-market tool—search is the established exception—so “quite a lot should go into the second category”: assistants to professionals. Full replacement “is still quite far away.” One additional warning: “Whether an agent is a technical term or a business-model term is a very big distinction.”
• The same applies to hardware startups. First build a good pair of glasses, a good pair of headphones, or the most interesting toy, then add some intelligence; that “may be more reliable than deciding today to build an emotional-companion robot that replaces your boyfriend, girlfriend, or best friend.” Trying to realize the grand vision immediately “could end very badly.”

14. Embodied-intelligence demos show hardware progress, not brain progress

• Investors are most willing to buy the story of “the robot’s brain”—transplanting the large-model narrative into embodied intelligence. But the viral side flips and handstand-to-stand demonstrations “actually have little to do with the brain”; they show hardware progress in degrees of freedom and joint motors, including high torque and balance support after a jump. Feng Shu’s old joke is that making a robot with feet squat fully and then stand from the squat are both extremely difficult, because the bottleneck is the motors in the knees, ankles, and hips.
• 3 developments before and after the Lunar New Year changed the consensus. First, adding advanced software capabilities to hard industrial supply chains to create new markets: “this is something China likes, is good at, and can ultimately win at against almost the whole world,” a view even the U.S. shares. Second, embodied-intelligence robots entered the government work report as a future industry for the first time. Once the first 2 points became consensus, state and private capital rushed in. Robots going viral and appearing on the Spring Festival Gala added another layer of heat. But “are they being used, are they being used well, and how much are people willing to pay?” remain unanswered; today’s robots “haven’t even fully managed the task of looking like a person.”

15. This looks like China’s 2016 EV startups: the strongest players have yet to enter

• The financing landscape consists of mid-stage rounds at valuations of several hundred million yuan, over RMB1B, and low single-digit billions, with “many people pooling money together under every round.” For now, nobody is willing to make a large bet—exactly like the new-energy vehicle startups after their second round. Li Xiang argues that the EV supply chain was already mature at the time; Feng Shu counters that in 2016-17, “very few people even believed new-energy vehicles could be used and become widespread.” People doubted both consumer demand and whether companies that had never built cars could use China’s supply chain to build them.
• The key judgment: BYD will 100% build robots, Huawei 100%, DJI 70%-80%. The players closest to the core capabilities—and among the most capable robot makers—have not formally entered, and will be the last to do so.
• Tesla is the cautionary example. By mid-2024 it had delivered the first wave of robot hype; by the end of the year it had not delivered on it, and expectations were cut from December onward. “It probably would have cooled off gradually at that point,” but China’s technology stories around the Lunar New Year, including robots on the Spring Festival Gala, reignited the fire. “Tesla’s grand blueprint has indeed delivered on many things—full self-driving has not been delivered, and robots have not been delivered either.”

16. FSD enters China: from unmatched to no obvious lead

• When Tesla began rolling off its China line in October 2019, it had “a lead in China’s new-energy vehicle market that nobody could even come close to.” Today FSD is entering a China that has already gone through several rounds of intelligent-driving competition, and “it absolutely has no obvious lead”—even if localization and data issues are resolved. More than 1 month after FSD entered China, it was initially attacked as “a complete mess” because of missing map data; Li Xiang guesses that “after driving it for a while, people found it was actually okay.” BYD’s early reputation for making intelligent driving broadly accessible was “pretty good,” though he cautions that large volumes of such vehicles have not yet reached the market.
• The structural reason is that intelligent driving builds on electric-drive architecture. China’s breadth and depth of electrification “should be far beyond anyone else’s,” layered on top of the sensor build-out and the intelligent-driving race that began last year. Feng Shu asks whether this is an inheritance from the previous AI bubble; Li Xiang explains the electric-drive architecture, electrification base, and accumulated intelligent-driving capabilities. It is the same model as robotics: advanced software technology combined with a long industrial chain.