E229｜From Handmade Workshops to Extreme Manufacturing: How China’s Power Batteries Built a “Moat”
Summary
- China’s power-battery takeoff was not driven by a single technical breakthrough, but by policy creating demand first and then giving domestic manufacturers a window to grow. 张从志 sees the 2009 “Ten Cities, Thousand Vehicles” program as the demand inflection point; 何倩然 points to the 2015 “whitelist,” under which only new-energy vehicles using domestic batteries qualified for subsidies, shutting Panasonic, LG and Samsung SDI out of the system. By 2023, China took just 17 months to go from its 10 millionth to its 20 millionth new-energy vehicle, while complete-vehicle exports also surpassed Japan.
- The real moat is not who invented the material first, but who can execute micron-scale manufacturing at PPB-level quality. Errors accumulate from materials and coating through assembly and Pack, with the entire system ultimately constrained by the lowest-capacity cell—the “barrel effect.” Tier One manufacturers are working to move from PPM to PPB: one defective cell for every billion produced. The hardest front-end coating process determines thickness, density, dust and burr control, which is why buying all the equipment still cannot reproduce CATL’s or BYD’s yields.
- China now leads not only in end-market share, but across almost every critical industrial link in the power-battery chain. In 2022, China accounted for roughly 70% of global cathode-material shipments, more than 90% of anode materials, 83% of separators and over 86.7% of electrolytes; domestic lithium-battery equipment penetration exceeded 90%, while localization of key processes surpassed 80%. 何倩然 describes Chinese companies as “leading the race and controlling the pace,” while the US is like “a very smart but not hardworking student”: rich in original technology, but unable to push TRL and MRL from early-stage Tier 1 and Tier 2 to 8-9.
- Starting from a position of low cost forced Chinese companies to disassemble entire production lines, turning scarcity into process knowledge that cannot simply be bought. BYD replaced Japanese equipment with labor, fixtures and small desiccant-filled boxes, cutting a roughly $5M production line to $140,000; ATL spent about $1M on a Bell Labs license and then ran into gas-generation problems because it “couldn’t just throw it away,” ultimately becoming one of the few among more than 20 licensees to persist and solve the problem. 何倩然’s conclusion: all know-how comes from “real hands-on” experience.
- The contest between cylindrical, Blade and prismatic cells is fundamentally a trade-off between manufacturing capability and application boundaries, not a simple energy-density ranking. Tesla evolved from more than 7,000 18650 cells to 21700, 4680, tabless and structural batteries, extending a mature cylindrical production line; BYD rebuilt its equipment and processes for Blade batteries, using higher volumetric utilization to preserve the LFP route. Prismatic cells also retain flexibility for battery swapping, recycling and energy storage—more options, in 何倩然’s view, than “welding the entire chassis shut.”
- Solid-state batteries remain a long-term R&D project; over the next 5-10 years, sodium-ion batteries are more likely to become a scaled commercial variable. The guests believe solid-state commercialization “may still have to wait until 2030,” and even then it would not immediately replace liquid lithium-ion batteries; QuantumScape may still need roughly 3 atmospheres of external pressure, while interface separation, cliff-like decay, supply chains and new safety standards remain unresolved. By contrast, some sodium-ion chemistries contain no lithium, cobalt, nickel or even copper, and polyanion systems are said to reach roughly 30,000 cycles, making them suitable for stationary storage.
- The investable opening for Europe and the US may lie not in replicating China’s wet-process lines, but in switching platforms or shortening the supply chain. 何倩然 favors dry electrodes, direct lithium extraction and battery recycling; Ascend Elements has extracted lithium carbonate from black mass, while one Lithium Americas project spent 7-8 years advancing without producing output. Northvolt failed despite raising about $20B and securing roughly $50B in orders. 曾毓群’s summary is the sharpest: “You use the wrong design, use the wrong process, and then you use the wrong equipment.”
Deep dive
1. Three inflection points turned technical experiments into a scaled industry
杨璐 sees 2023 as the year market confidence was confirmed: China’s 20 millionth new-energy vehicle rolled off the line in Guangzhou, just 17 months after the 10 millionth; complete-vehicle exports also surpassed Japan that year. The deeper shift, in her view, was the industry’s move from “trading market access for technology” to indigenous innovation.
张从志 takes the timeline back to the 2009 “Ten Cities, Thousand Vehicles” program. Technology and markets were both immature at the time; the central obstacle was a lack of demand. Subsidies “created demand by brute force,” giving companies, talent and capital a reason to enter and making a subsequent innovation ecosystem possible.
何倩然 sees the Ministry of Industry and Information Technology’s 2015 “whitelist” as more decisive: only new-energy vehicles fitted with domestic batteries qualified for subsidies, excluding Panasonic, LG and Samsung SDI. She estimates the subsidy was close to $400/kWh at the time, while explicitly warning, “I don’t know if that’s accurate.”
She compares this long-running policy support with the US: the IRA was not passed until 2022, was less forceful and lasted roughly two years; after Trump took office, many policies were rolled back, and the OBBBA also canceled many EV subsidies. Her core judgment is that policy continuity gave Chinese battery makers a growth window that Japanese and Korean manufacturers could not enter.
2. China has moved from chasing to “leading the race and controlling the pace”
何倩然 uses a marathon analogy to describe the national standings: Chinese companies are “leading the race and controlling the pace,” while Japan and South Korea are in a stable but increasingly constrained second tier, and the US—having fallen behind—is trying to change tracks, switch platforms or rebuild its base.
China’s advantage is its depth of coverage. From lithium, nickel, cobalt and graphite to cathodes, anodes, electrolytes and separators, then cells, Pack and BMS, the guests consider China the only country that has achieved scaled commercialization and continued iteration across every critical link at the same time.
杨璐’s 2022 figures put that coverage into concrete terms: cathode materials accounted for roughly 70% of global shipments, anodes more than 90%, separators 83% and electrolytes over 86.7%; domestic lithium-battery equipment penetration exceeded 90%, and localization of key processes surpassed 80%.
The global top-10 battery-maker rankings are also being reshuffled. 张从志 observes that Chinese, Japanese and Korean companies were present in roughly equal numbers in the early years; after 2015, CATL gradually took the top spot, and from 2020 or 2022 onward the top three were mainly CATL, BYD and LG, with more Japanese and Korean players retreating.
3. “Hand-built” was not a joke about backwardness, but the starting point for China’s process knowledge
BYD initially could neither afford Japanese equipment nor tolerate long delivery times, so it replaced automation with “manual labor plus fixtures.” Processes requiring a dry room were converted into small boxes filled with desiccant, with workers reaching their hands inside to operate them. That production line later won an order from Motorola.
A fully automated Japanese line cost roughly $5M at the time. BYD used a half-manual, half-mechanical approach to reduce the cost to about $140,000; workers earned roughly $60 a month, while 何倩然 notes that $60 could be the current hourly wage of a US engineer.
More important than the low cost was the change in how the company learned. Once the line was broken down into individual steps, engineers had to understand firsthand how materials, equipment and processes fit together. 何倩然 says all know-how comes from “real hands-on, step-by-step, extremely detailed understanding and research.”
ATL spent about $1M to license lithium-ion battery technology from Bell Labs, only to discover a gas-generation problem. Because “a million means a lot,” the team could not walk away. 何倩然 says ATL was one of the licensees among more than 20 that persisted in solving the core problem and commercializing the technology, eventually winning consumer-electronics customers including Apple.
4. Power batteries look like chemistry; the mass-production moat is consistency
杨璐 says the battery industry often describes its work as “micron-scale manufacturing.” Deviations in the positive or negative electrode, electrolyte, separator, manufacturing process or Pack compound with one another; a vehicle must also make thousands of cells operate together, so voltage, internal resistance and capacity have to be highly consistent.
The lowest-capacity cell determines the usable capacity of the entire battery pack, creating a “barrel effect.” Early vehicles whose range suddenly collapsed or whose charge display reached full unusually quickly were often suffering from cell inconsistency and inadequate manufacturing, not merely weaker nominal material performance.
杨璐 cites exploding Samsung phone batteries as an example, attributing the problem to a high concentration of magnetic foreign matter in the material. Such contaminants can trigger thermal runaway during use. Consumer electronics may tolerate defect rates measured in dozens per million; automotive batteries have to push that down to only a few per million.
Production is divided into three stages. The front end handles batching, slurry mixing, coating, drying, calendaring and slitting; the middle end handles winding or stacking, tab welding, casing and electrolyte injection; the back end handles formation, aging and grading. 刘一鸣 recalls that some early manufacturers would simply let batteries sit for a while: “If nothing happened, then they could go into the car.”
5. Front-end coating is the core process that equipment alone cannot reproduce
何倩然 considers front-end electrode manufacturing the hardest part, especially moisture, burrs, dust, slurry dispersion and thickness control during coating. US startups, including Tesla, import rolls of electrode sheets from China and perform winding and assembly domestically.
TDK’s acquisition of ATL brought with it precision-coating capabilities accumulated in the tape business. Magnetic tape and battery electrodes both require sub-micron control of thickness, local defects and particle agglomeration across an extremely long flexible substrate; the engineering logic is “almost the same problem.”
This coating uniformity and process control pushed ATL’s yield, reliability and replicability to the limit, and later provided a foundation for CATL’s manufacturing system. The difference between Tier One players is not whether they can make a cell, but whether they stop at PPM or move toward PPB.
何倩然 uses this to explain Northvolt’s predicament: even with all equipment imported, without fundamental understanding, optimization capability and decades of accumulated experience, a company may be “losing money on every meter of material it coats.” Equipment is a purchasable asset; stable yield is not.
6. The moat comes from collective engineering optimization across hundreds of factories
Borrowing the phrase from Breakneck, 何倩然 describes China as governed by engineers, while the US is more lawyer-led. The same generation of technology has been repeatedly refined by dozens or hundreds of factories from different angles, driving down costs, raising yields and iterating equipment. That is “collective engineering optimization.”
She contrasts Chinese engineers’ long stays at factories and continuous station coverage with the US model of leaving at 5 p.m., working less on weekends and changing jobs every two years in Silicon Valley. Her conclusion is that manufacturing know-how must accumulate across generations, while rapid employee turnover makes experience difficult to retain.
杨璐 uses BYD’s mask project to demonstrate capability transfer. Engineers watched and sketched equipment beside a mask factory that never stopped running, then built and improved the machinery themselves. In less than two months, daily output reached roughly 5 million masks, briefly making BYD the world’s largest mass producer.
Factories protect their newest processes, but knowledge remains densely distributed across the industry. A Tesla employee searching Douyin in China for calendaring and coating machines can find thousands of results; the same search on TikTok in the US returns nothing. 刘一鸣 jokes that if every component of a new product can be found on Taobao or Douyin, investors should instead watch out for “involution.”
7. The cell-format debate is about historical paths and application boundaries
The principle behind BYD’s Blade battery is not mysterious: continue using LFP, but change cell shape and arrangement within a constrained Pack volume to raise volumetric utilization. The real challenge was making an ultra-long cell; almost all existing equipment and processes had to be rebuilt. Project lead 宋华军 says the difficulty was “figuring out how to make the principle work after understanding it.”
Tesla had few options when it started and could only pack more than 7,000 Panasonic 18650 cylindrical cells into a car. It then followed the same path through 21700, 4680, tabless and structural batteries. Cylindrical winding is efficient and highly automated, while variation across small cells is easier to average out.
何倩然 believes prismatic and Blade cells offer higher volumetric utilization, customizable dimensions, and compatibility with battery swapping, recycling and grid storage. Tesla embeds large cylindrical cells into the vehicle floor and makes them part of the structural design, but once the entire chassis is welded shut, it leaves less room to “take the battery out.”
8. Changzhou turned policy into a supply chain through a “four-hour industrial circle”
Changzhou joined the trillion-yuan GDP city club in 2023, with new energy as its flagship industry. A “four-hour industrial circle” means that most components for an NEV can be assembled within four hours; BYD sits north of the city and Li Auto to the south, drawing in thousands of companies covering drivetrains, brakes, steering, electrical systems, instruments, lights and automotive accessories.
The city introduced three rates: the on-road EV rate, meaning the share of new-energy vehicles in the total vehicle fleet; the charging-pile construction rate; and the solar-installation rate. 杨璐 sees these metrics as evidence of the southern Jiangsu model of “big-government leadership,” where local officials’ understanding of an industry directly affects execution speed.
Changzhou backed Li Auto early, when the market still doubted whether a new automaker could build cars. Li Auto initially planned an A0-class low-speed vehicle, but pivoted quickly after policy tightened, eventually finding success with the Li ONE. The local government also acknowledges that before the product was validated, it “wasn’t confident either.”
An earlier failed Boston Battery project nevertheless left behind “seeds”: a group of officials began studying the battery industry. The government also appears to have invested several hundred billion yuan in local industrial-investment funds. For companies including SVOLT Energy and CALB, Changzhou offered capital, headquarters relocation and existing supply-chain advantages to bring them in.
9. Yibin shows how an anchor company, talent and green power can build a cluster from zero
Yibin was previously known for Wuliangye and black tea, but lacked a manufacturing base, talent, young people and an industrial chain. 杨璐 believes its unusual advantage was the large share of state-owned enterprises in the city economy, allowing the government to mobilize resources for a transition and demonstrating the efficiency of “concentrating strength to accomplish major tasks.”
Yibin first built a university town to attract young people, then invited 欧阳明高 to establish an academician workstation, bringing research resources, student entrepreneurship and the World Power Battery Conference to the city. It did not wait for a complete industry before recruiting companies; it built talent, research capacity and urban appeal in parallel.
The true anchor company was CATL, with suppliers following the main plant. Yibin used hydropower, green electricity, low power prices and efficient government service to win the project. 刘一鸣 describes the capabilities required for local investment promotion as “a copper head, an iron mouth, keen ears, a rubber waist, a teapot belly and rabbit legs.”
10. Heavy-asset industries need concentrated resources; original exploration cannot have the answer predetermined
At a Stanford conference, 何倩然 asked why the US could not have the DOE and venture capital jointly select a route such as sodium-ion batteries and “concentrate strength to accomplish major tasks.” The response was: “You can’t predict the future.” The DOE should give each technology a fair chance to compete rather than designate a winner from the top down.
She accepts that the US model is better at incubating breakthrough technologies, but emphasizes that batteries are extremely capital-intensive, with long cycles from laboratory to pilot to commercial production. If funding is spread across 10-20 routes with only a small amount going to each, “not a single technology may emerge” after 10 or 20 years. Many VCs also lack the patience to keep funding losses.
11. China’s “planning” happens at scale-up; early innovation still relies on a race
杨璐 rejects reducing China’s model to a top-down process of selecting a single technology. The early objectives were to reduce oil imports and strengthen energy security, so multiple routes—including coal-to-oil and batteries—were researched simultaneously. Different experts led different directions, the results were consolidated, and technology performance and the market determined what stayed.
陈立泉 once asked 万钢 to include lithium batteries in a major automotive program and was told, “Just go ahead and work on lithium batteries.” There were multiple routes within batteries at the time; lithium-ion was not the predetermined winner. It emerged from the race after proving technically viable and gaining market validation.
杨璐 compares the Power Battery Conference to a parent-teacher meeting: officials from the Ministry of Industry and Information Technology act like a homeroom teacher, reading out market shares, company rankings and international gaps one by one, while each speaker reports on progress toward self-reliance and controllability. Once the route becomes clearer, regulators set targets, push to close gaps and drive scale-up.
12. The solid-state battery timeline gets rewritten every three years
何倩然’s judgment is direct: solid-state batteries did not hit a bottleneck only in the past year or two; they have been “in a bottleneck for the past 50 years.” Many major manufacturers once said mass production would arrive in 2019 or 2021, but claims of that kind are “renewed every three years.”
Even the definition is inconsistent. Companies use different thresholds for all-solid-state, semi-solid-state and whether liquid electrolyte accounts for 10% or 5%. The guests believe meaningful industrial progress may not arrive until around 2030, and even if scaled rollout begins then, solid-state batteries will coexist with mature liquid lithium-ion technology for a long time.
何倩然 uses QuantumScape, a public company, to test the marketing claims. Existing lithium-ion cells with silicon anodes can already exceed 400, 450 or even 500Wh/kg, so solid-state is not automatically denser. Its cells may still require roughly 3 atmospheres of external pressure—equivalent to “several people standing on that tiny piece of battery.”
Beyond lithium dendrites, the lithium-metal interface may separate after several dozen cycles, followed by “cliff-like decay.” The supply chain for new materials overlaps little with today’s liquid systems, while sulfides may release toxic gases after impact. CATL and BYD continue to invest in R&D, but neither has committed to a specific mass-production year.
13. Sodium-ion batteries are closer than solid-state to a supply-chain inflection point
何倩然 personally favors sodium-ion batteries and says both CATL and BYD have strategic plans. Her view is that over the next 5-10 years, sodium-ion batteries could displace LFP, which currently holds more than 30% and possibly half of the market—but the forecast is explicitly conditional on both timing and scale.
Sodium-ion technologies include layered oxides, Prussian blue analogues and polyanion systems. Some chemistries contain no lithium, cobalt or nickel, and do not even require copper, allowing them to bypass critical-metal constraints. That matters especially for China, which still relies on South America and Australia for lithium resources.
The supply chain has matured faster than she initially expected. CATL is involved in hard-carbon development linked to Tianmu Lake, with hard-carbon investment reaching roughly RMB9B; leading cathode-material companies including Ronbay have also begun developing sodium-ion cathodes.
NFPP in the polyanion family is described as having an olivine structure similar to LFP, with almost no degradation during charge and discharge. Existing LFP can reach roughly 15,000 cycles, while some polyanion sodium-ion systems are said to reach about 30,000. Their energy density is lower, but they are well suited to stationary, grid-scale storage.
14. AI data centers are turning the US grid into a real use case for sodium-ion batteries
何倩然 believes the US has weaker momentum behind vehicle electrification because oil and gas are cheap; the more urgent problem is the grid. Much of the infrastructure was built in the 1960s and is nearing end of life, while new transmission projects are slowed by “not in my backyard” opposition, Native American reservations and ecological-protection zones.
Even in California, Los Angeles may experience dozens of outages in a year. At the same time, Nvidia, OpenAI and Oracle are driving the construction of hyperscale AI factories, where GPU clusters create power spikes and load swings in the hundreds of megawatts, further stressing the grid.
She believes the fastest and least expensive solution is to combine cheap solar and batteries into a “plug-and-play” system rather than wait for nuclear power, geothermal projects or major transmission lines. Stationary storage does not need automotive-grade energy density; cost, cycle life and safety matter more.
Sodium-ion batteries may also be better suited to the frequent, shallow charge-discharge cycles of data centers, buffering load fluctuations. This is the core application logic behind TDK Ventures’ interest in sodium-ion storage paired with large AI data centers.
15. Europe and the US should seek shortcuts in rebuilding the chain; Northvolt marked the boundary of failure
Beyond sodium-ion, 何倩然 is betting on dry electrodes. If wet-process coating know-how and IP have long been concentrated in China, the US can use a totally different platform to bypass the legacy process. Tesla highlighted dry electrodes at Battery Day, while AM Batteries is also pursuing the technology to reduce energy use and manufacturing costs.
Another upstream shortcut is direct lithium extraction and recycling. One Lithium Americas project spent 7-8 years advancing without producing output; Ascend Elements has already extracted lithium carbonate from recycled black mass in the US, creating the possibility of repeatedly putting materials from a Tesla into future vehicles instead of restarting the entire exploration, mining and refining chain.
Northvolt had a team led by former Tesla VP Peter Carlsson, raised roughly $20B and secured about $50B in orders, yet still went bankrupt. 何倩然 believes this dealt a major blow to private-sector investment in Europe and the US because it showed that the combination of “too big to fail” funding, talent and orders still does not equal manufacturing capability.
曾倩群’s summary of this class of failure is: “You use the wrong design, use the wrong process, and then you use the wrong equipment.” 何倩然 concludes that Europe and the US must do more than buy Chinese equipment; they also need to attend venues such as CIBF to learn materials, processes and engineering know-how.