How AI Is Forcing China to Reshape Industrial Policy
How AI Is Forcing China to Reshape Industrial Policy
3 hours ago•Odd Lots•Bloomberg
Podcast45 min 40 sec
Listen to Episode
Note: AI-generated summary based on third-party content. Not financial advice. Read more.
Quick Insights
  • Treat ASML (ASML) as a strategic semiconductor-supply-chain investment, while weighing its China exposure and export-control risks.
  • Monitor TSMC (TSM) as the benchmark for advanced chip manufacturing; China’s domestic-chip efforts still face major equipment, capacity, and HBM constraints.
  • Track SK Hynix, Samsung Electronics, and Micron (MU) for HBM supply, a key bottleneck for expanding AI-compute capacity.
  • Be selective with Chinese technology and industrial stocks: policy support and production growth do not guarantee durable profits, so prioritize companies with strong margins and defensible advantages.
Detailed Analysis

NVIDIA (NVDA)

  • Chinese AI companies have wanted access to NVIDIA GPUs, and some were reportedly eager to order H-200 chips when exports were allowed.
  • NVIDIA’s CUDA software ecosystem is a major advantage: Chinese firms are accustomed to building on it, while domestic alternatives have had to develop their own software libraries and user base.
  • The discussion contrasted the appeal of NVIDIA’s established platform with Beijing’s push for AI companies to use domestic chips.

Takeaways

  • NVIDIA’s position in China may face pressure from export controls and China’s drive for self-reliance, even as Chinese AI firms continue to value its products.
  • Investors should distinguish demand for NVIDIA chips from the separate question of whether Chinese companies can legally obtain them.

Huawei and China’s Domestic AI-Chip Ecosystem

  • Huawei’s Ascend chips were described as increasingly usable for AI inference, with efforts underway to make them more capable for training large models.
  • Huawei is working with Chinese AI firms, including DeepSeek, to adapt models and chips together. Some Chinese companies are also preparing models to run on domestic chips from launch.
  • Huawei is trying to compensate for less powerful individual chips by connecting larger numbers of them into “super pods.”
  • The discussion identified major constraints: chip-production capacity, semiconductor equipment, lithography, and high-bandwidth memory (HBM). Huawei has reportedly told some customers they may need to source HBM themselves.
  • Chinese firms including CXMT are working to develop HBM, but the guest said they still lag the leading suppliers. The discussion described a two-to-three-year effort to reach “good enough” HBM.
  • U.S. export controls have accelerated China’s effort to build a more self-reliant semiconductor supply chain. The guest also emphasized that catching the global frontier is difficult because that frontier keeps advancing.

Takeaways

  • The domestic chip effort presents a long-term opportunity tied to government priorities, but execution depends on expanding production and addressing equipment and memory bottlenecks.
  • Watch for evidence that domestic chips can support both model training and inference at scale—not just announcements of partnerships or new designs.
  • The discussion points to a trade-off: using domestic chips may support self-reliance but could leave Chinese AI firms behind the frontier in the near term.

SMIC (0981.HK / 688981.SS)

  • SMIC was described as China’s leading chipmaker and a key manufacturing partner for Huawei.
  • Its capacity is constrained by available equipment and supply chains, and some of its capacity is allocated to Huawei.
  • The guest highlighted lithography equipment as a critical limitation for China’s semiconductor ambitions.

Takeaways

  • SMIC’s strategic importance does not automatically translate into unconstrained growth: capacity and access to advanced manufacturing equipment are central issues to monitor.
  • The transcript suggests evaluating chipmakers on production scale and supply-chain access, not only on the policy support they receive.

ASML (ASML)

  • ASML’s lithography machines, including DUV immersion equipment, were identified as important bottlenecks for China’s ability to expand domestic chip production.
  • Restrictions on access to advanced lithography equipment helped spur China’s efforts to develop its own semiconductor equipment and supply chain.

Takeaways

  • The discussion underscores ASML’s strategic role in the global semiconductor supply chain.
  • Export restrictions and China’s efforts to develop substitutes are important considerations when assessing the company’s long-term exposure to China.

TSMC (TSM)

  • TSMC was presented as a formidable competitor because it operates at the chip-manufacturing frontier and can use advanced equipment and suppliers from around the world.
  • The guest compared China’s effort to catch up with trying to catch a moving train: Chinese capabilities may improve while the global frontier advances at the same time.

Takeaways

  • The transcript supports a view of TSMC as a key benchmark for advanced manufacturing, while emphasizing that China’s catch-up effort is uncertain and ongoing.
  • Consider the difference between China producing “good enough” chips domestically and matching the global frontier in both technology and scale.

HBM Suppliers: SK Hynix, Samsung Electronics, and Micron (MU)

  • The discussion identified SK Hynix, Samsung, and Micron as the leading HBM suppliers.
  • HBM was described as a bottleneck for pairing memory with Huawei’s AI chips. Chinese supplier CXMT is trying to develop HBM but was said to remain behind the leaders.

Takeaways

  • HBM is a key part of the AI-chip supply chain, not just a secondary component.
  • Monitor whether domestic Chinese memory production can close the gap; until then, access to leading HBM suppliers remains relevant to AI-compute capacity.

BYD (1211.HK / BYDDY)

  • BYD was described as an EV company that began as a battery maker and also makes chips and is moving into autonomous driving and robotics.
  • The discussion used BYD as an example of how Chinese companies can build on capabilities from adjacent industries.
  • At the same time, the guest warned that Chinese industrial sectors can face overcapacity, low margins, and difficulty sustaining competitive advantages.

Takeaways

  • BYD illustrates the potential advantages of integration across batteries, vehicles, chips, and related technologies.
  • The discussion also cautions against equating industrial capability or market growth with durable shareholder returns; margins and competitive barriers matter.

Xiaomi (1810.HK)

  • Xiaomi was described as a smartphone and consumer-electronics company that expanded into electric vehicles, drawing on China’s existing supply-chain capabilities and technical expertise.
  • The episode framed Xiaomi as an example of a firm moving into an adjacent industry rather than making an isolated leap into a new sector.

Takeaways

  • Xiaomi’s expansion shows how established technology and supply-chain capabilities can support entry into new markets.
  • Investors should still assess whether expansion produces sustainable profits, given the transcript’s broader concerns about intense competition and low margins in China.

Baidu (BIDU / 9888.HK)

  • Baidu was cited as a full-stack technology company with its own AI model, Ernie, chip-design business, data-center business, search business, and self-driving-car efforts.
  • The hosts noted that Ernie receives less attention than some other AI models.

Takeaways

  • Baidu illustrates the breadth of China’s full-stack AI strategy, but the transcript did not provide evidence about Ernie’s commercial performance or Baidu’s investment outlook.
  • Consider adoption and monetization alongside the company’s range of AI capabilities.

Alibaba (BABA / 9988.HK)

  • Alibaba was compared with U.S. technology firms pursuing multiple layers of the technology stack, including cloud, chips, and AI.
  • The discussion said Chinese companies are often more involved in manufacturing hardware and components, rather than outsourcing as much of the physical production.

Takeaways

  • Alibaba is presented as part of a broader full-stack technology trend, not as a specific stock recommendation.
  • The transcript offers no company-specific valuation, earnings, or price-target analysis.

DeepSeek and Z.AI (Private / Company Name as Stated)

  • DeepSeek was described as working closely with Huawei to adapt AI models to domestic chips, including technical collaboration between the companies.
  • The guest said Huawei chips had been more suitable for inference than for training very large models, while efforts were underway to improve training capability.
  • Z.AI was cited as an example of a Chinese AI company preparing a model to run on domestic chips from day one.

Takeaways

  • These examples show how China’s AI-model developers and chipmakers may coordinate to build a domestic ecosystem.
  • The main issue raised is the trade-off between using domestic hardware for long-term independence and using established alternatives to pursue model improvements more quickly.

Kweichow Moutai (600519.SS), CATL (300750.SZ), and Chinese Equities

  • The hosts said China’s stock market had experienced a difficult year and described the long-term Shanghai Composite chart as broadly sideways, despite periodic surges.
  • They noted that Kweichow Moutai had again become the largest constituent of the CSI 300, with CATL among the next-largest companies.
  • The guest argued that it can be hard for Chinese companies to sustain competitive advantages because competitors can enter, recruit talent, and obtain local-government financing. This can lead to overcapacity, low margins, and weak stock returns.
  • The discussion emphasized that strong industrial progress does not necessarily translate into strong shareholder returns.

Takeaways

  • The transcript is cautious on broad Chinese-equity performance: industrial strength alone may not produce durable profits or rising valuations.
  • For company research, pay close attention to margins, overcapacity, and the durability of competitive advantages—not just policy support or production scale.
  • The episode did not offer price targets or a specific recommendation for Moutai, CATL, or the Chinese market.

DJI (Private)

  • DJI was cited as an unusual example of a Chinese company with a strong position in the drone market.
  • The guest said he did not know why the market had become so concentrated and noted that new competitors could emerge.

Takeaways

  • DJI was presented as an example of a potentially durable competitive position, but the discussion did not establish why that position exists or whether it will persist.
  • The possibility of new entrants is a risk to consider when evaluating concentrated markets.

AI, Robotics, EVs, Batteries, and Biotech

  • China’s industrial strategy focuses on connected sectors such as AI, semiconductors, EVs, batteries, robotics, and biotech. Firms may use capabilities from one industry to enter another.
  • Government plans and state-backed investment funds can direct resources toward designated priority sectors.
  • The guest said Beijing sees AI as a tool to support other industries—including manufacturing, drug discovery, healthcare, and robotics—rather than as the only technology that matters.
  • The discussion also noted that Chinese companies may move from lower-margin markets such as EVs toward faster-growing areas such as robotics.
  • A recurring risk is that many companies may enter the same subsidized or strategically favored sector, leading to excess capacity and weak returns.

Takeaways

  • The potential opportunity is in cross-sector spillovers: for example, whether AI, batteries, and manufacturing expertise help companies compete in robotics or other emerging industries.
  • Treat policy support as a source of demand and funding, not as proof that an industry will be profitable.
  • Track whether new sectors develop sustained customer demand and pricing power, rather than focusing only on production capacity or government priorities.

Virtual Assets, Tokenization, and Stablecoins

  • A brief sponsor segment mentioned growing adoption of virtual assets, tokenization, and stablecoins, and raised the possibility that regulation could help the sector develop.
  • The episode itself did not discuss specific cryptocurrencies, companies, investment products, or regulatory changes.

Takeaways

  • The transcript provides no asset-specific basis for an investment decision in crypto or tokenization.
  • Any assessment would require additional information about the relevant assets, regulation, and business models.
Ask about this postAnswers are grounded in this post's content.
Episode Description
Can Chinese tech companies help the country achieve technological self-reliance while keeping up in the AI arms race? On this episode, we speak with Kyle Chan, a fellow at the Brookings Institution’s John L. Thornton China Center and author of the High Capacity newsletter, about Beijing’s response to the global AI boom and the tensions between Chinese companies’ commercial ambitions and the government’s long-term plans. He explains how companies draw on shared expertise and supply chains to expand across industries including smartphones, robotics, EVs and semiconductors — and we also discuss Huawei’s efforts to compete with Nvidia and why technological success doesn’t necessarily translate into strong returns for stock investors in China. Read more: AI Threatens to Make China’s Jobs Slump Much Worse China AI Biotech Firm Raises $200 Million in Oman-Backed Round Only http://Bloomberg.com subscribers can get the Odd Lots newsletter in their inbox each week, plus unlimited access to the site and app. Subscribe at  bloomberg.com/subscriptions/oddlots Subscribe to the Odd Lots Newsletter Join the conversation: discord.gg/oddlots See omnystudio.com/listener for privacy information.
About Odd Lots
Odd Lots

Odd Lots

By Bloomberg

<p>Bloomberg's Joe Weisenthal and Tracy Alloway explore the most interesting topics in finance, markets and economics. Join the conversation every Monday and Thursday.</p>