AI-Wiki

How China is viewing the AI race | The Dip Podcast

8 min read

Chu & Liu — How China is viewing the AI race (DW The Dip, Oct 2026)

▶ Watch on YouTube · DW News · 28:10

China’s AI ambitions are no longer just about catching up to Silicon Valley. From chips and data centers to energy, talent and global markets, Beijing is building an entire ecosystem designed to challenge US technological dominance. As Chinese companies roll out increasingly powerful AI models, the question is no longer whether China can compete, but who will shape the AI rules of the future. The US? China? Some combination of the two? The Dip’s Kassandra Sundt is joined by Claire Chu from Janes and the Council on Foreign Relations’ Zongyuan Zoe Liu.

— channel description, DW News (chapter list and links omitted)

TL;DR

A ~28-minute episode of The Dip, a business podcast on the DW News channel, published 4 October 2026. The host is Kassandra Sundt. The guests are Claire Chu, senior China analyst at Janes (a defence and security intelligence company) and a member of the Atlantic Council’s Global China Hub, and Zongyuan Zoe Liu, senior fellow for China studies at the Council on Foreign Relations, who leaves at 14:21. The captions are auto-generated, with proper nouns corrected at acquire time. There are no figures beyond one investment estimate.

1. Which models they use (1:30–2:53). Liu uses ChatGPT or Claude in the US and DeepSeek, Kimi and others in China, where she also asks e-commerce platforms’ assistants what to buy. Chu calls herself “a bit of an AI skeptic in my personal life.”

2. China’s stacked advantages (2:53–4:08). Chu: China stacks several competences to scale AI. It powers data centres with renewable energy, builds generation next to them, and routes training workloads to Inner Mongolia, Xinjiang and Gansu, where land is cheap, the climate cool, and wind and solar in surplus.

3. Finance is the constraint (4:08–5:26). Liu has moved from asking which country has the better model to asking who can pay: “a lot of this comes down to money… can you make the investment sustainable or not.” China’s system is “bank dominated… which is by default not very good at financing technology and innovation,” and Beijing is “rebuilding a mission-driven financial system to fund this tech competition.”

4. How Chinese people see AI (5:26–7:10). Chu: broadly more acceptance of new technology, partly because of the government’s emphasis on control and partly because of an official push to integrate AI into all sectors; AI-generated mini-dramas are a visible example. The government reads US guardrail talk as close to a security move, with sanctions used to suppress Chinese AI firms.

5. Building on Chinese models abroad (7:10–9:27). As Beijing moves away from the large infrastructure projects of the early Belt and Road years, companies pursue “more commercial opportunities, less sensitive opportunities” that fit countries’ own development priorities. “A lot of countries using Chinese models… building on Chinese models… to reduce the cost and… to close a technical gap.” Her example: in Kenya, “a lot of projects are being built on the backs of Chinese models,” with Chinese partners working with local universities and operators.

6. Productivity, not an arms race (9:27–12:45). Liu on her recent trip: managers in Chinese firms are embracing AI efficiency gains and tasking employees more, while employees have not yet learned to use it well, so AI is “accelerating that kind of anxiety” about involution (neijuan, intensifying competition without gain). Ordinary Chinese consumers and companies “do not view US and China are engaging in this fierce AI competition,” especially not as arms control. Technocrats see AI as total factor productivity and as an industrial-policy lever to “innovate out of China’s structural problem.” On overcapacity: by 2030 the US may invest “over 10 trillion” across AI-related industry, China far less, because “the technology is changing so fast they do not want to overinvest in the early stage” and risk stranded assets.

7. Chips (12:50–14:21). Liu: “chips are… fundamentally goods and the goods are fungible,” so export controls leak through shell companies in Malaysia or Hong Kong. China attacks the problem from two sides: buy chips “as long and as fast as they can,” and redesign chips and the industrial ecosystem around mature nodes. People she spoke to were confident of their own lithography machines and of “breakthroughs at 7 nanometers chips” soon.

8. Open weights as the distribution strategy (15:07–18:42). Chu: “rather than competing on high-end superiority, Chinese providers are marketing affordability, adaptability of their open-source and open-weight AI models.” Countries that make AI a national priority invite Chinese firms to support upskilling and localise AI across industry and government. Examples:

  • factories in Vietnam and Indonesia using Chinese models on the floor (anecdotal);
  • a new global development centre in Kuala Lumpur for AI payments and fintech;
  • Huawei working with Nigerian agencies on agentic-AI cloud infrastructure, and with Nigeria’s national commission for colleges of education to upskill teachers;
  • similar upskilling in Egypt and Kenya.

Asked what is in it for Chinese firms, Chu says China “missed the boat” on earlier generations such as software, and Beijing wants to “leapfrog” in emerging technology. She cites the World AI Cooperation Organization founded in Shanghai this summer with 29 members.

9. Kazakhstan (19:09–22:35). A joint venture of the Kazakh government and 01.AI (backed by Alibaba and Xiaomi) runs Kazakhstan’s national AI platform. The 01.AI CEO sits on the president’s AI council and helped build an AI board director for the sovereign wealth fund on a Kazakh-language model, able to vote. A month earlier Kazakhstan had joined a US-led initiative (“Pax Silica”, inferred from the ASR), and a US-backed hyperscale data centre is planned there with Nvidia. Chu: “I don’t know if there’s a predetermined end to this race.” She sees sub-Saharan and North Africa as “largely overlooked by the US.”

10. Data centres without pushback (22:35–24:08). Asked whether rural Chinese resist data centres as Americans do, Chu says China has equipped local governments to address concerns “before they can create real obstacles.”

11. Who sets the rules (24:26–27:04). Chu expects an ongoing struggle, with the two economies intertwined (Visa and Mastercard work with Chinese fintech firms on auditing standards). In Europe there is “emerging distrust of Chinese hardware, of US software,” and strategic autonomy has been “the phrase of the last year or two” for middle powers. Meanwhile China’s soft initiatives embed its standards without formal alliances: training government employees, university partnerships, and vocational training on Chinese systems.

12. In five years (27:04–27:35). Chu: for herself, US AI; “but for a lot of the world, likely not,” and payment systems in particular “are very likely to be backed by Chinese AI… and Chinese hardware.”

Dynamic-capabilities reading

  • digital-transforming/navigating-innovation-ecosystems. The episode describes China building an ecosystem rather than selling a model: open weights, partnerships with local universities and operators, joint ventures with governments, teacher upskilling, and vocational training on Chinese systems. For a buyer outside the two powers, the choice of model is a choice of ecosystem.
  • contextual/external-triggers. Export controls, chip smuggling, US-China summits, and the spread of strategic-autonomy thinking among middle powers are external conditions on any firm’s or government’s AI choices.

Linked entities and concepts

  • Entities: DW News (channel; entity page created with this ingest, its second source as author).
  • Concepts: open-source-ai (open weights as a geopolitical distribution strategy), ai-sovereignty (strategic autonomy for middle powers; national AI platforms built with a foreign partner).
  • Dangling (single-source mention, deferred): Kassandra Sundt, Claire Chu, Zongyuan Zoe Liu, Janes, Council on Foreign Relations, 01.AI, World AI Cooperation Organization.

Debates and supersession

  • Analysts’ accounts, few sources. The episode is two analysts’ views, with anecdotes from a trip and from reporting. The one figure (US investment over $10 trillion by 2030) is given without a source, and Liu corrects her own year mid-sentence.
  • Pushback in China. Chu’s claim that local governments head off resistance to data centres is an assessment, not evidence that there is none.
  • Names inferred from the ASR. “Pax Silica” is inferred from “PACILA”; the Kazakh model, the US data-centre backer, and the Kuala Lumpur centre’s owner are left as captioned in the raw file. None of these details is used beyond this page.
  • No supersession.

What was actually ingested

The full auto-generated English track: 678 segments, 0:00–28:05 of 28:10, thirteen creator chapters. Three voices; attributions are clear from direct address. No stills: the video is a three-way video call with a studio host.

Relationships

Supports

  • How China Plans to Win the Global AI Race — Shared topic: Chinese open-weight models as a strategy for reaching markets outside China. Bloomberg gives usage and cost figures and a switching case; Chu describes the same strategy country by country in the Global South and Central Asia.
  • Inside China's AI race | On Assignment — Shared topic: how China approaches AI, told by people who report on it. Reuters' correspondents describe pre-release testing and the domestic market; Liu describes Chinese developers' view of AI as productivity policy rather than an arms race.
  • China, Open Source & AI Competitiveness I Andrew Ng — Shared topic: open models and US-China competition. Ng argues open models are not dangerous and that US restrictions slow the US; Chu describes China using open-weight models to win adoption abroad.

Graph View

Backlinks