# Kinling Lo and Tiffany Ap report in Rest of World: DeepSeek shifts China’s AI race toward applications [Canonical reading](https://escapemeridian.com/readings/kinling-lo-and-tiffany-ap-report-in-rest-of-world-deepseek-689183bf) Rest of World published Kinling Lo and Tiffany Ap’s report on April 14, 2025. The authors argue that DeepSeek’s success, alongside widening U.S. chip bans, has accelerated change in China’s AI industry: startups are being pushed toward applications and services, while major technology companies continue investing in AI models. They describe a tension between building costly models and finding practical ways to earn revenue from AI. Investors and analysts quoted in the article say DeepSeek’s efficiency claims challenge expectations for model-training budgets, and that smaller firms may struggle to compete in foundation models. The report also describes startups shifting toward areas such as medical AI and finance, while some continue developing models or AI agents. Its closing argument, voiced by venture capitalist Celia Chen, is that companies combining AI with domain expertise may find opportunities in solutions that big tech cannot easily replicate, including medical diagnostics and business workflows. The report notes that Chinese AI venture funding remained muted despite greater state backing. It does not establish whether application-focused startups will succeed, or whether their products will deliver the cited benefits. Headline and summary by Escape Meridian, prepared by Luna from the attributed source. A source check is not independent confirmation of the source’s claims. ## Original source Rest of World · Article Original source date: Apr 14, 2025 (2025-04-14T10:00:00+00:00). Date basis: Date stated on the publisher’s page. Added to Escape Meridian: 2026-11-14T15:03:23.853Z. First observed: 2026-11-04T05:55:22.836Z. Retrieved: 2026-11-04T05:57:20.190Z. - [DeepSeek and chip bans have supercharged AI innovation in China](https://restofworld.org/2025/china-ai-boom-chip-ban-deepseek/) ## Limits Text as retrieved on November 4, 2026; originally published April 14, 2025. The report attributes the claim that DeepSeek’s V3 training cost about $6 million to DeepSeek; it does not provide independent verification in the retrieved text. The report gives no measurement basis for Meituan’s claim that LongCat is improving operational efficiency. The report does not establish whether startups shifting toward applications will succeed or whether their products will deliver the outcomes described. ## Open question What evidence will show whether startups that shift from developing models to building AI applications can sustain businesses and deliver the outcomes the report describes? ## Source-stated Kinling Lo and Tiffany Ap report that DeepSeek’s success and widening U.S. chip bans have pushed China’s AI ecosystem to innovate more quickly and startups toward practical applications. > Widening U.S. chip bans, combined with DeepSeek’s success, have pushed the Chinese AI ecosystem to innovate more quickly - [Supporting source](https://restofworld.org/2025/china-ai-boom-chip-ban-deepseek/) ## Source-stated The report says Chinese technology companies are releasing competing AI models and investing in research, while startups face pressure to focus on applications and services. > Chinese tech giants, meanwhile, are launching new AI models and pouring billions into research, signalling a future where only major players will compete on AI model development - [Supporting source](https://restofworld.org/2025/china-ai-boom-chip-ban-deepseek/) ## Source-stated Kevin Xu argues that DeepSeek’s efficiency claims make it harder for smaller AI firms focused on foundation models to compete for investment. > Investors in China would now be wary of backing smaller AI firms still focused on foundation models, since DeepSeek’s technology makes it harder for them to compete, said Xu. - [Supporting source](https://restofworld.org/2025/china-ai-boom-chip-ban-deepseek/) ## Source-stated The report describes Baichuan halting model pretraining in mid-2024 to focus on medical AI services, according to Chinese tech media outlet 36Kr. > Baichuan halted its model pretraining in mid-2024 to focus on medical AI services, according to Chinese tech media outlet 36Kr. - [Supporting source](https://restofworld.org/2025/china-ai-boom-chip-ban-deepseek/) ## Source-stated Celia Chen argues that opportunities may lie in combining AI with domain expertise to provide solutions that big tech cannot easily replicate. > The real winners will be those who combine AI with domain expertise to deliver solutions that big tech can’t easily replicate - [Supporting source](https://restofworld.org/2025/china-ai-boom-chip-ban-deepseek/) ## Source-stated The report says venture capital investment in Chinese AI and machine learning remained muted, with deal value down compared with the same period the prior year. > The total deal value fell 30% compared with the same period last year. - [Supporting source](https://restofworld.org/2025/china-ai-boom-chip-ban-deepseek/) ## Source and contributor context Rest of World · Publisher The publisher is identified here. No individual author has been inferred. ## Source media Media retrieval status: unchecked. ## Terms and references Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement. ## Foundation model A large model trained on broad data that can be adapted to many different tasks. - [Definition reference](https://developers.google.com/machine-learning/glossary#foundation-model) ## AI agent An AI system that chooses and carries out actions toward a goal, often using software tools. - [Definition reference](https://www.ibm.com/think/topics/ai-agents) ## Rest of World Rest of World reports technology and its social effects outside a US-only frame. - [Explore this context](/entities/rest-of-world) - [Reference](https://restofworld.org/) - [Reference](https://restofworld.org/feed/latest/rss.xml) ## Editorial provenance Version 1. Publication ID: 103eb875-5c98-a511-b1ff-6f8bee6662c8. Prepared by openai/gpt-6-luna; grounding review by anthropic/claude-opus-5.5. Source check: uncertain by anthropic/claude-opus-5.5 on 2026-11-14T14:58:47.786Z. This checks fidelity to the source, not whether its claims are independently true. Editorial selection: accept; reasons: bridge\_admitted\_by\_model. Observation: aa3eeffe392f93ae68436be08299318b1f28cee24ecae31e3d66ba4eadb9828b. Source text hash: 73f1c3d07f490187963592d78ddcfa8b496ff7e7ccd99f46f4a75880b7297eb5. ## Revision and source-check history ### Version 1 — 2026-11-14T15:03:23.853Z Rest of World published Kinling Lo and Tiffany Ap’s report on April 14, 2025. The authors argue that DeepSeek’s success, alongside widening U.S. chip bans, has accelerated change in China’s AI industry: startups are being pushed toward applications and services, while major technology companies continue investing in AI models. They describe a tension between building costly models and finding practical ways to earn revenue from AI. Investors and analysts quoted in the article say DeepSeek’s efficiency claims challenge expectations for model-training budgets, and that smaller firms may struggle to compete in foundation models. The report also describes startups shifting toward areas such as medical AI and finance, while some continue developing models or AI agents. Its closing argument, voiced by venture capitalist Celia Chen, is that companies combining AI with domain expertise may find opportunities in solutions that big tech cannot easily replicate, including medical diagnostics and business workflows. The report notes that Chinese AI venture funding remained muted despite greater state backing. It does not establish whether application-focused startups will succeed, or whether their products will deliver the cited benefits. ## Original source Rest of World · Article Original source date: Apr 14, 2025 (2025-04-14T10:00:00+00:00). Date basis: Date stated on the publisher’s page. Added to Escape Meridian: 2026-11-14T15:03:23.853Z. First observed: 2026-11-04T05:55:22.836Z. Retrieved: 2026-11-04T05:57:20.190Z. - [DeepSeek and chip bans have supercharged AI innovation in China](https://restofworld.org/2025/china-ai-boom-chip-ban-deepseek/) ## Limits Text as retrieved on November 4, 2026; originally published April 14, 2025. The report attributes the claim that DeepSeek’s V3 training cost about $6 million to DeepSeek; it does not provide independent verification in the retrieved text. The report gives no measurement basis for Meituan’s claim that LongCat is improving operational efficiency. The report does not establish whether startups shifting toward applications will succeed or whether their products will deliver the outcomes described. ## Open question What evidence will show whether startups that shift from developing models to building AI applications can sustain businesses and deliver the outcomes the report describes? ## Source-stated Kinling Lo and Tiffany Ap report that DeepSeek’s success and widening U.S. chip bans have pushed China’s AI ecosystem to innovate more quickly and startups toward practical applications. > Widening U.S. chip bans, combined with DeepSeek’s success, have pushed the Chinese AI ecosystem to innovate more quickly - [Supporting source](https://restofworld.org/2025/china-ai-boom-chip-ban-deepseek/) ## Source-stated The report says Chinese technology companies are releasing competing AI models and investing in research, while startups face pressure to focus on applications and services. > Chinese tech giants, meanwhile, are launching new AI models and pouring billions into research, signalling a future where only major players will compete on AI model development - [Supporting source](https://restofworld.org/2025/china-ai-boom-chip-ban-deepseek/) ## Source-stated Kevin Xu argues that DeepSeek’s efficiency claims make it harder for smaller AI firms focused on foundation models to compete for investment. > Investors in China would now be wary of backing smaller AI firms still focused on foundation models, since DeepSeek’s technology makes it harder for them to compete, said Xu. - [Supporting source](https://restofworld.org/2025/china-ai-boom-chip-ban-deepseek/) ## Source-stated The report describes Baichuan halting model pretraining in mid-2024 to focus on medical AI services, according to Chinese tech media outlet 36Kr. > Baichuan halted its model pretraining in mid-2024 to focus on medical AI services, according to Chinese tech media outlet 36Kr. - [Supporting source](https://restofworld.org/2025/china-ai-boom-chip-ban-deepseek/) ## Source-stated Celia Chen argues that opportunities may lie in combining AI with domain expertise to provide solutions that big tech cannot easily replicate. > The real winners will be those who combine AI with domain expertise to deliver solutions that big tech can’t easily replicate - [Supporting source](https://restofworld.org/2025/china-ai-boom-chip-ban-deepseek/) ## Source-stated The report says venture capital investment in Chinese AI and machine learning remained muted, with deal value down compared with the same period the prior year. > The total deal value fell 30% compared with the same period last year. - [Supporting source](https://restofworld.org/2025/china-ai-boom-chip-ban-deepseek/) ## Source and contributor context Rest of World · Publisher The publisher is identified here. No individual author has been inferred. ## Source media Media retrieval status: unchecked. ## Terms and references Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement. ## Foundation model A large model trained on broad data that can be adapted to many different tasks. - [Definition reference](https://developers.google.com/machine-learning/glossary#foundation-model) ## AI agent An AI system that chooses and carries out actions toward a goal, often using software tools. - [Definition reference](https://www.ibm.com/think/topics/ai-agents) ## Rest of World Rest of World reports technology and its social effects outside a US-only frame. - [Explore this context](/entities/rest-of-world) - [Reference](https://restofworld.org/) - [Reference](https://restofworld.org/feed/latest/rss.xml) ## Editorial provenance Version 1. Publication ID: 103eb875-5c98-a511-b1ff-6f8bee6662c8. Prepared by openai/gpt-6-luna; grounding review by anthropic/claude-opus-5.5. Source check: uncertain by anthropic/claude-opus-5.5 on 2026-11-14T14:58:47.786Z. This checks fidelity to the source, not whether its claims are independently true. Editorial selection: accept; reasons: bridge\_admitted\_by\_model. Observation: aa3eeffe392f93ae68436be08299318b1f28cee24ecae31e3d66ba4eadb9828b. Source text hash: 73f1c3d07f490187963592d78ddcfa8b496ff7e7ccd99f46f4a75880b7297eb5. ### uncertain — 2026-11-14T14:58:47.786Z — anthropic/claude-opus-5.5 Publication: 103eb875-5c98-a511-b1ff-6f8bee6662c8 attribution — uncertain: Authors and publisher are correct. However, the summary says 'The authors argue' and claim 1 says the authors 'report' that DeepSeek and chip bans pushed faster innovation. The source attributes that assessment to 'experts and investors told Rest of World', and claim 1's excerpt cuts off before that attribution. Smallest repair: 'the report, citing experts and investors, says...'. > pushed the Chinese AI ecosystem to innovate more quickly, strengthening its position in the global tech landscape, experts and investors told Rest of World summary — uncertain: The reading is otherwise faithful and balanced, but it never says what DeepSeek is, though it is the central actor. A Chinese AI company whose high-performing open-source models were reportedly built cheaply is essential context for a non-specialist. 'Pretraining' (claim 4) is also undefined and missing from definitions. Repair: add one clause identifying DeepSeek, and a term-library entry for pretraining. > build products and services on top of its high-performing open-source technology date — pass: April 14, 2025 matches originalPublished and the article dateline at day precision. Forecasts, such as where investment will flow, are kept as attributed views. No hindsight framing. The limitation discloses the later retrieval date, and kind is publisher-page, so no archive disclosure is needed. > 14 April 2025 support — uncertain: Claim 3 says Xu argues that DeepSeek's 'efficiency claims' make it harder for smaller firms to 'compete for investment'. Xu actually says DeepSeek's technology makes it harder for them to compete, and that investors will be wary. The efficiency-claims framing belongs to PitchBook's Melanie Tng. Repair: 'Xu says investors will be wary of smaller foundation-model firms because DeepSeek's technology makes it harder for them to compete.' The page was edited after publication, but the cited passages appear to be original article text. The other claims are grounded. > since DeepSeek’s technology makes it harder for them to compete, said Xu. media — pass: No media was selected, and no verified candidates were supplied, so nothing is misidentified. ## Recorded references Explicit citations in public readings and public file attachments to this reading. Shared topics are not citations. Other artifact types need an explicit reference adapter. Some references could not be checked. Reload to try again. [Reference data (JSON)](/readings/689183bf6387a125fa1ea854ef350b369e20570e6f17b9be2f7231f784341489/references.json) ## Corrections and challenges [Send a correction or an opposing source](https://escapemeridian.com/corrections/) [JSON record](https://escapemeridian.com/readings/kinling-lo-and-tiffany-ap-report-in-rest-of-world-deepseek-689183bf.json)