Chinese AI is pulling global competition into a different game, Reuters Breakingviews argued on September 17, noting that U.S. AI giants have raced forward on massive spending — over $380 billion in venture funding from 2023 to mid-2026, with U.S. tech majors expected to invest nearly $1 trillion in AI infrastructure by 2027, versus roughly $165 billion for their Chinese peers.
Yet China has not been left behind. Its strongest models trail OpenAI and Anthropic's top models by only about seven months on average, while DeepSeek's latest V4-Flash model runs at more than 100 times lower cost than Anthropic's Claude Fable 5.
Chinese companies are winning on a different axis — efficiency, low price and scale. Alibaba said its models were downloaded over 3 billion times in the past six months, surpassing Meta's and Google's free models. OpenRouter data showed that as of September 14, DeepSeek, Zhipu, Tencent and Alibaba together held about 45% of API market share.
Costs keep falling: Zhipu said single-inference cost dropped 80% this year and its API gross margin turned from negative to 25%. Breakingviews' verdict: this low-cost, high-efficiency competition is teaching OpenAI and Anthropic an important lesson.
