Global AI Price War Intensifies as Chinese Models Gain Ground

The DeepSeek logo. Growing adoption of low-cost Chinese AI models is intensifying global price competition, while concerns over service reliability persist. The global race for dominance in the artificial intelligence market has moved beyond performance enhancement and entered a phase of fierce price competition. As Chinese companies lure developers with usage fees as low as one-tenth of those charged by leading U.S. models, American tech giants including OpenAI and Anthropic have begun mounting their defense by rapidly rolling out low-cost models of their own. According to OpenRouter, a global AI model aggregation platform, Chinese AI models accounted for approximately 48% of all AI task requests recorded from Oct. 5 to 11, surpassing the combined 40.3% share held by OpenAI, Google and Anthropic. By company, China's DeepSeek ranked first at 24.5%, followed by Zhipu AI at 13.4% and Alibaba's Qwen at 5.4%. The primary reason developers are gravitating toward Chinese AI is its favorable cost-to-performance ratio. The new model "V4.1 Flash-1017," unveiled by Chinese startup DeepSeek on Oct. 17, outperformed OpenAI's "GPT-5.6 Sol" and Anthropic's "Claude Opus 5" in coding, reasoning and AI agent tasks, according to the company's own evaluations. Notably, it scored 77.0 on DeepSWE v1.1, a software development benchmark, placing it ahead of competing U.S. models. The cost gap is even more striking. DeepSeek's enterprise usage fee starts at as little as $0.15 per million input tokens — more than 20 times cheaper at the widest price gap than high-performance U.S. models, which are priced between $2 and $4. Even as it releases new models, DeepSeek has kept its pricing unchanged, maintaining its cost competitiveness. As competition from China intensifies, U.S. companies have responded with price cuts of their own. In September, OpenAI launched "GPT-6 Luna," a low-cost model priced at around $0.10 per million tokens, while Anthropic followed with the release of "Claude Haiku 5.5," a smaller model in a similar price range. Google has also joined the price war, temporarily offering its top-tier model at less than half its regular price as competition over profitability intensifies. This phenomenon is closely tied to the growth of the AI agent market, in which AI systems perform tasks autonomously on behalf of users. Because AI agents call on models multiple times to handle a single task, even small price differences can have an enormous impact on a company's operating costs as usage scales up. As a result, companies have a strong incentive to choose lower-cost models once performance meets their requirements. Despite rapid growth, service reliability remains a pressing challenge. DeepSeek has experienced five outages in October alone, and on Oct. 15, both its chatbot and API services were disrupted for approximately five hours. The cumulative number of outages this year has reached 44. Industry observers have pointed to U.S. export controls on AI chips as a possible reason why China's computing infrastructure is struggling to keep pace with surging demand. DeepSeek is betting heavily on infrastructure expansion, partnering with Huawei to reduce its dependence on Nvidia and constructing a 1-gigawatt data center in Inner Mongolia. While the company is pursuing a breakthrough through large-scale fundraising, concerns have been raised that recurring service disruptions could become a significant obstacle to future valuations or a potential initial public offering (IPO).