Chinese AI models up to 50 times cheaper as enterprises reassess OpenAI, Anthropic costs: JPMorgan

Enterprises are increasingly exploring lower-cost open artificial intelligence (AI) models as the price of using leading AI systems continues to rise. Some Chinese models now cost 10 to 50 times less per token than frontier models while delivering comparable performance on several benchmarks, according to a new JPMorgan analysis.

The report, titled "Semiquincententacles: The US Grip on Global Markets," stated that growing cost pressures are prompting businesses to reconsider their dependence on premium AI models from companies such as OpenAI and Anthropic.

The shift comes as major AI providers move towards usage-based pricing and raise charges to recover the massive infrastructure investments required to build and operate advanced models.


JPMorgan cited examples of rising AI costs across the industry, including OpenAI doubling token prices between GPT-5.4 and GPT-5.5, and Microsoft increasing Copilot pricing and introducing higher charges across several AI model families. Some enterprise users reported AI-related cost increases of up to 100-fold following pricing changes.

The report points to early signs of migration towards cheaper alternatives.

AI startup Lindy said it shifted its services from Anthropic's Claude models to China's DeepSeek, claiming millions of dollars in savings while improving performance.

Coinbase chief executive officer Brian Armstrong has also said he expects most AI workloads to move to significantly cheaper models over the next year.

According to JPMorgan, the cost-performance equation is increasingly favouring open-weight models.

One comparison found that Anthropic's Claude Opus 4.8 achieved a benchmark score of 56 at a cost of about $3,700, while DeepSeek V4 Pro scored 44 for roughly $186, making it nearly 20 times cheaper.

The report noted that Chinese models from companies including DeepSeek, Alibaba, Xiaomi, MiniMax and Kimi now dominate what it describes as the industry's "intelligence-per-dollar" frontier, where performance is balanced against operating costs.

JPMorgan cautioned that frontier models from OpenAI and Anthropic will remain essential for advanced scientific research, cybersecurity and high-end AI agents.

However, it argued that many enterprise workloads may not require the most powerful models, creating a significant opportunity for cheaper open alternatives as businesses seek to control AI spending.