Serenity: Many US Companies Turning to DeepSeek to Cut Costs, Leading to Revenue Strain on High-End Models

BlockBeats News, June 26th, Serenity tweeted about the phenomenon of U.S. companies extensively using Chinese AI models. Serenity cited a UBS report indicating that many enterprises have started routing simple tasks to cheaper Chinese open-source models, with some teams spending up to $35,000 in tokens per month and exceeding 200%. As a result, high-end model revenue growth is under pressure. Serenity confirmed the UBS report and attributed it to a capitalist dilemma—where the market naturally selects the cheapest option, and Chinese models like DeepSeek are indeed much cheaper than Gemini, OpenAI, Anthropic.

Serenity also stated that the Trump administration's previous suspension of Fable/Mythos access was the right move, as the repeated distillation of top models poses significant risks, and models approaching AGI levels should have increased access barriers. Serenity believes the U.S. needs to do two things: further develop models specializing in low-cost reasoning and impose bank-level identity verification for AI model access.