The economic calculus of the AI industry has reached a breaking point. Startups are realizing that even top-tier model performance cannot justify an astronomical burn rate. Flo Crivello, CEO of Lindy—a company with 25 employees—confirmed that his project has completely ditched Anthropic's Claude in favor of DeepSeek. The reason is pragmatic: the bills from Anthropic exceeded the company’s total payroll. Crivello described this cost structure as "unsustainable," noting that switching to the Chinese model collapsed the expense curve to near zero, saving the organization millions of dollars. In an interview with CNBC, he summarized the move not as a matter of preference, but as a matter of business survival.

This shift exposes a critical vulnerability for Western giants like Anthropic and OpenAI. While Anthropic prepares for an IPO and investors expect growth, the company is facing a severe deficit in customer loyalty. The migration to DeepSeek was largely enabled by the model being hosted on U.S. soil by American providers, which simultaneously removed regulatory hurdles and eased the concerns of compliance officers.

Key Shifts in Market Trends

Economics trumps branding: API costs for top-tier models can now exceed a company's entire salary fund. Infrastructure pivot: Hosting Chinese models on U.S. servers eliminates legal and compliance risks. The inference crisis: The high cost of token generation is becoming a barrier to the deployment of autonomous AI agents.

"Unless elite AI providers learn to sell intelligence for less than the cost of human labor, they risk becoming an unjustifiably expensive luxury for those who don't track their margins."

As Snowflake’s CTO points out, while affordable alternatives (including GLM-4) aren't yet identical to Claude in every metric, their price-to-performance ratio makes them the clear favorites for specific business tasks. Even OpenAI's Sam Altman admits that inference costs have become a "huge problem," particularly for agentic systems that consume tokens at an aggressive rate.

Anthropic’s pricing power is being squeezed between benchmark scores and harsh market realities. Crivello is transparent about his pragmatism: he is willing to return if Anthropic slashes prices, but for now, unit economics beats brand loyalty. The market was promised an era of high-margin automation, but we are instead seeing a price war over a basic resource. We are witnessing "revolutionary" solutions lose ground to aggressive discounting that is rapidly turning cloud computing into a commodity.

AI in BusinessCost ReductionLarge Language ModelsAnthropicDeepSeek