Anthropic executives said companies should not respond to rising artificial intelligence costs by reducing or halting use.

Business Insider reported on Tuesday that they stressed operating strategies that deliver the same results at lower cost are more important than blanket budget caps.

Angela Zhang (앙겔라 장), head of product for the Claude platform, said on a Sequoia Capital podcast that “stopping all AI use is the wrong idea,” adding that she is seeing cases where some clients respond that way. Caitlin Lese (케이틀린 레세), head of platform engineering, said it is natural for companies to weigh costs when deciding how to deploy AI, but added that “setting only a cap and being trapped inside it is the real risk factor.”

Anthropic's concern comes as debate has grown across U.S. companies about recouping AI investments. AI bills are swelling quickly, while some executives have not confirmed a return on investment, or ROI, sufficient to justify the spending.

Anthropic presented operational efficiency, rather than cutting usage, as a solution to cost control. Zhang explained that AI spending can suddenly rise due to so-called shadow IT, where employees adopt AI models individually outside formal internal approval processes. She added, “What we recommend to customers is not to stop innovating,” and said that if AI adoption allows them to ship products faster and raise operational efficiency, that itself is an outcome.

Lese said operating strategies that deliver the same results at lower cost will become the next competitive point. She said running high-performance models overnight differs from producing the same result with a smarter strategy, adding that companies will now move toward designs that match performance with cost.

The discussion also connects to competition in the AI infrastructure market. Companies such as Vercel are targeting cost-sensitive corporate demand by promoting routing approaches that send requests to the most suitable model depending on the nature of the task. The market expects such routing demand to continue as long as AI token costs remain high.

Anthropic is also reviewing this direction within its own ecosystem. Zhang said, “A router within the Claude domain is meaningful,” adding that the platform is being designed around Claude and that the company is focused on enabling Claude to solve a wide range of problems well. The remarks are seen as keeping in mind a strategy to balance cost and performance within its model lineup, rather than simply lowering prices.

AI chatbot price competition is also intensifying. Sam Altman (샘 알트먼), OpenAI's chief executive, claimed the company's new GPT-5.6 family model charges half the price when performing the same tasks as Anthropic's Fable 5, with token efficiency about twice as high.

As AI companies must respond at the same time to corporate customer spending pressure and ROI validation, the competitive focus is increasingly likely to shift beyond model performance to cost efficiency and operating methods. For companies preparing to list, the cost debate remains a factor that could affect market assessments.