OpenAI Discounts Lure Anthropic Customers, Citing Up to 90% Savings

What happened

OpenAI is offering steep discounts to win over customers of rival Anthropic, and some of those customers say the move has cut their AI costs by as much as 90 percent. Reporting from The Decoder names Rippling, Replit and DataCamp as companies that switched and reported savings of up to that level.

The move is a commercial one. Anthropic's growth has reportedly slowed, and OpenAI appears willing to accept thinner margins to gain market share. The pricing push is not happening in isolation. Chinese competitors, including DeepSeek, are adding further downward pressure on the price of AI model access.

The Decoder's account rests on vendor and customer statements about savings, and the broader sources do not provide independent cost audits or a full breakdown of the discount terms.

Why it matters

For organisations deploying AI, this shows that the cost of running a model is becoming a negotiable variable rather than a fixed input. When a provider will trade margin for share, buyers gain leverage, and procurement teams that treat AI as a settled, single-vendor decision may find that assumption no longer holds.

The story also highlights a wider shift in how AI services are sold. As models become more interchangeable for many everyday tasks, price, switching ease and perceived reliability start to carry more weight in the choice. Leaders responsible for digital transformation will need to judge AI suppliers on total cost of delivering a good outcome, not on the headline rate per request.

By the numbers

• Up to 90 percent reduction in AI costs, as reported by customers that switched to OpenAI.
• Three named companies (Rippling, Replit and DataCamp) cited as examples of the savings.

The Renascence take

Most headlines will read this as a price war between two AI labs. The more useful reading is that a discount of this size tells you how much of your AI loyalty was about habit and convenience rather than the product itself. If a 90 percent saving can pull a customer across, the switching cost was never the real barrier.

Treat a headline discount as an anchor, not a benchmark. Customers who move because of price have shown that they can be moved again, which means the experience they get after the switch will decide whether they stay. Operators should measure cost per successfully completed customer task, track quality and escalation rates before and after any change of model, and run a controlled switch on a small journey before committing. Savings that degrade resolution or raise repeat contacts are not savings.