Corporate America’s approach to artificial intelligence has undergone a rapid reversal, shifting from a culture of “tokenmaxxing” — where companies wore heavy AI usage as a badge of honor — to one of “thrift-maxxing,” according to a report in The Wall Street Journal. Fed up with ballooning costs, companies big and small are starting to use lower-priced models, including some built in China, alongside premium offerings from OpenAI and Anthropic.
Mike Saeks, a field chief technology officer at Cursor, a startup Elon Musk’s SpaceX is acquiring for $60 billion, said the most powerful systems are unnecessary for routine tasks. “It’s like driving a Lamborghini to go to the grocery store to pick up milk when that was designed to be raced around a track,” Saeks told the Journal.
Cursor recently ran an experiment to evaluate the cost savings. Building a web browser from scratch on OpenAI’s GPT-5.5 cost a little more than $10,000. Using Cursor’s Composer model in combination with Anthropic’s Opus 4.8 cost $1,339.
“There’s zero loyalty that I’m seeing,” Marty Kausas, chief executive of the AI customer support platform Pylon, told the Journal. “It really feels like a bloodbath right now.” Kausas said his company has received months of unlimited free usage from vendors, including around $1.6 million in free tokens from one vendor, $65,000 from another, and $10,000 from another.
David Casem, chief executive of Telnyx, a startup that provides infrastructure for AI agents, said his company was forced to switch models after Anthropic stopped allowing third-party operating systems on its subscriptions. “We did the math and it was going to be like 100 grand per day,” Casem told the Journal. Telnyx now uses a family of models from Chinese startup Z.AI to power its 1,400 agents, costing around $100 per agent per day, while Anthropic’s most powerful model, Fable, acts as a conductor and OpenAI’s Sol handles review.
The trend is threatening the business models of OpenAI and Anthropic as they prepare for public listings. To fight back, the Journal reported, the companies are offering partnerships, tens of thousands of dollars in incentives and heavily subsidized AI usage to lock in customers. OpenAI released a newer model, GPT 5.6 Sol, which it said is trained to be much more token-efficient. Anthropic released a powerful, lower-cost model on Friday. Both companies said they support open-weight models.
The adoption of cheaper models has injected a geopolitical dimension into enterprise technology decisions. Anthropic and OpenAI have accused Chinese AI startups, including DeepSeek, MiniMax and Moonshot AI, of ripping off their technology. Some executives and Trump administration officials have suggested a ban on such models, the Journal reported, while others argue that those favoring restrictions are trying to stifle competition. On Friday, a group of tech companies including Nvidia, Microsoft and Palantir signed a letter in support of open models, urging U.S. policymakers to exercise caution on potential restrictions.
Executives across sectors said they expect the era of cheaper AI to persist regardless of the geopolitical fight. Barry McCardel, co-founder and CEO of the data analytics platform Hex, said about 50% of his company’s customers have adopted Kimi, a model produced by China-based Moonshot AI, in the last two weeks. “We are continually assessing how much we want to commit to any one lab given how dynamic a moment this is,” McCardel told the Journal.
Zoom’s chief technology officer, Xuedong Huang, said the company has saved substantially by fine-tuning Meta’s open-weight Llama model for three years, alongside a combination of Anthropic, OpenAI and open models. Gabe Pereyra, president of the legal AI startup Harvey, said the company trained its own model, GLM-5.2, and equipped it with a tool that can “call” Anthropic’s Fable 5 model when it determines a task is hard.
Saeks said the pace of change in the AI model market is accelerating, making long-term commitments risky. “The best model for a task used to change every few months,” he said. “Now it feels like it’s happening multiple times per week.”