OpenAI, a new, humane AI that makes real spending as users value efficiency

Flo Crivello’s expenses were insane and there was only one way to manage them.

Earlier this month, the AI ​​startup’s CEO Lindy, 34, separated his company’s traffic from Anthropic’s Claude model and shifted 100% of it to DeepSeek, a Chinese company that makes a cheaper open-weight alternative.

“We did that, and you could see the cost curve come crashing down into the ground,” Crivello said in an interview at the company’s San Francisco headquarters. He said the decision will save Lindy millions of dollars in the coming months, but said the company, which has about 25 employees, expects to spend more on AI than on payroll.

“This is a matter of survival for the business,” Crivello said. “That’s it.”

Mr. Crivello previously spent about five years with the company. Uberis one of a growing number of founders and executives across the country trying to rein in spending on artificial intelligence. Since OpenAI first captivated Wall Street with its ChatGPT chatbot in 2022, AI billings have ballooned, sometimes reaching billions of dollars. This started a rush for companies to adopt technology in areas such as customer support, marketing, and finance.

In particular, costs rose in the area of ​​AI-assisted coding as developers injected tokens into the creation of new tools and services. Previously, you needed a team of coders. This ushered in the era of so-called token maxing and AI leaderboards, where employers encouraged developers to use as much AI as possible without worrying about the consequences.

A crackdown is underway. Uber announced this month that it is introducing a series of spending tiers for some of its AI tools, starting at a basic level of $1,500 per month, but that employees can request access to higher tiers. In April, Uber CTO Praveen Neppalli Naga revealed to The Information that the ride-sharing company exhausted its annual AI budget in just four months.

OpenAI and Anthropic have been the main beneficiaries of the spend-whatever-costs mentality, which has fueled their exponential growth rates and pushed both leaders in AI models to valuations approaching $1 trillion.

Now, as they prepare for potentially historic IPOs (both secretly filed in early June), the mood around AI is changing, with business leaders like Crivello no longer willing to put money into Anthropic or OpenAI unless they can see a clear return on investment.

“Anthropic and OpenAI’s current growth rate is the fastest it’s ever been. This is largely a basic math problem,” Gil Luria, an equity analyst covering tech companies at DA Davidson, told CNBC. “This is a good reason to go public now, as well as concerns that some of our largest enterprise customers may begin to limit uncontrolled token spending.”

Anthropic last reported an annual run rate of $47 billion in May, up from the roughly $10 billion in revenue it posted for all of last year. OpenAI’s run rate is reportedly on pace to approach $25 billion earlier this year, up from the $13.1 billion in revenue it generated in 2025.

Although the numbers are still dizzying, going public soon could be strategic.

“There’s going to be a certain point in the future where companies are going to streamline their spending, and that could be a hot spot for Anthropic and OpenAI,” Luria said in an interview. “There’s a sense of urgency to make it public before that happens.”

Anthropic declined to comment for this story. OpenAI did not respond to a request for comment.

Crivello is a big fan of Anthropic, but said his company has long grappled with “unsustainable” AI costs.

Lindy was built on the idea that the cost of a token, a unit of data processed and generated by an AI model, would decrease dramatically over time, Crivello said. While that has proven true for a while, major model developers like Anthropic and OpenAI have slowed down in price reductions in recent months.

Crivello said he would be willing to switch the Lindy back to the Claude model if prices come down.

“Hopefully at some point they’ll cut costs again, but until then you have options,” he said.

Jeff Henry, president of Highspring’s consulting division, said some of his clients are holding off until “they can actually prove ROI,” while others are waiting another 12 to 18 months before making big spending decisions.

“Everyone is experiencing the same spending squeeze on AI,” he said.

However, he said there are countless mid-sized companies that have not even started experimenting with AI yet.

“AI is not going away,” Henry said. “That toothpaste never goes back into the tube.”

Darren Kimura, CEO of enterprise AI company AISquared, said one area where AI spending is “definitely” peaking is in the use of cutting-edge models, also known as frontier models, for simple tasks that can be accomplished with cheaper alternatives.

Some companies are turning to so-called model routing, which matches the right tasks to the right models. Because this is such a new technology, Glean CEO Arvind Jain says that about 95% of enterprise AI usage is still performed in frontier models.

Kimura said this approach would be “unacceptable” to most companies in the long term.

DA Davidson’s Luria said pricing in the market is still “unsophisticated,” but both OpenAI and Anthropic are adapting to an increasingly budget-conscious environment.

Earlier this month, OpenAI introduced analytics and modern management capabilities for businesses, allowing managers to analyze credit spending across the workplace, set spending limits, and give employees visibility into available budgets. In August, Anthropic introduced a set of controls that allow customers to provision users, view analytics, and set spending limits at the organizational and individual level.

Eric Greiman, co-CEO of expense management startup Lamp, said finance departments are paying close attention after being hit with surprisingly large AI bills.

“Most CFOs not only didn’t plan for this rapid growth in their annual plans, but they also don’t have good tools to manage it,” Greiman said in an interview. “Suddenly there’s a third pillar of spending through tokens and intelligence, which is not a pretty spending area.”

As companies become more price-sensitive to AI, OpenAI and Anthropic will have to contend with deep-pocketed competitors looking to develop lower-cost models.

Microsoft, which has poured more than $13 billion into OpenAI, as well as $5 billion into Anthropic, announced a series of new lower-cost models earlier this month. The company also highlighted that its AI coding product, GitHub Copilot, routes users to the best model for the task.

Microsoft CEO Satya Nadella said in a June essay that the industry needs to avoid concentration of power in the hands of a few large providers.

“No one wants a world where every company in every sector cedes value to a few models that eat up everything in sight,” Nadella wrote. “Political economy would not tolerate it if all the value was created by just a few models.”

Amazon and google The company is also stepping up investment in models for business users.

Amazon’s head of AI, Peter DeSantis, told CNBC this month that he hopes the company will be able to compete with OpenAI and Anthropic’s frontier models “next year.” Like Microsoft, Amazon also invests in these companies.

DeSantis said in February that Amazon would use its own chips to develop models that are cheaper than its competitors.

“AI has a cost problem,” he said in an interview with the Wall Street Journal. “If we ultimately want AI to transform everything, the costs have to be different.”

google made a concerted effort to highlight affordable AI products at its annual developer conference last month. CEO Sundar Pichai said the company introduced the Gemini 3.5 Flash, a lightweight addition to its model suite, available at half, and in some cases nearly a third, the price of comparable Frontier models.

“Microsoft and Google have the infrastructure and capabilities, the entire stack, to step in and take a hard line on both OpenAI and Anthropic,” PitchBook analyst Harrison Rolfes said in an interview. “They’re probably waiting on the sidelines to take the fight and see what’s going wrong.”

Regarding listings, none of the major modeling companies has disclosed the exact timing of their planned debuts. The New York Times reported on Thursday, citing people involved in the deliberations, that OpenAI is leaning toward delaying it until next year.

Pressure to go public may revolve around the need for capital. As Anthropic and OpenAI increasingly compete with their biggest backers, the IPO market could become the best vehicle for raising new capital, especially since their capital needs are too large for most venture and private equity firms.

“A lot of the traditional capital is drying up,” said Dharmesh Thakkar, general partner at Battery Ventures. “All the institutional investors who can invest in these companies have already saved up.”

WATCH: Anthropic slams Alibaba over campaign to extract AI capabilities