OpenAI and Anthropic built empires valued at nearly 1 trillion dollars by riding a wave that few dared to question: no matter what it costs, as long as the AI works.
But that logic is starting to crack.
The story of Flo Crivello, CEO of startup Lindy, sums up what is happening in the market right now. Earlier this month, he simply decided to migrate 100% of his company’s traffic from Anthropic models — the well-known Claude family — to DeepSeek, a Chinese company that offers open-source alternatives at a much lower cost. The result was immediate and visible.
We did that, and you could see that cost curve just, like, fall off a cliff, Crivello said in an interview from the company’s San Francisco headquarters. According to him, the decision will save Lindy millions of dollars within a few months, although he still expects the roughly 25-person company to spend more on AI than on payroll.
It is a matter of business survival, he said bluntly. That is all it is.
And Crivello is far from alone. He spent nearly five years at Uber before founding his own company and is part of a growing group of founders and executives in the United States trying to rein in AI spending. AI bills have ballooned — sometimes into the billions of dollars — since OpenAI dazzled Wall Street with ChatGPT in 2022 and kicked off a corporate race to deploy the technology in areas like customer support, marketing, and finance.
Uber, for example, blew through its entire annual AI budget in just four months. Meanwhile, OpenAI and Anthropic are racing to go public, having confidentially filed their IPO paperwork in June — timing that does not seem like a coincidence. The market is shifting, the numbers still look impressive, and the window may be smaller than it appears. 👇
The problem nobody wanted to see
For years, the argument was simple: spending heavily on AI was synonymous with staying ahead of the competition. Companies opened their wallets with little hesitation, whether to train their own models or to consume APIs from giants like OpenAI and Anthropic. The reasoning made sense early on, when the competitive edge still justified every line item in the budget. But nobody stopped to calculate where that bill would end up once the technology was actually scaled into everyday business operations. And that is exactly what happened: scale arrived, and costs exploded right along with it.
Receive the best innovation content in your email.
All the news, tips, trends, and resources you're looking for, delivered to your inbox.
By subscribing to the newsletter, you agree to receive communications from Método Viral. We are committed to always protecting and respecting your privacy.
Spending skyrocketed particularly in the area of AI-assisted coding. Developers began consuming tokens — the units of data processed and generated by models — in massive volumes to build tools and services that previously would have required entire teams of programmers. This gave rise to the era of so-called tokenmaxxing and AI leaderboards, where employers encouraged developers to use as much artificial intelligence as possible without worrying too much about the practical outcomes.
The crackdown on that behavior has already begun. Uber reported that it implemented a series of spending tiers on some AI tools, starting with a base level of 1,500 dollars per month, with the option for employees to request access to higher tiers. And the case of the annual budget being blown in four months, revealed by the company’s CTO Praveen Neppalli Naga, is both emblematic and alarming.
Burning through an annual AI budget at that rate is not a simple planning mishap — it is a clear sign that the current pricing models were not designed for large-scale operations. And when a company the size of Uber feels that squeeze, you can imagine what happens to smaller startups operating on much tighter margins without the same financial cushion to absorb surprises like that.
Flo Crivello was straightforward when he explained his decision. Lindy was built around the idea that token costs would drop dramatically over time. That proved true for a while, but the leading model developers, including Anthropic and OpenAI, have been slower to cut prices in recent months. He even left the door open: I hope they cut costs again at some point, but until then, we have options.
The cost pressure hits the finance department
This squeeze is not exclusive to any one company. Jeff Henry, president of consulting at Highspring, says some of the firm’s clients are holding off on investments until they can really start proving an ROI, while others are waiting 12 to 18 months before making major spending decisions. Everyone is going through the same AI spending crunch, he summed up.
On the other hand, Henry makes an important point: there are still countless mid-sized companies that have not even begun experimenting with the technology. AI is not going away, he said. There is no putting this toothpaste back in the tube.
Darren Kimura, CEO of enterprise AI company AISquared, points out that one area where spending is clearly peaking is the use of frontier models for simple tasks that could be handled by cheaper alternatives. In budget terms, it is like using a cannon to kill a fly.
The approach many companies are adopting is called model routing — a technique that matches each task to the most appropriate model for it. It is such a recent concept that, according to Arvind Jain, CEO of Glean, about 95% of enterprise AI usage still runs on frontier models. Kimura is emphatic that this pattern will be unsustainable for most companies in the long run.
Eric Glyman, co-CEO of expense management startup Ramp, notes that finance departments are paying close attention after getting spooked by surprisingly high AI bills. Most CFOs not only did not plan for this accelerated growth in annual budgets, but they also do not have good tools to manage it, he explained. Suddenly you have this third pillar that showed up, which is spending on tokens and intelligence. It is not a clean expense category.
The race to IPO and what is at stake
The confidential IPO filings by OpenAI and Anthropic in June did not come out of nowhere. There is a very clear strategic rationale behind the move, as Gil Luria, equity analyst at D.A. Davidson, points out. According to him, the current growth rates at these companies are the fastest they will ever have, as a matter of basic math. That is a good reason to go public now, as is the concern that some of their biggest enterprise clients may start clamping down on runaway token spending, he said.
The numbers, to be fair, are still impressive. Anthropic reported an annualized revenue run rate of 47 billion dollars in May, a massive leap from the roughly 10 billion recorded for the entire previous year. OpenAI, meanwhile, had its annualized run rate tracking toward about 25 billion dollars earlier this year, compared to 13.1 billion in revenue generated in 2025. Listing now, while the numbers still shine, could be a smart move.
Luria reinforces this reasoning: There has to be some period in the future where there is a rationalization of spending by enterprises, and that could be a headwind ahead for Anthropic and OpenAI. That creates some urgency to go public before we see that happen. When contacted, Anthropic declined to comment and OpenAI did not respond to requests for a statement.
The competition is heating up
As companies become more price-sensitive, OpenAI and Anthropic have to contend with deep-pocketed competitors that are targeting low-cost model development. And the curious detail is that these competitors are, in many cases, their own investors.
Microsoft, which has already poured more than 13 billion dollars into OpenAI and up to 5 billion into Anthropic, unveiled a series of new low-cost models. The company also emphasized that its coding product, GitHub Copilot, will route users to the most appropriate model for each task. In an essay published in June, CEO Satya Nadella made his vision clear: The last thing any of us want is a world where every company in every industry is ceding value to a handful of models that devour everything they see. If all value is accrued by just a few models, the political economy simply will not tolerate that.
Amazon and Google are also ramping up investments in models aimed at enterprise users. Peter DeSantis, Amazon‘s top AI executive, said he expects the company to be able to compete with frontier models from OpenAI and Anthropic by next year. Amazon‘s bet is on its in-house chips to develop models at a lower cost than its rivals. AI has a cost problem, DeSantis said. If we really want AI to transform everything, costs need to be different.
Google, for its part, highlighted affordable offerings at its annual developer conference, introducing Gemini 3.5 Flash — a lighter addition to its model lineup — available at half or, in some cases, close to a third of the price of comparable frontier models, according to CEO Sundar Pichai.
PitchBook analyst Harrison Rolfes neatly sums up the comfortable position these tech giants hold: Microsoft and Google have the infrastructure and the capacity, the entire stack, where they can step in and push both OpenAI and Anthropic into a corner. They are probably waiting on the sidelines to watch the two fight it out and figure out where they are not doing well.
The market is changing fast
What the move by companies like Lindy reveals is a behavioral shift that could reshape the entire competitive landscape of the AI industry. When startups and mid-sized companies start swapping out major proprietary models for open-source alternatives, they are not just saving money — they are sending a clear signal to the market that technical quality alone is no longer enough to justify the price difference.
At the same time, it is important not to oversimplify this equation. Open-source models like DeepSeek have come a long way, but there are still enterprise use cases where OpenAI and Anthropic models deliver a level of sophistication, reliability, and support that open alternatives have not fully replicated yet. Companies dealing with sensitive data, strict regulations, or complex integrations tend to keep prioritizing the major players — at least in the short term.
It is worth noting that both OpenAI and Anthropic are already reacting to this more cost-conscious environment. OpenAI rolled out analytics features and updated controls for businesses, allowing administrators to track credit spending, set usage limits, and give employees visibility into their budgets. Anthropic, in turn, released a set of controls in August that let organizations provision users, view analytics, and set spending limits at both the organization and individual levels.
As for going public, neither company has given an exact timeline for their stock market debuts. The New York Times reported that OpenAI is leaning toward waiting until next year. The pressure to go public ultimately comes down to the need for fresh capital. With Anthropic and OpenAI increasingly competing against their own biggest backers, the IPO market may be the best avenue for raising funds, especially since their capital needs have grown too large for most venture capital and private equity firms. Many of the traditional pockets of capital are drying up, noted Dharmesh Thakker, a partner at Battery Ventures.
What becomes clear when looking at this landscape as a whole is that the narrative that AI is worth any price is being replaced by a far more pragmatic question: what is the real return on this investment? That shift changes everything — from how companies choose their AI vendors to how investors will evaluate the IPO filings from OpenAI and Anthropic. The sector is growing, that much is undeniable, but the next chapter will be written by those who can balance innovation with financial sustainability. And that balance, for now, is still being found. 🧩
