What Anthropic’s IPO could mean for your Claude bills

Now that Anthropic’s filed its S-1, people are asking the big questions: Is the company overvalued? Can its growth last? Is AI a bubble? It’s also likely more than one CFO is thinking, “I wonder what this is going to do to my Claude bills?”

That’s anybody’s guess, Justin Etkin, CEO and cofounder of spend management platform Tropic, told CFO Brew. But he doesn’t expect dramatic price hikes in the near future.

That’s because customers treat LLMs more like commodities than like traditional software, Etkin said, readily switching between models and offerings from rival labs. As Siqi Chen, CEO of CFO.ai, told us, “Internally, for developer purposes, we probably use Anthropic models more than OpenAI at this point. But it changes from week to week. Whoever has the newest models that week, we will go ahead and spend all of our budget on there.”

Price wars ahead? “Pricing does play a major factor in who’s ultimately going to be the preferred option,” Etkin said. Competition between AI companies is already causing price wars, he added, pointing out that OpenAI made “some very drastic price cuts” when Anthropic’s growth began to take off, “and Anthropic had to respond.”

CFOs and CIOs have also grown more sophisticated and stricter about token usage in the past year, Etkin said. “The process for accessing and requesting more tokens is very secure,” he said, predicting that we won’t see as many stories of companies receiving shockingly high AI bills as we did early in 2026.

But the same phenomenon may throttle LLM labs’ growth. “The meteoric growth that we saw for Anthropic,” Etkin observed, “was built on a lot of consumption that was completely unmitigated within organizations.”

Expect more complex pricing. Increased caution around token use, as well as the IPO, may force Anthropic to become savvy about pricing. “When they’re owned by the public and these shareholders, they are probably going to have to do things differently on both sides of the ledger,” Jason Flynn, AI evangelist at agile training firm Scaled Agile, told us. Anthropic may have to focus both on earning more and spending less on training, which could mean it rolls out new models less frequently.

Flynn predicts that costs will go up for enterprise and power users, and that Anthropic’s pricing will become more complex, perhaps involving different tiers and premium options. If it does release new models less frequently, it will need to “make what we do have at the frontier, at the leading edge, seem more exclusive,” he said. “I would bet they’re going to start paying hundreds of thousands of dollars to go-to-market specialists who are really good at rapidly testing a lot of different offers, packages, and pricing.”

What CFOs can do to keep costs down. To retain the greatest pricing flexibility, both Etkin and Flynn recommend that companies use multiple LLM models. “The more deeply you get embedded with one provider over another, the more challenging it might be to make a switch,” Etkin said, adding that he tells customers to diversify their workloads across multiple AI companies’ tools.

CFOs should also keep an eye on rollouts of frontier models. When LLM labs introduce new models, Etkin said, prices for earlier models often go down. “What you can bank on is there will be price compression for older versions, more legacy versions of these models that are really quite good at what they do.”

Open-source models can form part of the mix, too. “I think there is a lot of work that is done in large enterprises that could be handed off to open source models,” Flynn said. Etkin said he’s seen more companies opting for them, “particularly those that are heavy spenders,” but that security and compliance concerns are, for now, keeping them from becoming more pervasive.

Sometimes, Flynn said, users believe they need to use the latest LLM models available—the problem being that these are often the most expensive. Training staff on AI can help them learn which models are actually the most appropriate for their needs. Companies can also install apps called routers, which Etkin says direct people to the most practical LLM model to use.

But don’t neglect productivity when making pricing decisions, Chen said. His company’s AI spend has gone up dramatically this year, he said, and he individually spends around $100,000 a month on LLMs. The cost is worth it, he argues, stating that AI has given his team of 10 engineers the “measurable output of 800 engineers,” as gauged by his productivity tools.