My AI Bill Skyrocketed — and I’m Glad It Did

I’m a technologist leading a custom software studio, and AI is woven into nearly every product we build and most of our daily work processes. So when my team’s monthly Anthropic bill spiked sharply recently, frustration would have been the expected reaction. Instead, I was relieved.

The price most businesses pay for AI right now is not the real price—it’s a lie. The longer this false pricing persists, the more dangerous it becomes for the long-term future of human expertise and collaborative work teams. The sooner companies start paying the actual cost of the AI they use, the better the outcome will be for everyone.

Let’s start with the hard, unvarnished math. By 2025, OpenAI generated roughly $13 billion in annual revenue, but posted a $21 billion operating loss. That adds up to the company spending almost $1.60 for every single dollar of revenue it brought in. This is no accident or temporary rough patch. Top frontier AI labs are intentionally pricing their services below the actual cost to serve customers, all to capture market share before businesses fully understand the technology’s true value and long-term cost. Venture capital firms and major cloud providers are currently covering that massive pricing gap, and they will eventually want their money back.

Most industry analysts already expect AI API prices to climb steadily once this subsidy-fueled market share race comes to an end. When AI feels absurdly cheap today, that’s exactly what it is: cheap for you, temporarily, and being subsidized by someone else’s dime.

Here’s why I’m happy to see that my bill is finally going up. This artificially low subsidized price distorts the most critical question every organization is currently navigating: should this work be done by a human, or by a machine? At today’s fake, suppressed rates, AI wins cost arguments that it has no business winning. Move the price to its actual, real-world level, and many of those “obviously cheaper to automate” decisions quietly flip back to favoring human workers.

My biggest concern isn’t AI becoming too expensive. It’s what we’re choosing to do right now, before prices normalize. Companies are being pushed to make permanent, irreversible business decisions based on a temporary pricing anomaly. If you eliminate your team of writers, customer support specialists, and in-house analysts today while AI is on sale, you aren’t just cutting a budget line item. You lose the people, their client relationships, your company’s hard-won institutional memory, and the nuanced human judgment that AI still cannot touch, even with the latest model advancements.

You can renegotiate or cancel a software contract whenever you need to. But you can’t easily rehire the relationships and institutional knowledge you once had.