In today's Finshots, we explore why the AI boom may be on the verge of creating its own FinOps moment.

But before we dive in, if you’re someone who loves keeping tabs on the world of business and finance, hit subscribe if you haven’t already. And if you’re already a subscriber, thank you! Maybe forward this to someone who’d enjoy our stories but hasn’t discovered us yet.

Now, onto today’s story.

The Story

Silicon Valley spent two years convincing companies to use more AI. Now the companies that listened are the ones in trouble.

A few weeks ago, Microsoft was scaling back access to Anthropic's Claude Code for its engineers. Around the same time, Uber revealed that it had already exhausted its entire 2026 AI budget within the first four months of the year. Its COO (Chief Operating Officer) Andrew Macdonald admitted that it was becoming difficult to connect rising token (data units used by AI models) usage with meaningful outcomes. That’s simply to say that the costs didn’t justify the results.

And some companies like Amazon have even started pulling back on this “tokenmaxxing” culture, where employees were encouraged to maximize AI usage wherever possible.

Even if AI makes employees more productive, then encouraging people to use more AI should create better outcomes.

But what happens when the AI bill arrives?

It's a question that's becoming harder to ignore. According to a recent report, One AI consultant told Axios about a client who spent more than $500 million in a single month serving Claude users. That's burning through nearly $17 million every day just to keep the models running.

To understand why this matters, we need to talk about tokens.

Think of tokens like a taxi meter that never stops running. Every time you ask an AI agent a question, upload a document or make follow up requests, the meter ticks. One ride may be cheap. But when thousands of employees are taking similar rides all day and everyday, the fleet bill becomes something else entirely.

To users like you and me, the tokens feel invisible. But to AI companies, they're anything but that.

Behind every token sits a real-world cost — GPU time, electricity, and data center capacity. And because most AI providers charge businesses based on token consumption, tokens have become the closest thing the industry has to a common currency.

The more AI a company uses, the more tokens it spends. And the more tokens it spends, the larger its AI bill becomes.

Of course, none of this means AI spending is a bad thing. After all, if AI helps a software engineer complete ten hours of work in two, the return could easily justify the cost.

But when something becomes cheap and easy to access, people usually end up using far more of it than they planned.

This isn’t the first time something like this has happened. When cloud computing was still new, companies were sold on the idea of paying for the computing power they actually used instead of buying expensive servers.

Netflix for example, became one of the poster children when it migrated its infrastructure to Amazon Web Services (AWS) after years of managing its own data centres. The move allowed the streaming giant to expand its computing capacity whenever demand surged and scale back when it didn’t.

To put that in perspective, if 50 million people decide to watch the season finale of a hit show on the same evening, Netflix doesn't suddenly need to build a brand-new data centre. It can simply tap into more cloud resources. It sounded brilliant. No more guessing future demand or buying hardware years in advance. And as you can imagine, businesses loved it.

But companies soon discovered a little problem. While spinning up new resources was easy, keeping track of them wasn't. That’s simply because as computing resources became easy to access, usage exploded. Developers spun up new workloads and forgotten virtual machines kept running for months. This meant that storage bills ballooned in the background.

Before long, many companies discovered that managing cloud costs had become a challenge of its own.

The situation got so bad that McKinsey found some companies could cut nearly 20% of their cloud spend if they knew where to look.

But abandoning the cloud wasn’t a solution or even an option. If anything, companies wanted to use more of it. You could say that cloud's success had created an entirely new challenge.

So the only way out was to create an entirely new discipline dedicated to managing it.

Today, that discipline has a name. It’s called FinOps.

But nobody sat down one day and decided to invent FinOps. It came up because companies found themselves in an unusual situation ― the cloud cost crisis.

And that slowly turned into an industry of its own. Today, companies hire FinOps specialists, software vendors sell cloud cost-management tools, and professionals can even earn FinOps certifications. In other words, what started as a billing headache eventually became a career path.

And that brings us back to AI. Just like cloud computing before it, AI is becoming easier to access, scale, and consume.

Today, most companies still measure AI adoption. They want to know how many employees are using AI tools, how often they're using them, and whether AI has become part of everyday workflows.

That's a perfectly reasonable goal when a technology is new.

But mature technologies can’t be judged by usage alone. They have to be judged by outcomes.

And that simply means that if companies want employees to use AI, then they may have to ask and answer a more difficult question: Which AI usage is actually worth paying for?

But this isn’t just a story about AI costs and budgets. Over the past two years, many companies have cut jobs believing AI would help the remaining employees do more work with fewer people. Fair enough. But like any investment, it has to show results.

Jefferies estimates that AI-related spending could touch $4.7 trillion globally by 2029. That’s a massive amount of money being poured into a technology whose business value companies are still trying to figure out.

And as the spending rises, so does the pressure to prove that AI is actually helping.

Otherwise, companies could end up in an awkward situation where they have fewer employees, bigger AI bills, and no real improvement in results.

So maybe, just like cloud computing once did with FinOps, AI will also have to prove that the value it creates is worth the cost. And how it does that, is something we’ll have to wait and see.

Until next time…

If this story helped you understand why companies are cutting back on AI usage, share it with your 'tokenmaxxing' friends, family members or even strangers on WhatsApp, LinkedIn and X.

How Strong Is Your Financial Plan?

You've likely ticked off mutual funds, savings, and maybe even a side hustle. But if Life Insurance isn't a part of it, your financial pyramid isn't as secure as you think.

Life insurance is the crucial base that holds all your wealth together. It ensures that your family stays financially protected when something unpredictable happens.

If you’re unsure where to begin, Ditto's IRDAI-Certified insurance advisors can help. Book a FREE 30-minute consultation and get honest, unbiased advice. No spam, no pressure.