Higher AI spending ≠ Higher ROI: Here’s what leaders are learning the HARD WAY about AI-related costs.

From the desk of Miles Everson:

Happy Friday!

I’m excited for another edition of "Mindfulness by Miles."

In these articles, I talk about health, career longevity, AI, business, and the future of work. My goal is to equip you with the insights needed to navigate both your personal and professional life with greater clarity and energy.

Today, we’ll talk about the skyrocketing costs of AI usage.

Eager to know more?

Continue reading below.

Ridesharing giant Uber made headlines recently when reports surfaced that the company burnt through the entirety of its 2026 artificial intelligence (AI) coding tools budget in just four months.

Yes, you read that right. Uber’s AI budget for a whole year was spent in the span of four months.

In response to this, Uber president and CEO Andrew Macdonald said it’s difficult to find a link between his company’s AI usage and innovations meant to better serve customers.

He said “that link is not there yet” and “maybe implicitly there’s more that is getting shipped, but it’s very hard to draw a line between one of those stats and ‘Okay now we’re actually producing like 25% more useful consumer features.’”

Reports state Uber employees’ usage of AI skyrocketed after the firm incentivized employees to increase usage through the creation of an internal leaderboard that tracked AI tool utilization.

Uber isn’t the only firm who is actively promoting tracking AI usage.

Big companies like Alphabet , Meta , and even banking giant JPMorgan Chase have been reported to have set up mandates that track the AI utilization of their employees.

With companies mandating AI usage, the rise of “tokenmaxxing” —the practice of tracking and maximizing AI token usage has gained popularity.

While this practice looks good on paper, it has brought up a bunch of issues that companies are struggling with right now.

Chief among those issues is cost .

The Price of AI Adoption and Usage

AI adoption is becoming more expensive right now due to changes in how companies are charged.

AI model usage is billed through what’s called token-based billing. Before, users were charged a flat fee for using models. However, this changed this year when AI companies started charging based on tokens consumed.

Token-based billing is more expensive since users are charged based on the consumption of AI tokens.

For example: If a user sends a prompt that costs 600 tokens and receives an answer that’s worth 210 tokens, here’s how this interaction will be billed:

• Let’s assume that the cost of one token is USD 0.00300
• The total price per request, based on our example, would be around USD 0.00825.

That total may look cheap, but imagine if that were multiplied by 15,000 users, that cost will amount to USD 123 per day or roughly USD 3,700+ per month.

Unfortunately, the example we provided is just a conservative estimate. Token consumption can vary widely depending on use case and prompt.

Taken together, it becomes clear just how expensive AI usage can be when complexity and scale are both factored in.

It’s no wonder then that Uber managed to burn through its AI budget in just four months.

… and it’s not the only company that’s facing the exact same problem. Microsoft , for example, recently canceled a vast swath of its Claude code licenses for employees due to how expensive Anthropic’s AI model is.

This brings us to our next problem: Cost savings .

AI Automation = Higher Cost Savings

AI gained widespread adoption and hype because of the promise of cost savings. Unfortunately, instead of this expectation, reality has taken a different turn.

According to data from Bain & Company , among companies that targeted cost savings of between 11% to 20%, 40% of them managed to save only 10% or less.

That means there’s a disconnect between reality and expectation.

AI budgets and related investments continue to rise. However, instead of automation leading to higher cost savings, it seems companies are saving less and spending more.

The AI Spending Problem

As demonstrated by Uber, rising AI expenditures are increasingly difficult to defend, given the challenge of measuring their effect on metrics such as customer satisfaction. Compounding this, Bain & Company's research indicates the anticipated cost savings from AI have not yet been realized.

So, does this mean companies should just scale back their AI adoption efforts altogether?

The simple answer is NO .

AI is still a transformative technology that holds lots of promise.

What needs to happen is a mindset shift.

Instead of operating with a blank-check AI budget, leadership teams and managers should shift from unchecked adoption to disciplined and mindful utilization.

Doing this includes:

• Setting reasonable targets and expectations for AI usage.
• Adopting the attitude that AI’s role is to augment human labor, not replace it.
• Identifying which business problems need AI augmentation.

The true measure of successful AI integration isn’t about maximizing token usage or operating an AI budget with an eye-watering amount of money; it’s about mindfully deploying AI to solve specific business problems.

That way, leaders, managers, and even independent workers in some cases could set reasonable targets for tracking AI’s impact and keeping their budgets in check.

AI is one of today’s great innovations. However, that doesn’t mean financial sustainability should be set aside in the race to adopt this new technology.

For a daily version of this newsletter, please subscribe here.