Uber Blew Its AI Budget by April – What It Means for AI Usage Costs in 2026

If you thought rolling out AI in your business would be a neat, predictable “add-on cost” like an extra Microsoft 365 licence… we’ve got some news.

You’re not alone – even Uber has reportedly blown through its entire 2026 AI budget by April.

Yes. April.

And this isn’t because someone in finance forgot a zero. It’s because AI doesn’t behave like traditional software. Especially once people actually start using it at scale.

So what does that mean for businesses rolling out Microsoft Copilot, ChatGPT-style tools, or any AI-driven workflow?

Let’s break it down.

The Uber problem: when AI usage scales faster than budgets

Uber recently gave thousands of engineers access to AI coding tools like Claude and Cursor.

Within a few months:

• 84% were actively using “agentic” AI workflows
• AI-related costs reportedly increased ~6x since 2024
• Their full-year AI budget was exhausted before Q2

Not ideal.

But here’s the important bit – it wasn’t waste. It was adoption working exactly as intended.

The problem is that traditional budgeting assumes:

“If 1 user costs X, then 1,000 users cost 1,000X.”

AI doesn’t do that.

Why AI usage costs break traditional IT budgeting models

Most software costs are predictable:

• Per user
• Per month
• Fixed licensing tiers

Microsoft 365 Copilot (and most modern AI systems) introduces something more fluid at the advanced end (while entry-level plans often remain fixed per user):

• Usage-based processing (prompts, queries, content generation)
• Variable workloads per employee
• “Agentic” behaviour (AI doing multi-step tasks automatically)

In plain English:

Some employees will use Copilot occasionally. Others will absolutely send it.

And that’s where budgets start getting… unpredictable.

What this means for real-world business adoption

If you’re rolling out AI tools across your organisation, you’re not just buying “seats”.

You’re introducing:

•  A new way of working
• A new consumption model
• A cost variable that scales with enthusiasm (and curiosity)

And if you’ve ever seen a team discover a tool that makes their job easier… you already know how that ends.

One person: “I’ll use it for emails.”

Another person: “I rebuilt our entire reporting workflow with it.”

Finance: “Sorry, what happened to the budget?”

The hidden AI usage costs that nobody’s modelling properly yet

The biggest mistake we’re seeing isn’t overpaying for licences.

It’s underestimating:

• How much people will actually use AI once it’s good
• How quickly usage spreads across teams
• How “small tasks” turn into continuous AI workloads

AI doesn’t sit quietly in the background.

It gets used.

A lot.

So… should businesses be worried?

Not necessarily – but they should be intentional.

The winning approach isn’t “limit AI usage”.

It’s:

Treat AI like a business resource, not a software checkbox.

That means:

• Defining high-value use cases first
• Understanding where Copilot actually drives ROI
• Avoiding “let’s turn it on everywhere and see what happens”
• Planning for usage growth, not static adoption

Because the uncomfortable truth is:

If AI is working well, your costs will go up before they stabilise.

What we recommend

For organisations introducing AI productivity tools, a staged approach usually works best:

1. Start with a defined pilot group (not full deployment)
2. Focus on high-impact roles first (knowledge workers, analysts, ops teams)
3. Monitor usage patterns and cost behaviour for 30–90 days
4. Refine governance rules around how and where it’s used
5. Expand licensing in waves based on proven ROI and stable usage patterns

Or more simply:

Don’t turn it on everywhere and hope for the best. Prove value first, then scale it.

Final thought

Even the most sophisticated organisations are still learning how AI behaves at scale.

If a company like Uber can underestimate AI usage costs with a multi-billion-dollar R&D budget, most businesses should treat that as a signal to be deliberate – not cautious, but structured.

Because in AI, the real budget risk isn’t the licence cost; it’s uncontrolled success.

Want to get ahead of the cost curve?

If you’re rolling out AI tools across your organisation and trying to figure out what “controlled adoption” actually looks like in practice, it’s worth getting your licensing and rollout strategy right from the start.

We’re helping businesses think through:

• how to structure phased AI rollouts
• how to avoid uncontrolled usage blowouts
• and how to align licensing with actual business value, not guesswork

If you want to sense-check your approach (or avoid learning the hard way), feel free to reach out to the team.

Online Enquiry

Contact Us Today

Got questions, ideas, or just want to chat? We'd love to hear from you! Reach out to us anytime, and we'll get back to you with all the help and information you need.