On May 26, The Verge ran a one-line headline: "Uber's president says AI spending is getting 'harder to justify.'" It would be easy to wave this off as just another executive soundbite, but the implications run deeper than that.
Uber spent $3.4 billion on research and development in 2025, up 9 percent from the year before. Then, just four months into 2026, came reports that the company had already burned through its entire annual AI budget. Inside the company, a question started growing louder: is all this spending actually creating value?
The answer that Andrew Macdonald, Uber's president and chief operating officer, gave in an interview is the heart of the story: "That connection isn't there yet. Implicitly, more may be shipping, but drawing a line from a single metric like token usage to 'we're now actually building 25 percent more useful consumer features' is very hard."
There's a reason this remark carries weight. It signals the start of round two of AI adoption — the "prove it" phase.
Does More Token Consumption Mean More Value?
Macdonald's diagnosis, restated, goes like this: token usage for AI tools like Claude Code is growing astronomically. But there's no visible, direct link between that growth in usage and any growth in the value users actually receive.
This isn't one company's private worry. Nearly every business that has seriously adopted AI tools eventually arrives at the same question. In the early days, simply "using AI" looked like an achievement in itself. Code got generated automatically, reports came together faster, design mockups arrived several at a time. That alone felt like results.
Things are different now. A year into adoption, the invoices have become unmistakable. And looking at those invoices, executives have started asking: how much did this spending grow our revenue? How much happier are our users? How much better is our product?
Surprisingly few companies can answer those questions clearly. Even a data-savvy giant like Uber struggles to.
Uber's Structural Dilemma
There's another revealing thread in Macdonald's comments. Uber CEO Dara Khosrowshahi said earlier this month: "We're offsetting increased AI investment with reduced hiring."
The implication is plain. Uber is cutting labor costs and spending that money on AI tools. Instead of hiring more people, it's buying more tokens.
For that trade to be rational, one condition has to hold: AI has to genuinely do the work people would have done — and that has to show up measurably in revenue or user experience.
One line from Macdonald captures the weight of that condition: "We have to start talking about token consumption, and the cost of that versus headcount. If you can't draw a direct line to how many useful features and how much functionality you're shipping to users, that trade becomes hard to justify."
The decision to hire fewer people is hard to undo. If a company shrinks its workforce first and only later concludes that "AI didn't deliver as expected," its operations start to wobble. Some big tech companies have already lived this: layoffs justified by AI, followed by a scramble to rehire.
The Measurement Trap
Why is measuring the ROI of AI investment so hard? There are several structural reasons.
First, efficiency gains at the front line don't carry through to final outcomes. Say one developer starts coding 30 percent faster. That doesn't translate into the company shipping new products 30 percent faster. In between sit planning, decision-making, review, deployment, marketing, and sales — and each stage is a bottleneck. If coding speeds up but everything else stays the same, final output barely changes.
Second, acceleration casts a shadow. When AI generates code faster, there's that much more code to review. When AI produces more content, there's that much more content to vet. The human workload doesn't shrink — it changes shape. And that new work is poorly measured.
Third, the baseline disappears. "How long would this have taken without AI?" is a hard question to answer, because you can't compare before and after under identical conditions. The market shifts, the product shifts, the makeup of the team shifts.
Fourth, productivity gains aren't permanent. The utility of an AI tool peaks early in adoption, and its marginal benefit declines over time. Many teams see remarkable change in the first month, then watch the cost-benefit picture grow murkier by month six.
Add these together, and AI investment becomes one of the trickiest areas in business to measure ROI on.
What "Harder to Justify" Actually Means
Macdonald's phrasing is worth a second look. He said the spending is getting "harder to justify" — not that it "can't be justified." That nuance matters.
This isn't a decision to stop investing in AI. It means accountability for that investment is tightening. Last year, the single pressure of "fall behind if you don't invest in AI" was enough to get budgets approved. From this year on, companies have to answer the question "so what did it produce?" Fail to answer, and budgets get cut, usage gets restricted, or someone has to take the fall.
What this signals is that the sales round of AI adoption is over. We've moved from deciding whether to adopt to justifying what's already been adopted. And this phase demands hard numbers, not flashy demos.
Where Small Companies Hold the Edge
This shift is a burden for large companies — but for small businesses and solo operators, it can actually create an opening.
Big companies decide on AI adoption at the enterprise level. The budgets are large, the use cases sprawl, and measuring ROI is complex. At Uber's scale, simply tracing the causal chain from token usage to user value is a massive project in itself.
Small companies are different. A solo founder pays for a Claude Code subscription, builds a feature with it, and can trace what revenue that feature generates — all inside one person's head. The unit of measurement is small, so the causality stays clear.
"Did revenue go up after a month with this tool, or didn't it?" A small company can answer that within a month. A big company often can't answer the same question even on a quarterly or annual basis.
That asymmetry is the small company's unexpected strength. The harder ROI becomes to measure in the AI era, the more the advantage shifts to organizations that can run small, fast experiments and check the results.
A Practical Way to Measure ROI
For anyone trying to measure the ROI of AI in their own business or team, here's a practical approach.
1. Keep the unit of measurement small. Measure "the effect of AI on this one workflow," not "the total effect of our AI investment." One customer-support automation, one type of content generation, one stage of code review. The narrower the unit, the clearer the effect.
2. Don't count only the direct costs. Beyond subscription fees, include learning time, system integration costs, the extra hours spent on review, and recovery costs when something goes wrong. Only the sum of all of it is the real cost.
3. Measure value as changes in user behavior. "A 30 percent internal efficiency gain" has no value on its own. What matters is whether it led users to get answers faster, come back more often, or buy more.
4. Reassess every three months. The utility of AI tools changes over time. A tool that looked impressive at first can deliver less per dollar six months in. Don't adopt once and use forever — build the habit of periodically re-examining the value.
5. Consciously ask, "What if we did this without AI?" Throwing AI at every task isn't the answer. Some work is faster and more accurate when a person does it directly. Deliberately separating the places where AI belongs from the places where it doesn't is the most direct way to raise ROI.
The Second Round Begins
Here's the AI market landscape right now, in brief.
Big tech is pouring enormous capital into infrastructure. Nvidia has spent $40 billion on AI equity investments this year alone, and OpenAI and Anthropic have partnered with private equity firms to build consulting organizations worth billions of dollars. The companies making the tools are still on offense.
The companies using the tools are another story. Even a giant like Uber has started asking whether these tools genuinely deliver value. Put more precisely: the sales pitch is over, and the justification phase has begun.
The most interesting place to watch is where these two currents collide. Tool companies push "use more"; user companies ask "is it worth what we're spending?" That tension will shape pricing, packaging, and differentiation strategy across the AI market for the next several quarters.
The first round of AI adoption ran on excitement. The fear of falling behind was the most powerful marketing there was. That round is now winding down.
The second round is about verification. "We tried it — so what happened?" becomes the central question. The tools that survive this round will be the ones that can prove clear value, and the companies that survive will be the ones that can measure that value inside their own business.
Uber's AI budget, gone in four months, is the outcome of round one. And Macdonald's "harder to justify" is the starting gun for round two.
It's a good moment to check which of these two rounds your own business or team is standing in. Still swept up in the excitement, or already moved on to calm, deliberate verification? Whether you can answer that question honestly will determine your competitiveness in the next stage of the AI era.
The more powerful the tools become, the more the judgment of the person deciding where to use them matters. The question Uber is asking right now is, in the end, a question every business owner will face soon enough.
