Microsoft is fundamentally rethinking how it measures AI success, moving away from traditional software licensing metrics toward a model based on how intensively customers actually use AI tools. CEO Satya Nadella made this shift explicit during the company's latest earnings call, signaling a major strategic pivot as Microsoft grapples with the reality that artificial intelligence could cannibalize its traditional software business.

The numbers tell part of the story. Microsoft reported $82.89 billion in revenue for its third fiscal quarter, with Azure cloud services growing 40 percent year over year. Microsoft 365 Copilot, the company's AI assistant suite, now has more than 20 million paying users, up from 15 million just four months earlier. But here's what's notable: Microsoft isn't breaking down how much of Azure's growth actually comes from AI workloads, how profitable Copilot is on its own, or what direct revenue OpenAI generates as a customer.

That opacity hints at a deeper concern. If companies need fewer employees to accomplish the same work thanks to AI, they'll need fewer software licenses. This threatens the per-seat licensing model that has powered Microsoft's profitability for decades. Nadella's answer is to abandon that model entirely.

Why Is Microsoft Shifting Away From Per-Seat Licensing?

The traditional software business assumes a simple equation: more employees equals more licenses. But AI disrupts that math. A smaller team augmented with AI agents might accomplish what previously required a much larger workforce. If Microsoft continues charging by the seat, it loses revenue as headcounts shrink. If it charges by usage instead, it can capture value from the productivity gains AI enables.

Nadella explained the strategy directly: "Any per-user business of ours, whether it's productivity, coding, security, will become a per-user and usage business." Microsoft has already moved GitHub Copilot, its AI coding assistant, to this license-and-usage model. The company is essentially betting that companies with smaller teams using more AI will generate the same or higher revenue than larger teams using traditional software.

Nadella

This isn't just theoretical. Microsoft is backing the bet with massive capital expenditure. The company plans to invest $190 billion in infrastructure during 2026, well above what Wall Street analysts expected. Even with that buildout, Nadella warned that computing capacity will remain tight through the end of 2026.

How to Evaluate AI's Real Business Impact?

• Measure Local Productivity: Track how quickly individual workers complete specific tasks when using AI tools, which is relatively straightforward to quantify through time-tracking and task completion metrics.
• Connect to Company Outcomes: Determine whether those individual productivity gains translate into measurable improvements in overall business results, such as revenue per employee or project completion rates, which is significantly harder to isolate.
• Calculate Usage-Based ROI: Compare the cost of AI usage fees against the value generated, ensuring that AI spending doesn't exceed the cost of hiring additional full-time employees to accomplish the same work.

The challenge Nadella faces is real. Measuring whether AI-driven productivity actually shows up in a company's bottom line is genuinely difficult. Local effects, like how fast a single worker moves through tasks, are easy to track. Whether and how those gains compound across an entire organization is a much tougher question.

"At the end of the day, it's going to come from some eval and outcome that a business has, where these agents that are working on behalf of users or with users has created value," said Satya Nadella.

Satya Nadella, CEO at Microsoft

Nadella sees usage intensity as the key metric. "It's more about getting intense users and intense usage, and that's what we're focused on," he stated. In other words, Microsoft doesn't care if you have 100 Copilot licenses sitting idle. It cares if you have 20 users who rely on Copilot multiple times per day to accomplish critical work.

What Does This Mean for Enterprise AI Adoption?

The shift has immediate implications for how companies should think about AI investments. If Microsoft is moving to usage-based pricing, other enterprise software vendors will likely follow. This means companies can't just buy AI licenses and hope for the best; they need to demonstrate concrete value or face escalating bills.

The pressure to prove AI's dollar value will only intensify. If AI spending soon matches the cost of several full-time hires at midsize companies, executives will demand clear evidence that the AI investment is actually delivering productivity gains. Nadella's framework suggests that evidence will come from measuring how deeply and frequently employees use AI tools, not from theoretical benchmarks or vendor promises.

Microsoft's earnings report also revealed a softer-than-expected revenue and margin forecast for the next quarter, causing the stock to drop more than 5 percent. This suggests that investors are still uncertain about how AI will ultimately impact Microsoft's profitability. The company's refusal to break out AI-specific revenue numbers only deepens that uncertainty. But Nadella's strategic pivot makes clear that Microsoft is betting on a future where AI usage, not software seats, drives the company's growth.