Microsoft CEO Satya Nadella has posed a deceptively simple question that could reshape how companies think about artificial intelligence: if you swapped out your AI model tomorrow, would your business still deliver value? His answer reveals a fundamental shift in where competitive advantage actually lies in the age of AI.

What Does Nadella's "Model Swap" Test Actually Mean?

At first glance, Nadella's test sounds obvious. If removing an AI model means you lose all your value, then the model was doing your job, not enhancing it. But his insight goes deeper. As frontier AI capabilities become commoditized, with every major cloud provider eventually accessing similar computing power and open-source models closing the gap, the real defensible advantage shifts away from the model itself.

Nadella argues that companies should build AI systems where a generalist model can be swapped out without losing the unique "company veteran" expertise embedded in their learning systems. This reframes the entire competitive landscape. If the model is truly replaceable, then by definition it's a commodity input, however capable it might be.

The deeper structural shift is this: as AI drives the cost of execution toward zero, the binding constraint stops being raw intelligence and becomes verification. Tasks with measurable inputs and outputs get cheaper. But in many critical domains, checking the work still requires scarce experience, long feedback loops, and someone willing to stand behind the result.

Where Is the Real Intellectual Property in AI?

Nadella identifies what he calls "verification infrastructure" as the true crown jewel. This encompasses everything a firm has measured and can measure, plus everything its top experts have learned. It's the accumulated experience that shapes how they verify, judge, curate, and apply judgment on the job.

He describes this as a "hill climbing machine" that compounds over time. Every improved workflow generates better training signals, which accelerates the accumulation of tacit knowledge unique to the firm. The companies that build this early will have an advantage that is hard to replicate, regardless of any new individual model capability.

This creates what researchers call "verification-grade network effects." Every dispute resolved, fraud caught, and expert correction becomes reusable precedent. A settled case lets the firm safely automate the next case faster. That asset deepens with use and cannot be manufactured with compute alone, because it is earned one verified outcome at a time.

How Should Companies Build This Competitive Moat?

• Private Evaluation Systems: Rather than relying on public leaderboards, firms should build private evals and private reinforcement learning environments where performance is benchmarked against outcomes the business actually cares about.
• Capture Failed Experiments: Some of a firm's most valuable records are its failed experiments, errors, and missteps. Rejected output, captured correctly, becomes the best training signal for institutional judgment.
• Codify Expert Judgment: Encode expert judgment as reusable ground truth so the next AI agent can apply it without the expert in the loop, freeing the scarcest resource the firm has and widening the share of work it can trust at scale.

What's Happening at Microsoft Right Now?

While Nadella articulates this vision, Microsoft itself is facing significant challenges in executing it. The company is confronting multiple AI-related headwinds that complicate its path forward.

A class action lawsuit filed by the City of St. Clair Shores Police and Fire Retirement System alleges that Microsoft executives, including Nadella, made "materially false and/or misleading" statements about adoption of the company's Copilot technology. The complaint claims that "Microsoft's flagship proprietary AI model ranked well below competitors on a number of benchmark tests" and that "Microsoft had failed to convert a significant percentage of its commercial Microsoft 365 users to paid Copilot subscriptions".

The numbers underscore the challenge. When Microsoft announced its fiscal second quarter results on January 28, the company disclosed that paid Microsoft 365 seats had reached only 15 million out of 450 million total Microsoft 365 users. The company's shares subsequently declined by more than $48 per share, roughly 10 percent of their value at the time.

"We are aware of the complaint and believe the claims are without merit. Microsoft stands by the integrity of its public statements and will vigorously defend itself in court," stated a Microsoft spokesperson.

Microsoft Spokesperson

Why Is Microsoft Turning to Amazon for Help?

Beyond the lawsuit, Microsoft faces a more immediate infrastructure crisis. GitHub, the coding platform Microsoft acquired in 2018, is buckling under the weight of AI-driven demand. The platform is now turning to Amazon Web Services (AWS), Microsoft's biggest cloud rival, to help address capacity issues.

The scale of the problem is staggering. GitHub commits, which serve as a proxy for development activity, were on pace to reach 14 billion in 2026, up from 1 billion in 2025. This explosion reflects how AI coding tools have made it easier for developers to write more software, flooding GitHub with new code and straining its compute resources.

Microsoft had originally planned to move GitHub entirely to its Azure cloud service by 2027. Instead, the company is now adding extra computing capacity via AWS. A Microsoft spokesperson confirmed that GitHub is tapping multiple cloud providers but declined to comment on specific Amazon involvement.

"The incredible spike in agentic development that began late last year has tested our infrastructure's limits. To meet this demand, we are both accelerating our move to Azure and continuing to explore a multi-cloud strategy to ensure we have the future capacity, compute elasticity and horizontal scale required to support continued growth," stated a GitHub spokesperson.

GitHub Spokesperson

This move is notable because Microsoft is actively competing with AWS for cloud market share. Giving an arch rival more business, rather than addressing GitHub's needs through its own Azure service, underscores how severe the capacity crunch has become.

GitHub has suffered dozens of major outages in 2026, prompting criticism from developers. Mitchell Hashimoto, cofounder of startup HashiCorp, wrote in April that GitHub was "no longer a place for serious work if it just blocks you out for hours per day, every day".

What Does This Mean for Nadella's Vision?

The gap between Nadella's strategic vision and Microsoft's operational reality is striking. He argues that companies should build proprietary verification infrastructure and private evaluation systems that make them resilient to model changes. Yet Microsoft is struggling to deliver reliable infrastructure for one of its most important developer platforms.

Nadella's test assumes companies have the luxury of focusing on long-term competitive moats. But when your flagship AI products underperform competitors, your conversion rates disappoint investors, and your infrastructure can't handle the demand you're trying to create, the immediate crisis takes precedence over architectural elegance.

The irony is that GitHub itself could become a case study in Nadella's framework. If Microsoft can build the right verification infrastructure around developer workflows, it could create exactly the kind of defensible advantage he describes. But first, it needs to solve the capacity problem, even if that means relying on AWS to do it.