Microsoft is exploring a fundamental shift in how it prices its software, moving away from traditional "per user" licensing toward a model based on "per agent." CEO Satya Nadella revealed this strategic pivot during an episode of the "Dwarkesh Podcast," released on Thursday, as the company rethinks its business model in response to the rise of artificial intelligence. Nadella explained that Microsoft is transitioning from being an end-user tools company to an infrastructure provider for AI agents—autonomous systems capable of performing tasks and making decisions independently. "Our business, which today is an end-user tools business, will become essentially an infrastructure business in support of agents doing work," he said. He emphasized that the concept of "per user" should now be redefined as "per agent," reflecting the growing role of AI in daily workflows. This shift mirrors a broader industry transformation. As AI systems increasingly use software to execute tasks rather than simply being powered by it, companies are struggling to maintain flat, seat-based pricing models. With each AI inference or query consuming significant compute resources, traditional licensing is becoming unsustainable. ServiceNow CEO Bill McDermott previously told Business Insider that when usage exceeds what companies can afford under fixed pricing, they need metered billing. This trend is already visible in enterprise solutions: Deloitte and EY have launched agentic AI platforms that allow AI systems to generate reports, schedule tasks, and interact with clients autonomously. Instead of charging for human labor hours, these firms are moving toward a "service-as-a-software" model, where clients pay for outcomes delivered by AI agents. Nadella noted that Microsoft’s existing infrastructure—such as the tools within Microsoft 365—will remain essential, even for AI agents. "You need some place to store it, some place to do archival, some place to do discovery, some place to manage all of these activities, even if you're an AI agent," he said. He believes the demand for this underlying infrastructure will grow faster than the number of human users. Microsoft has already begun testing this new approach. Earlier this year, it introduced a pay-as-you-go pricing model for its AI agents, layered on top of the free Copilot chat experience available to Microsoft 365 users. This allows organizations to pay based on the actual work performed by their AI agents. Other tech giants are following similar paths. Anthropic charges customers based on the number of tokens processed by its Claude models, while Google uses a usage-based structure for its Gemini API. As AI becomes more embedded in enterprise operations, the shift from per-user to per-agent pricing is emerging not just as a possibility, but as a necessity for sustainable growth.