Intelligence is about to get cheap. Really cheap.
In a candid conversation following Microsoft's latest developer conference, CEO Satya Nadella painted a picture of computing's next chapter—one where artificial intelligence agents don't just assist with tasks, but fundamentally rewire how we think about software, productivity, and the very architecture of business technology.
His vision? We're witnessing the collapse of traditional software into something far more fluid: orchestrated networks of AI agents that can reason, collaborate, and execute complex workflows with minimal human intervention.
Every layer of the tech stack has to be reimagined
"Every layer of the tech stack has to be reimagined," Nadella told Matthew Berman, outlining Microsoft's massive infrastructure overhaul across 70 global data center regions. These aren't just getting AI upgrades—they're becoming what he calls "AI factories."
When Excel Becomes an IDE
Here's where things get interesting. Nadella described Microsoft 365's evolution as something he calls a "tri-modal experience"—a fresh AI-first interface, deep integration into collaborative platforms like Teams, and perhaps most intriguingly, the transformation of familiar productivity tools into development environments.
"Every Office canvas is now an IDE," he explained, drawing parallels between modern document editing and software development. Picture this: you're working in Excel, but instead of just manipulating spreadsheets, you're orchestrating AI agents that can analyze data, generate insights, and even write code—all within the familiar interface you've used for years.
"It's like having a data scientist sitting next to me while I'm analyzing a spreadsheet," Nadella said. "But it's not just Copilot—it's orchestration."
This isn't just feature creep. It's a fundamental reimagining of what productivity software can become when intelligence becomes abundant and cheap.
The Great Software Collapse
If you're building vertical SaaS right now, Nadella's next point might make you uncomfortable.
"The application layer is going to collapse down into agents," he stated matter-of-factly. Traditional software applications—those neat, bounded programs we've organized our digital lives around—are giving way to something more flexible: multi-agent systems that can dynamically configure themselves around business processes rather than forcing processes to conform to rigid software constraints.
Think about what this means for enterprise software. Instead of switching between different applications for CRM, project management, and analytics, you'd have agents that understand your business context and can fluidly move between data sources and capabilities as needed.
SaaS companies can no longer afford to be
just 'a system of record'
For SaaS companies, the message is stark: become part of the orchestration layer or risk obsolescence. "SaaS companies can no longer afford to be just 'a system of record,'" Nadella emphasized. They need to integrate into what he calls the "agentic web"—platforms that support Microsoft's Multi-Agent Control Plane (MCP) and other orchestration frameworks.
The Personal-Professional AI Divide
But here's a tension worth exploring: as AI agents become more capable and personalized, the boundary between personal and professional computing starts to blur. What happens when your personal AI assistant—the one that knows your habits, preferences, and communication style—wants to follow you to work?
Nadella's answer reveals Microsoft's strategy for maintaining enterprise security while embracing AI's personal nature. "We have to make sure these two worlds don't have data leakage," he said, pointing to tools like Edge profiles and the separation between work identities (Entra ID) and personal accounts.
More importantly, he staked out a clear position on intellectual property: agents used within an enterprise become part of that company's IP, with all the security controls that implies—identity management, conditional access, and threat protection extending to the agents themselves.
The Energy Reality Check
Now, about that elephant in the room: energy consumption. While some tech leaders dance around AI's environmental impact, Nadella addressed it head-on with characteristic directness.
"Yes, it'll double," he acknowledged, referring to energy usage. "But it needs the social permission to double."
His argument is essentially utilitarian: if AI agents can transform healthcare outcomes, democratize education, and supercharge small business productivity, then the energy cost becomes justifiable—even necessary. Microsoft is hedging this bet with investments in energy-efficient hardware, innovative cooling systems, and what Nadella calls "sustainable abundance."
It's a pragmatic stance that recognizes both the environmental reality and the transformative potential of AI at scale.
Intelligence as the New Electricity
Perhaps the most profound shift Nadella described isn't technological—it's philosophical. As the marginal cost of intelligence approaches zero, we're entering what he calls an era of "abundance."
Do we need more abundance of something like intelligence that can drive productivity and economic growth? Absolutely
"Do we need more abundance of something like intelligence that can drive productivity and economic growth? Absolutely," he declared, citing Stanford Medicine's use of multi-agent systems to streamline oncology workflows as an early example of this potential.
But abundance creates its own challenges. Nadella mused about the fundamental nature of computing itself: "Sometimes we overstate the determinism of what we call deterministic systems." As AI models become more sophisticated, even operating systems might evolve into probabilistic environments—imagine an OS that generates each frame dynamically, much like the AI-recreated version of Doom that captured the internet's imagination.
The Architecture of Everything
The killer app of the AI era, according to Nadella, isn't an app at all—it's an architecture. We're moving toward a world where agents don't just assist with computing; they are the computing layer, orchestrating resources, data, and capabilities in real-time based on intent rather than predetermined workflows.
This represents a fundamental shift from software as distinct applications to software as orchestrated intelligence. The implications stretch far beyond Microsoft's ecosystem—they suggest a future where the boundaries between productivity tools, development environments, and business process management dissolve into something more fluid and responsive.
For businesses, this means rethinking not just which tools to use, but how work itself gets structured. For individuals, it suggests a future where personal productivity scales to enterprise-level sophistication without the complexity.
We're still in what Nadella calls the "middle innings" of the AI transformation. But if his vision holds, we're not just witnessing the evolution of software—we're seeing the emergence of intelligence as the new operating system of everything.
