Microsoft’s bet: What’s left when AI models are commodities?

This article summarizes an episode of Dwarkesh Patel’s video series featuring Satya Nadella, CEO of Microsoft.

Satya Nadella, CEO of Microsoft / Photo credit: Microsoft

Microsoft CEO Satya Nadella sees the race to build AI systems differently. He believes the opportunity is not just serving a few large AI labs but playing a long game. This plan challenges the industry’s costly race.

A planned slowdown in building data centers

Spending billions on hardware built for today’s AI could be a mistake if the next version of AI needs a different design.

Building for an unknown future
Nadella states, “I didn’t want to get stuck with a massive scale of one generation… The pacing matters; the fungibility and the location matter; the workload diversity matters; customer diversity matters. And that’s what we’re building towards.”

Avoiding reliance on a few customers
The goal is to build a system that serves thousands of companies with different AI needs, not just a few companies making the newest AI.

Nadella notes, “What we have said is that we’re in the hyperscale business, which is at the end of the day a long tail business for AI workloads… we’re not in the business of just doing five contracts with five customers being their bare-metal service.”

Using money wisely is the new advantage

This decision not to spend too much created a new focus. Microsoft is now using software to improve efficiency and get the most from its large investments.

Turning money into an advantage
The AI boom has made cloud computing a very expensive industry. This change could hurt the profits of a large software company.

Nadella argues, “We are now a capital-intensive business and a knowledge-intensive business. In fact, we have to use our knowledge to increase the ROIC [return on invested capital] on the capital spend.”

The software advantage
This focus on efficiency shows a clear difference between a modern cloud company and an older data center company.

“Some people ask me, what is the difference between a classic old-time hoster and a hyperscaler? Software. Yes, it is capital intensive, but as long as you have systems know-how, software capability to optimize by workload, by fleet… That’s why when we say fungibility, there’s so much software in it.”

AI models risk becoming a commodity

This plan for spending money is tied to a specific view of the market. Nadella believes the companies that create AI models might not have the power in the end.

The problem for AI model makers
While the world is focused on the race to build the most powerful new AI model, Nadella questions if that is where the long-term value will be.

Microsoft’s future is building systems for AI agents

Global trust is key