'AI will be...': Satya Nadella's candid confession on AI’s next gold rush, hyperscalers, and the battle for dominance

Hyperscalers will be the next big thing in creating value in artificial intelligence, Microsoft CEO Satya Nadella said in a podcast with Dwarkesh Patel, a San Francisco-based Indian-origin AI podcaster, on Wednesday.

“Because the fundamental thing is if intelligence is a log of compute, whoever can do lots of compute is a big winner. Everybody’s excited (not) about what’s happening on the GPU side (of models like ChatGPT), but the ratio of AI-accelerated storage to compute at scale. AI workloads are hungrier for more compute, not just for training but also for test time. So, (agents are) going to create massive demand and scale for compute infrastructure,” Nadella said in the podcast.

Hyperscalers are massive cloud service providers that offer computing and storage services at an enterprise scale.

However, Nadella argued that in the value-creation phase of hyperscalers, within the domain of cloud computing and AI models, there wouldn’t be a winner-takes-all scenario.

“After that, it becomes a little fuzzy because you could say there is a winner-takes-all model, but I just don’t see it. The buyers will not tolerate a winner-takes-all. Hyperscalers will never be a winner-takes-all because buyers are smart consumers,” Nadella said.

Nadella recalled the time when Microsoft entered the cloud computing market with Azure, at a time when Amazon Web Services (AWS) already had a strong lead. Many assumed AWS would dominate entirely. However, based on past experience competing with Oracle and IBM, Nadella knew enterprise buyers prefer multiple suppliers—they don’t accept a single dominant player. This is because corporations and IT departments seek alternatives for reliability, cost competition, and flexibility.

“Markets sometimes can be winner-takes-all, but the buyer is a corporation and an enterprise. They will want multiple suppliers. And so, you’ve got to be one of the multiple suppliers,” Nadella said.

Applying this to AI models, Nadella argued that while there may be dominant proprietary (closed-source) models, open-source alternatives will always emerge. Just as Windows coexisted with open-source operating systems, the AI model space will also have both proprietary and open-source options. The presence of open-source models ensures that no single closed-source model can achieve total dominance.

Additionally, Nadella said that during this AI boom, private players should not underestimate the role of governments.

“By the way, if this thing (AI and hyperscalers) is really as powerful as people make it out to be, let’s not discount the state. (The state) is not going to sit around and wait for private companies to go around and take over the world. So, I don’t see it as a winner-takes-all,” Nadella added.

However, he argued that ChatGPT is an example of an at-scale consumer property that has already gained real escape velocity. But what happens at the consumer level, he believes, may not necessarily happen at the enterprise level, paving the way for multiple winners.