Yotta 2026: The Race for Power, Infrastructure and Public Trust

A conference that brings together 6500 attendees to discuss, explore, and conduct the business of one of the largest infrastructure investment cycles in history is the perfect opportunity to reflect on the physical, energy, and financial infrastructure needed to bring the future of AI to life.

What was made clear at the conference is that the next frontier of AI is physical infrastructure. Prior conventional wisdom took for granted that compute and computing were the main constraints limiting the potential of AI. However, it is now apparent that the continued growth of the AI economy rests on the industry’s ability to deliver electricity to data centers, partner with utilities and government to build the infrastructure to connect and manage these massive loads, raise or borrow the capital to finance potentially trillions of dollars of capex, and engage communities and rally their support for the data center buildout. In other words, AI may be powered by data and algorithms, but it will live and die by electrons, engineering, and communications.

If there was one ubiquitous slogan that permeated the conference, it was “speed to power.” In crude terms, there were two types of companies at Yotta, those that are looking for ways to get power more expeditiously (hyperscalers, data center developers, neoclouds) and those with solutions to make that a reality. Unfortunately, utility infrastructure is not currently mature enough to meet this moment in history. As such, developers have to figure out ways to access power with behind-the-meter systems. Increasingly, natural gas is emerging as the major near-term solution. However, renewables, batteries, and microgrids are a major piece of the speed-to-power puzzle.

Another solution to the speed-to-power challenge is less well-known but powerful and technologically feasible. Flexible power surfaced throughout Yotta conversations because it leverages existing infrastructure to unlock significant capacity. Flexible solutions include AI workloads that can be shifted or temporarily suspended, technologies that can intelligently extract more from existing transmission infrastructure, and smart software that, by coordinating electricity, cooling, and computing demand, drives efficiency across the entire ecosystem. In addition, I heard a lot about grid-interactive data centers that, instead of draining the grid of load, can actually be an asset to the grid, unlocking potential economic benefits for utility customers. To be fair, that flexibility is here and ready to help solve the speed-to-power challenge was hardly unanimous. But the fact that it was a large part of the conference program speaks to its potential.

My last observation about Yotta 2026 was that it featured what might be the most vexing bottleneck confronting the data center industry, namely its ability to gain social license and win the approval of communities to build its centers. Of course, the data opposition movement has emerged as the hottest topic of conversation industry-wide; but what made its prominence at Yotta surprising was that it is a conference focused on AI, energy and infrastructure, not community relations. As someone who has been counseling the data center industry on communications for years, I know that storytelling is often lost on engineers, who made up the majority of conference attendees. Nonetheless, the community relations hornet's nest emerged as a topic at conference sessions, networking receptions and informal conversations. Its prominence is a strong signal that the industry has learned that the best hardware, software and energy generation solutions will mean nothing if it can't win the hearts, minds and pocketbooks of communities.