The scale of AI deployment and investment over the past two years has been astonishing. Despite concerns of a bubble and environmental questions, the potential for further innovation and economic growth from AI remains outstanding as we enter 2026. However, one issue presents an immediate and clear threat to AI growth: energy. Without a radical transformation in capacity, the overwhelming energy needs created by modern artificial intelligence could potentially slow its so far revolutionary economic advancement.
Hollow-Core Fiber and the Race to Power the AI Economy
Of course, the urgency of the problem is not being ignored. “It’s the gold rush, and I’m selling picks, shovels, and pipes,” said Jason Eichenholz, CEO of Relativity Networks. I sat down with him at Fortune Brainstorm AI this Monday, hoping to learn more about how AI infrastructure needs will be met. He’s confident. “They will figure it out because they have to, because of the level of investment. In the next five years, there will be a massive transformation in energy infrastructure and capacity.”
Eichenholz’s own company aims to play a crucial role in solving the energy problem. Relativity Networks is developing hollow-core fiber, a new type of fiber-optic cable where light travels through air instead of solid glass, allowing data to move much faster. For reference, conventional fiber-optic cable allows data to travel at two-thirds the speed of light, while hollow core is nearly at light speed. The resulting reduction in latency (about 33% over the same distance) can be significant for work such as high-frequency trading and telemedicine, which relies on rapid data exchange. For infrastructure build-out, this means that HCF can help increase geographic availability for potential data centers. “They are building data centers the size of Manhattan. Obviously, you cannot build those in Manhattan,” Eichenholz said. How much more land area could HCF potentially make viable? “Double.”
One of the issues, Eichenholz says, is a common refrain across development projects in or near major urban centers. “NIMBY (not in my backyard) is a big problem. And in some ways, it’s legitimate.” A proposed 290-acre data center near Tucson, called Project Blue, faced large-scale community pushback for what was seen as gratuitous energy and water usage in one of the driest parts of the country. Greater land flexibility may help avoid such conflicts. It could also create opportunities for projects to be built closer to sustainable energy sources, a benefit Eichenholz highlighted when asked about the relationship between HCF and environmental concerns surrounding AI.
The Sustainability Challenge Ahead
The energy crunch comes at a time when fossil-fuel usage must be reduced globally. Developing abundant green energy resources will be necessary to accommodate the rise of artificial intelligence, or else risk setting climate-related energy-use goals back substantially. Innovations, like HCF, will be necessary to help mitigate tradeoffs. Whether in water efficiency, renewables, infrastructure, or manufacturing, it will take a seismic effort to meet the energy needs of the next wave of AI build-out.
“I think of it like the space race, like with Velcro,” Eichenholz reflected. “You have a massive market incentive to create something, and the innovation follows. And not always just for where it’s intended.” There’s a hope that perhaps the massive external need for energy will spur investment and invention in green energy, but this, of course, remains to be seen.
I asked Eichenholz pointedly what the consequences would be if, in the next five years, the energy capacity didn’t materialize. He was adamant. “It’s 6.66 billion dollars of revenue per gigawatt. They will figure it out. But it will be a massive shift in power infrastructure.”
