AI is already causing a resource crunch. But we may only be at the beginning, one executive says.
Kevin Cochrane, chief marketing officer of neocloud Vultr, told The Deep View that AI demand has only hit the "tip of the spear." Because most of AI demand is coming from hyperscalers, developers and AI-native startups, several massive companies are still very nascent in their AI journey.
For every Fortune 500 company, there is likely a portfolio of thousands of applications for both internal and external uses, according to Cochrane. "Every single one of those applications needs to get modernized," he said. "That's a 10-year build cycle, and that's going to demand a tremendous amount of compute capacity all around the globe." It's no secret that enterprise could represent a cash cow for AI. That's why OpenAI and Anthropic are vying to serve major organizations, investing in forward-deployed engineers and signing partnerships with legacy tech organizations such as IBM to make that transition smoother.
And even beyond enterprises, Cochrane asserted, AI presents the opportunity for "unbounded demand," essentially only limited by humanity's capacity for creativity. "I think we're going to get to a point where we're going to see that there is a predictable pace of growth of demand," he said.
However, once that deployment comes to fruition, the industry runs into another problem: Compute capacity and energy constraints. There is a "theoretical limit" to how much power you can and should generate, or how much land and water you can and should use, Cochrane said.
With AI already impacting infrastructure, "something has to give," he said. The thing that gives may be the chips themselves, something that companies like Nvidia and AMD apparently realize as they shift focus to efficiency.
"Given the fact that there has to be some natural limit, and there's infinite demand … You have to be as efficient as possible," said Cochrane. "You have to drive down that efficiency curve year after year after year."
Cochrane isn't the only one reading the tea leaves on how AI will affect infrastructure. Investors have taken an interest in the technologies at the foundation of AI, such as energy and mining, seeking to get ahead of the problems that AI could create.
Our Deeper View #
Cochrane is striking a common theme in saying AI will drive fundamental changes in work and life. Fundamental changes, however, require innovation from all sides. Chips and data center architecture are just one piece of the puzzle in making this technology more efficient. At the utility level, more sources of clean energy could ease the burden that AI and data centers are placing on overstressed energy grids. And the models themselves could get more efficient, too, such as Pathway's 150-million-parameter small reasoning model, BDH-CQ, which costs a fraction of OpenAI's most cost-effective model, GPT 5.6 Luna and demands less compute and energy. Efficiency is a core part of innovation. Computers used to take up entire rooms, and mobile phones used to be giant bricks as long as your forearm. AI's version of this means trimming down every component of the stack, from energy to infrastructure to the models themselves.