You’re Worried About the Wrong AI Problem

It seems everyone is thinking about artificial intelligence. Investors are putting trillions of dollars to work building data centers. Venture capital firms are writing checks for AI startups, one of the few sectors that can still secure meaningful venture funding relatively quickly. And every cohort of society is viewing the impact of AI from their unique perspective.

Young adults are worried about entry-level jobs disappearing. Mid-career professionals are being told to adapt, adopt, or get out of the way. Corporate boards, never wanting to appear caught off-guard, are seeking AI-powered productivity gains.

Last week I heard about retirees leveraging AI to build financial retirement models, design personalized workouts, and hunt down Black Friday deals on Burgundy. It’s remarkable, really. AI has become ubiquitous. Trusted? Not exactly. But ubiquitous.

Some people fret about Hollywood scenarios—AI taking their jobs, Terminator-style rogue algorithms, existential threats. But Steve Lord, someone with a genuinely unique vantage point into the world of venture capital, thinks we’ll hit a different wall first.

“We are going to have a power problem with AI long before we have a Terminator-type scenario,” Steve told me during our latest Global Macro Update interview. “We’re not going to be able to generate the power necessary to run all this.”

Steve is COO at Burkland Associates—a firm providing fractional CFO services to hundreds of VC-backed companies. That position gives him a front-row seat to the startup ecosystem. His team works with dozens of top-tier VC firms. He sees what’s coming before most people do. And what he sees is a massive bottleneck around electricity.

The data backs him up. According to Pew Research, a typical AI-focused data center consumes as much electricity as 100,000 households. The larger facilities under construction could use twenty times that amount. The International Energy Agency projects that data centers will account for nearly half (half!) of US electricity demand growth through 2030. A RAND Corporation report found that AI data centers could require 68 gigawatts of power capacity by 2027—close to the current total power capacity of California.