Investors Stopped Counting GPUs. Here's What Counts Right Now.

"The simpler the ROI to explain, the realer it is."

That's Dave Easton, Growth Equity Partner at Generation Investment Management, the sustainability-focused firm behind checks into AlayaCare, Octopus Energy, and WEKA. On this latest episode of Deep Geeks, he tells host Serena Huang what he's actually underwriting when he evaluates an AI company. Not tokens shipped, not GPUs racked, but return on intelligence, how much value a company gets out of every unit of compute and power it burns.

The Octopus Energy test

Dave points to Octopus Energy, the UK electricity retailer that's been running AI through its Kraken platform for years. Its two North Star metrics have never changed. Does it improve the efficiency of the team, and does it improve the happiness of the customers. If a company can prove both, Dave says, that's ROI. No dashboard of activity metrics required.

His broader point. If explaining the value takes "a complicated logic chain," it's probably not going to land with customers, and it's probably not real.

Watch: Dave on the Octopus Energy test

Why Dave asks for the chief people officer

Dave says he routinely asks to speak with a portfolio company's chief people officer during diligence, and that it's "amazing how rarely investors do that." He's been told more than once, "You're the first person that's wanted to speak to our chief people officer."

What he's looking for is a company with a winning culture and a genuinely kind, mission-oriented one at the same time. In his words, companies "trying to have an impact in the world" don't need a lower bar. If anything, they need a higher one, because the stakes of what they're building are bigger.

Watch: Why diligence starts with the chief people officer

Intelligence per kilowatt hour

Dave came to AI through the energy sector, six or seven years ago, watching cluster sizes grow for autonomous vehicles and computer vision and asking what that would do to power demand. That question became Generation's "green data" thesis, and it's part of why the firm backed WEKA. As Dave puts it, GPUs burning power without enough data fed in fast enough are wasted spend.

The metric he says every AI company should be tracking is intelligence per kilowatt hour, how much intelligence you're getting out of the power you're pulling off the grid. He cites one WEKA customer that took a training run from two weeks down to four hours as the kind of result that should be showing up on a board scorecard.

His reasoning, in his own words. "The demand for intelligence seems to be unlimited. The supply of kilowatts is very much limited."

Watch: The two-week to four-hour training run, and intelligence per kilowatt hour

Two metrics that matter

Asked to pick one metric that separates real AI value from hype, Dave asked for two. Customer satisfaction and intelligence per kilowatt hour. In his words, "with the resources that I have, am I doing the maximum I possibly can to get the most intelligence out of the fewest resources."

Dave’s advice to founders

Asked what founders should get right before walking into a meeting with an investor, Easton's answer was about purpose, not metrics. "Why does your company exist, and what's its higher purpose and higher mission." He was direct that this isn't in tension with making money, companies "have to make money," but said the mission is what will "inspire your team" and "inspire your customers" when things get hard, which he said they will.

Watch the full episode

Growth investor Dave Easton on what he's actually looking for in an AI company, and why he thinks intelligence per kilowatt hour belongs on every board scorecard.

Watch now: “Why Investors Today Count on Kilowatts”