Bill Maris: How Google Could Crush AI Competitors, Why Small Funds Win, and AI's Atari Stage

Bill Maris, founding CEO of Google Ventures, had said he was out of investing. He is back with Section 32 and a $150 million first fund, and the size is the argument: a small fund lets him "be very selective in the companies that I invest in, the people that I hire." He also fixes his objective function up front — the fund will "invest for a financial return," because "any other metric is impossible to measure and therefore won't succeed." That premise is what later licenses his attack on companies that wrap themselves in public-benefit language while keeping the value creation private.

The scale claim attached to this is unhedged: AI means "the world is gonna change by orders of magnitude." It is a forecast, not a finding — the measured results arrive later and are considerably narrower.

Maris establishes himself as a builder before an investor: Section 32, then Google Ventures, then Google's VP of special projects, where he says he incubated Waymo, Google X and Calico, and before that a web hosting and data center company. The origin scene is a neuroscience graduate miserable on Wall Street until he looks into an office closet and finds a server under the jackets: "that's where our email and websites live."

The insight is arithmetical, not technical — if a website can live in a closet, how many can live in his? He quits, having "glimpsed through a keyhole." The comedy is the gap between vision and reality: the state-of-the-art 1997 data center was three servers, "a small, medium, and large," run on credit cards from his Vermont apartment, where opening the windows to cool the machines made the room cold enough that "by noon, if you had a glass of water in your desk, it would ice over."

A thunderstorm, a leaking roof, a landlord who says "well, that happens sometimes." Maris buys tar and a mop, climbs up in the lightning and seals the roof from the wrong corner: "I tarred myself into a corner. But the choice that I faced was either the servers get electrocuted or perhaps I get electrocuted." The lesson is lesson two — "to see the future, sometimes you need to be a little bit insane" — and its precise form is about perception: "It may appear to those around you that you are tarring the roof in a thunderstorm."

That is the weakest link in the talk. It asserts that correct foresight looks reckless from outside, but supplies no way to distinguish it from actual recklessness. Every failed founder also looked like a man on a roof in a storm.