For years, conventional startup wisdom said the real money was in software, and that hardware was a capital-intensive trap. One investor who left a top engineering leadership role at a major tech company is betting that thinking is now backwards — and that the businesses building the physical backbone of the AI era are about to become the most valuable companies on the planet.

The investor behind this firm spent over a decade running engineering and infrastructure at one of the world's largest tech companies, where he watched compute demand explode while the supply of power to run it all stayed painfully slow to build. That experience convinced him a massive supply-demand gap was coming — and that whoever solved it would have a defensible business. After leaving his corporate role, he used his own family's capital to make over twenty early-stage investments to prove that hardware-heavy startups could actually generate strong returns, despite years of investors insisting otherwise. That track record gave him the conviction to raise a proper institutional fund this year, and the firm has now backed more than two dozen companies rebuilding everything from power grids to manufacturing.

The central thesis is straightforward: as AI tools make it dramatically cheaper and faster to write software, the competitive advantage that used to live in clever code is shifting somewhere else — into the physical infrastructure that powers and supports that software. Building a chip, a battery, or a power line still takes real time, money, and engineering know-how that can't be replicated by typing a prompt. That, the investor argues, is exactly why infrastructure has become the new moat. He's quick to point out that hardware companies aren't simply software companies with bigger budgets — the failure modes, planning cycles, and testing processes are fundamentally different, and founders who treat them the same way are setting themselves up to struggle.

For roughly two decades, electricity demand in the U.S. barely moved year over year. That has flipped, with consumption now climbing several percentage points annually as data centers, electric vehicles, and factory electrification all compete for the same limited grid capacity. The problem is that building new power infrastructure — permitting a plant, laying transmission lines — takes years, while everyone racing to deploy AI wants capacity immediately. The investor pointed to long-duration battery storage as one near-term fix, describing portfolio companies building batteries that can discharge steadily for several days, giving utilities a cheaper alternative to rarely used backup gas plants. He also highlighted a company modernizing the decades-old power transformer using technology borrowed from electric vehicles, and a more experimental venture harvesting energy from ocean waves to power offshore AI computing hardware.

A recurring theme in the conversation was that almost everything that gets cheaper over time does so because it's mass produced — phones, batteries, solar panels — while custom-built, one-off equipment like grid transformers only gets more expensive every year. The investor's portfolio leans heavily into companies designing standardized hardware that can be manufactured at scale rather than custom-engineered for every project, since each doubling in manufacturing volume tends to shave meaningful cost off the next unit. He contrasted that dynamic with fossil fuels, whose extraction costs have stayed roughly flat for half a century, arguing that a technology getting cheaper every year against a flat-cost competitor is an unusually strong position to invest behind.

Having overseen the construction of millions of square feet of data centers and several generations of consumer hardware earlier in his career, the investor said the biggest lesson was that physical products fail in unpredictable, humbling ways — and that founders need to test relentlessly in the real world rather than over-planning on paper. He recalled small, almost embarrassing mistakes, like a data center loading dock built at the wrong angle, as reminders that scale magnifies every overlooked detail. That experience now shapes how he advises founders: ship something, break it, learn fast, and treat every real-world failure as cheaper to fix early than after mass production begins.