AI's Next Real Estate Bottleneck Isn't Space. It's Power.

For the past three years, AI has dominated commercial real estate headlines. First, it was the explosion of data centers. Then came the revival of San Francisco's office market as venture-backed AI companies rapidly expanded their footprints. Now, another trend is beginning to take shape, one that could have an even greater impact on Silicon Valley's industrial market. The biggest constraint is no longer finding space. It's finding enough power.

Etched AI recently signed an 80,000-square-foot lease in Milpitas to expand its chip research and production operations. While the transaction itself is significant, it represents something much larger than one growing company. As AI moves beyond software and into chip design, robotics, and advanced manufacturing, industrial real estate is being evaluated through a completely different lens. Companies are no longer searching for buildings that simply have enough square footage. They need facilities capable of supporting highly specialized operations that consume enormous amounts of electricity.

For decades, industrial leasing followed a familiar formula. Location, ceiling height, loading docks, office build-outs, and rent were among the biggest considerations for tenants. Electrical capacity certainly mattered, but it was rarely the factor that determined whether a deal happened. Today, that equation has changed. According to CBRE, there are currently no vacant industrial buildings anywhere in Silicon Valley with more than 4,000 amps of electrical service already installed. Many older buildings were designed with roughly half that capacity, long before anyone imagined the infrastructure demands of modern AI manufacturing.

That shortage is beginning to reshape the market in ways vacancy statistics alone cannot explain. On paper, an industrial building may appear ideal, but without sufficient electrical infrastructure it becomes unusable for many of today's fastest-growing companies. Two buildings with nearly identical locations and square footage can have vastly different values simply because one has the power capacity tenants need while the other does not. Infrastructure, something that has historically stayed in the background, is becoming one of the most valuable features of an industrial asset.

The leasing activity reflects this shift. According to Kidder Mathews, AI infrastructure and advanced manufacturing companies leased approximately 1.1 million square feet of R&D space during the second quarter, a 36% increase from the previous quarter. Even more telling is the pipeline of future demand. CBRE brokers are tracking companies collectively searching for nearly 12 million square feet of industrial space across Silicon Valley, much of it from AI-related users with significant power requirements. These aren't traditional warehouse tenants. They're companies building chips, testing hardware, and developing the physical infrastructure that will power the next generation of artificial intelligence.

Interestingly, Silicon Valley's R&D vacancy rate remains above 11%, and net absorption is still slightly negative for the year. At first glance, that might suggest supply is outpacing demand. In reality, the opposite may be true. Much of the available inventory simply wasn't designed for today's occupiers. Buildings constructed decades ago often require major electrical upgrades, utility coordination, and significant capital investment before they can support modern manufacturing operations. In today's market, vacancy doesn't necessarily mean availability.

That distinction has important implications for owners, developers, and investors. For years, landlords focused on renovating offices, improving facades, and modernizing common areas to stay competitive. Those investments still have value, but they may no longer be the biggest differentiator. Increasingly, the competitive advantage lies behind the walls. Properties with upgraded electrical service, redundant power, or the ability to quickly expand capacity will likely command stronger demand than comparable buildings without those capabilities.

This isn't just a Silicon Valley story. It's a glimpse of how AI is reshaping industrial real estate across the country. Just as e-commerce transformed logistics facilities and biotechnology drove demand for laboratory space, AI is creating a new class of industrial occupier with infrastructure requirements unlike anything the market has experienced before. Developers and investors who recognize that shift early will be better positioned as demand continues to grow.

For decades, commercial real estate has revolved around one familiar principle: location, location, location. AI isn't replacing that rule, but it is adding another. In the next cycle of industrial real estate, power may become just as important as the address itself.