Silicon Valley Discovers Electricity Requires Planning

For eighteen glorious months, venture capital has operated under a assumption so foundational it barely needed stating: electricity comes from the wall, infinity is optional, and permits are merely suggestions issued by people who don't understand exponential growth. Goldman Sachs has now gently informed the industry that this business model has encountered what engineers call 'a constraint' and what the rest of us call 'how the world actually works.' The U.S. electrical grid, it turns out, does not scale on startup timelines.

The collision is almost poetically stupid in its predictability. Silicon Valley's move-fast-break-things ethos has spent the better part of a decade breaking nothing of consequence—mostly just user agreements and product roadmaps—but has now crashed directly into something that cannot be disrupted: the physics of power distribution and the bureaucratic glacialism of permitting infrastructure. When your bottleneck is measured in kilowatts and your solution requires Environmental Impact Statements, no amount of deck-polishing or Series funding changes the fundamental timeline. The grid moves slowly because it has to; because when it breaks, actual people in actual homes lose actual electricity, not just access to a social feature.

This should have been obvious. The industry has known for months—perhaps years—that training large language models consumes staggering amounts of electricity, that data center buildout is energy-intensive, and that the continental power infrastructure was designed for a different era. Yet somehow, the dominant operating assumption has remained that capital deployment and technical ambition could outrun the laws of thermodynamics. They could not. They cannot. They will not.

The real comedy lies in what happens next. Venture firms will now announce new 'energy infrastructure' funds. There will be panels on 'solving the power crisis.' Startups will emerge claiming to have invented microgrids or fusion or battery technology that will definitely, absolutely, certainly be commercially viable in seven to ten years. None of this changes the fact that Goldman Sachs has just warned that energy bottlenecks are set to stall the entire AI boom—which is to say, the entire justification for recent venture returns.

What makes this particularly rich is the symmetry of the failure. The industry built its entire thesis on the ability to move faster than incumbent systems, to sidestep regulation, to operate in the margins while legacy players lumbered. But electricity is not a margin. It is the operating system. And when your strategy requires permits from the government, suddenly you are no longer disrupting—you are waiting.

The broader message is clarifying: the venture economy has spent a decade solving for capital and software while systematically ignoring the physical world. The grid cannot be disrupted because it is not a bug; it is infrastructure. And infrastructure, by definition, is the opposite of a startup—it is expensive, regulatory-bound, and measured in decades instead of quarters.

So Silicon Valley is about to discover what the rest of the economy has always known: sometimes the bottleneck is not a problem to be solved, but a reality to be accepted.