For decades, power was one of the simplest variables in site selection.

A utility either had capacity or it didn't.

Today, that assumption is beginning to break down. Across the United States, utilities are warning of multi-year delays for new generation and transmission as data centers, semiconductor facilities, AI infrastructure, and advanced manufacturers compete for limited grid resources. In some markets, power availability has become the deal killer. In others, projects never make it far enough to enter the queue.

A growing number of utilities are exploring a different approach. Rather than treating capacity as a fixed constraint, they are asking whether some of that capacity can be unlocked through flexibility.

The concept takes different forms depending on the market, but the premise is straightforward: large users may be able to gain faster access to power if they are willing to alter how and when they consume it.

Sacramento Municipal Utility District (SMUD) is among the utilities attempting to put that idea into practice.

From Fixed Load to Flexible Demand

SMUD is experimenting with “flexible load” structures, where facilities—particularly data centers—are not treated as constant, always-on demand. Instead, they operate as dynamic participants. In periods of constraint, a facility can temporarily shift to on-site generation, freeing capacity for other users. Once the constraint passes, it returns to grid power.

“There are a lot of data centers that are interested in that model,” Anguay said.

The concept is gaining traction nationally. In Electric Reliability Council of Texas, large users are increasingly required to curtail demand during grid emergencies. Elsewhere, utilities are introducing tariff structures that bundle power delivery with flexibility commitments and on-site generation strategies.

But the idea is evolving beyond demand response. Raja Kadiyala, global head of digital solutions at Sidara, describes the shift more broadly as a move toward “active infrastructure,” where buildings and facilities are no longer passive consumers of power, but active participants in how it is managed.

Instead of drawing constant load, facilities can adjust demand in real time—what he describes as a form of power cycling—helping stabilize the grid while reducing the need for new infrastructure.

“The fastest infrastructure to build is infrastructure you don’t have to build,” Kadiyala said.

In practice, that could mean shifting compute workloads across time zones, ramping down noncritical processes during peak demand, or using on-site storage and generation to smooth out spikes. The underlying idea is the same: flexibility at the facility level can unlock capacity at the system level.

For site selectors, that changes the equation. Power is no longer a simple yes-or-no constraint. It becomes something that can be structured, negotiated, and, in some cases, accelerated.

Smaller Loads, Faster Timelines

At the same time, Sacramento is shaping demand to match what the system can realistically deliver.

Rather than pursuing only hyperscale users, local officials describe a strategy built around mid-sized, distributed loads—often in the 20- to 50-megawatt range—spread across multiple sites. These projects are easier to permit, easier to power, and faster to bring online.

They also align with a different type of investment. Instead of purely back-end data centers, the focus is increasingly on research and development, semiconductor design and manufacturing, and what some in the region describe as “AI factories”—facilities that combine compute, memory, and engineering functions in a single location.

This shift reflects a broader trend across the industry toward modular, distributed infrastructure. It is not just about where compute happens, but how quickly it can be deployed and adapted.

When the Utility Becomes the Partner

Underlying SMUD’s approach is a governance model that allows it to move differently than many of its peers. As a municipal, nonprofit utility, it is not accountable to shareholders and does not operate as part of a city budget. Revenues are reinvested into the system, and rates are set with long-term stability in mind.

That structure enables a closer alignment between utility strategy and economic development goals. It also allows for more direct dealmaking—around flexible load, distributed capacity, and accelerated timelines—without the same level of regulatory friction seen in other markets.

In Sacramento, power is not just a constraint to be managed. It is becoming a lever to be pulled.

The Limits of Flexibility

The model is not without constraints.

Flexible load arrangements depend on customer participation and operational discipline. Not all workloads can be interrupted without consequence, particularly as AI applications move from training to real-time inference. Excess capacity can only be stretched so far before reliability is tested. And while distributed demand may be easier to serve, it does not eliminate the need for long-term investment in generation and transmission.

These are challenges utilities across the country are now confronting. The difference is how they respond.

A Different Kind of Competitive Advantage

What distinguishes SMUD is not that it has solved the power problem. It is that it is treating power as something that can be actively managed in the context of site selection.

In a market where timelines are slipping and capacity is scarce, that distinction matters. For years, economic developers competed on rates, incentives, workforce, and shovel-ready sites. Increasingly, they may find themselves competing on something less tangible: their ability to structure access to scarce power.

The regions that win projects over the next decade may not be the ones with the largest generating fleets. They may be the ones that learn how to make existing capacity go further.