While Northern Virginia remains the country's data center HQ, hyperscalers are rapidly moving inland to Texas and to states all across the Midwest (Wisconsin, Michigan, Missouri, and more). The reasons for this hardly need to be spelled out: energy demand has skyrocketed in tandem with widespread AI adoption, and existing infrastructure lacks the capacity to keep up.

As demonstrated by countless heated town halls across the country, this migration has hardly been uncontroversial. Hyperscalers' inland migration has increasingly been met with public hostility, leading to complex regulations and, in some cases, outright moratoriums on build-outs. Many of these data centers will indeed eventually come online, but what these experiences have made clear is the need for additional solutions in the hunt for increased capacity.

We're entering the third phase of the great data center build-out. The first concentrated capacity in the metro network hubs where users live; the second has been a land-and-power grab across the heartland. The third will be more thoughtful and deliberate, foregrounding things like local grid stability, water stewardship, and active community engagement. And it will be enabled, in large part, through new, more optimized methods of marshalling the energy that these data centers need to keep up with skyrocketing demand.

Building out moves inland

It's not a coincidence that, until recently, most data centers were built in major metropolitan areas: in addition to already hosting established network infrastructure, these places also happen to be where the end-users of this technology live – both individuals and enterprises.

Nor is it surprising that a fervent rush for market dominance has sent the major AI companies into the heartland, where land is plentiful and far less expensive.

What is surprising – what the hyperscalers could not have predicted – is the intensity of the backlash to these constructions, which has escalated rapidly over the past year. Almost the only thing vast swathes of Americans seem able to agree on, in these polarized times, is that they don't want data centers in their communities: per recent Gallup polling, seven in ten Americans oppose such construction. The reasons opponents give matter, too: resource use tops the list, with water and energy each cited by nearly one in five.

Some of these concerns are justified: issues of grid strain, water consumption, and noise pollution around data center construction are well-documented, with household electricity rates climbing in several markets alongside data center loads. Some, on the other hand, outpace the evidence and discount the genuine opportunities such constructions bring to local communities.

In any case, this sudden swing in public perception has created serious problems for hyperscalers, who now find themselves on the receiving end of a rash of painful regulations. Denver, Oklahoma City, and Aurora, IL have all approved moratoriums on data center construction. Oklahoma is weighing a statewide, three-year pause on large facilities, while South Dakota has enacted a law affirming local governments’ authority to restrict or ban data centers; and in Wisconsin, after three Manitowoc County towns passed resolutions opposing a prospective project, the county board adopted an 18-month moratorium of its own.

This conflict can and will be solved in time. It will require a new approach from the hyperscalers, one that centers on genuine engagement, not generic messaging around the economic benefits that these data centers can provide. In the meantime, though, demand for AI grows unabated: the AI companies have no time to wait. It is in this context that new, more sustainable solutions become all the more valuable.

Boosting inland transitions with power infrastructure solutions

Beyond the hyperscale campuses that dominate headlines sits a vast fleet of mid-market and colocation facilities – already built, already permitted, already connected to the grid. With new interconnection requests facing multi-year queues, this existing fleet is the fastest capacity available anywhere. A new approach to power and workload management can activate it now, when it is most needed.

These brownfield sites are considered "legacy" infrastructure, but this is a misnomer. They have a latent supply of power and compute, both of which can be effectively routed towards AI computing needs. Accessing that power and compute merely requires the use of intelligent orchestration.

Renewable energy sources play a key role here. US tech companies, to a large degree, still lag behind their European counterparts in deployment of sustainable energy solutions. That doesn't need to be the case. Newly affordable and easy-to-acquire Battery Energy Storage Systems (BESS) can charge up when power is abundant, when the sun is shining, when the wind is blowing, and discharge it when it isn't, helping to bridge the gaps when it comes to power demands. At the same time, grid-aware technology can dial back energy use when the grid is strained, and draw more energy when it's cheap and plentiful.

What intelligent orchestration does is synchronize these factors and countless others to ensure power is being drawn with maximum effectiveness. The idea here is to replace the old, lumbering in-and-out paradigm, wherein tasks are completed sequentially, regardless of urgency or energy needs, with dynamic scheduling, which is much more flexible. A chatbot query, for instance, might be routed instantly, while a more complex training operation can be saved for periods when energy is more affordable, and high-intensity jobs can be shifted to zones with more aggressive cooling. The results are higher utilization and a lower energy cost per unit of compute.

Demonstrating good faith through intelligent orchestration

To an extent, even the most fervent anti-data center advocates are aware that they are not going to meaningfully curtail the industry’s ambitions. What they want – and, frankly, what they deserve – is some input into how these data centers will affect their lives. That conversation will be a long one, but it will be resolved eventually.

Orchestration gives operators something concrete to bring to that conversation. A facility that can forecast its load profile and flex its demand becomes a grid asset rather than a grid burden: it can participate in demand response programs, shed load during peak hours, and help stabilize the very system its neighbors worry it will strain.

In the meantime, operators aren't exempt from addressing power consumption issues now. The rise of intelligent orchestration tools provides an opportunity for developers to, at one and the same time, meaningfully reduce energy costs while demonstrating good faith towards the communities they are entering.

Win-wins, in the context of data center construction, are few and far between. More often, on both sides and especially in the media, the issue is framed as winner-takes-all. Intelligent orchestration is one way, a highly effective one, that developers can demonstrate that this is a conversation, as opposed to a battle.