The AI infrastructure buildout has generated enormous analysis about hyperscalers, data center developers, GPU manufacturers, and cooling technology vendors. The entity that has received the least attention relative to its actual influence over what gets built, where, and when is the regulated utility. That is now changing rapidly, and the shift has implications that extend across every dimension of AI infrastructure strategy.
As the constraints on AI infrastructure development have shifted from capital and demand to power access and grid capacity, utilities have emerged as the decisive gatekeepers of the entire buildout. Their decisions, their timelines, and their regulatory positions now determine which AI infrastructure projects proceed and which do not. This is a structural shift, not a temporary bottleneck, and it is happening largely outside the attention of the technology press.
The Entity Nobody Is Watching
The shift is not immediately obvious from the surface of the AI infrastructure story. Headlines focus on hyperscalers announcing hundred-billion-dollar capex commitments and AI lab valuations reaching unprecedented levels. Utilities do not generate that kind of coverage. They are regulated monopolies operating under frameworks designed in a different era, managing assets that move slowly, and communicating through regulatory filings rather than press releases. Yet it is precisely this structural position, combining monopoly control over grid interconnection in most US markets with regulatory relationships that shape the economics of every project that needs grid power, that has given utilities leverage that no amount of capital or computing power can bypass.
Understanding why this happened, how utilities are responding, and what it means for data center developers, hyperscalers, and infrastructure investors is becoming essential context for anyone operating in this market.
The irony is that the utility’s leverage did not grow because utilities became more powerful organisations. It grew because everything else the AI infrastructure buildout depends on, capital, compute, talent, land, became more abundant precisely when power did not. In a market defined by a single constraint, the entity that controls that constraint holds disproportionate influence. Regulated utilities now occupy that position in the AI infrastructure market, and they are likely to hold it for at least the next five years because building the transmission infrastructure needed to relieve current congestion takes that long.
How Utilities Became the Bottleneck Nobody Planned For
For most of the commercial data center era, power was a procurement challenge rather than a strategic constraint. Developers selected sites based on available power, negotiated rates with utilities, and built out capacity against confirmed grid connections. The utility was a vendor in this relationship, providing a commodity service against established tariffs and interconnection processes designed for loads within the normal operating experience of regional transmission operators.
That model worked because data center demand, while growing, grew in ways that utilities could accommodate within their existing planning cycles. A 20-megawatt data center is a meaningful load, but one utilities had built interconnection processes to handle. Step up to 200 megawatts and the challenge grows, but it is not categorically different. A 500-megawatt AI campus, however, represents something genuinely new. It is a load that exceeds the total industrial consumption of many utility service territories.
When AI Infrastructure Changed the Rules
AI infrastructure changed the demand profile in ways that utilities had not modelled and were not structured to accommodate. Specific transmission infrastructure must be planned, permitted, and built to serve loads at this scale. Furthermore, the sharp load ramps associated with AI training workloads stress grid stability in ways that conventional industrial loads do not. A training cluster consuming 300 megawatts can ramp from idle to full load in minutes, creating voltage and frequency disturbances that ripple through regional grids.
Dominion Energy Virginia serves the highest concentration of data centers in the world and has disclosed interconnection wait times of up to seven years for loads above 100 megawatts. Lawrence Berkeley National Laboratory’s most recent Queued Up analysis found that only 13% of capacity entering interconnection queues between 2000 and 2018 was ever actually built. The median time from interconnection request to commercial operation has reached five years. That is not a temporary backlog. It is a structural mismatch between infrastructure that moves at regulatory and engineering timescales and demand that moves at technology company timescales.
The Planning Cycle Problem
For data center developers accustomed to 18-month construction timelines, a five-year interconnection process fundamentally changes how they must conceive and finance projects. Furthermore, utilities managing these queues operate under regulatory frameworks designed for a different era. The same regulatory processes that govern rate structures and capital investment programmes also constrain their ability to accelerate interconnection.
A utility that wants to build transmission infrastructure to serve a new AI campus at speed still needs regulatory approval for the capital expenditure. That means a rate case filing, a review period, and a determination that the investment is prudent and in the public interest. When ratepayer advocates and environmental groups intervene, as they increasingly do in proceedings that involve large new industrial loads, the timeline extends further. The regulatory calendar and the technology company calendar are simply not aligned, and there is no obvious mechanism to close that gap quickly.
The interconnection queue has also become a strategic chokepoint in ways that go beyond simple timing. Projects that are early in the queue have a structural advantage over those that are late, and that advantage compounds over time. The queue is not a waiting room where everyone eventually gets served. It is a filter through which only a fraction of applications ever become operational capacity. For data center developers, that means the decision to submit an interconnection application in a specific service territory is, in effect, a long-dated option on capacity in that market. The developers who recognised this early, and who systematically built positions in the most attractive service territories before the current wave of demand made those positions expensive to acquire, have a multi-year head start that is not easy to close.
Why Utilities Now Hold Structural Leverage
The leverage utilities hold is not merely a function of queue length. It is structural, and it operates through several distinct mechanisms that collectively give utilities influence over the AI buildout that no other market participant currently matches.
The first mechanism is interconnection approval. Without a confirmed grid connection, a data center developer cannot bring a facility online. In markets where queues stretch years, the utility’s decision about queue position and cost allocation directly determines a project’s commercial viability. Queue position matters as much as site selection. Moreover, utility relationships are not transactional assets that companies can acquire quickly. Developers build them through years of engagement, proven track records of successful project delivery, and accumulated goodwill with engineers and key account managers who, in practice, determine how complex interconnection negotiations proceed.
The Rate Structure Mechanism
The second mechanism is rate structure authority. Regulated utilities in most US markets have the ability to propose rate structures that affect how data center electricity costs are calculated and allocated. State regulators, in turn, have the ability to approve them. As electricity rate increases in data center markets have become a political issue, utilities are under increasing pressure to demonstrate that data centers are not shifting costs onto residential ratepayers.
The Ratepayer Protection Pledge signed at the White House in March 2026 reflects this pressure precisely. It also reflects the reality that utilities, not data center operators, are the entities with ongoing rate-setting relationships with state public utility commissions. Consequently, the terms on which AI infrastructure development proceeds are increasingly being negotiated through regulatory processes where utilities are the primary interlocutors and data center developers are, in effect, petitioners. A utility that determines large industrial customers should bear a larger share of transmission infrastructure costs is making a decision that can materially affect the economics of projects already under development.
The Transmission Investment Mechanism
The third mechanism is the timing and sequencing of transmission investment. Utilities that need to build new transmission infrastructure control the timeline of that investment. That timeline is subject to their own capital planning cycles, their regulatory approval processes, and the availability of specialised equipment. As we have covered previously in our analysis of the silent bottleneck of transformer and substation supply chains, the electrical equipment supply chain was already under severe pressure before the current wave of AI infrastructure investment added further demand.
The utility’s position within that supply chain adds another layer to its structural leverage. Its ability to prioritise transformer procurement for specific interconnection projects can determine which ones proceed on their planned timelines and which are delayed by equipment availability. A developer that has built a strong utility relationship is more likely to receive preferential treatment in procurement sequencing than one that has treated the utility as a commodity supplier.
The Geography of Utility Leverage
The leverage that utilities hold is not uniformly distributed across US markets. It varies significantly by service territory, by the degree to which the local utility has invested in interconnection capacity, and by the regulatory environment in which the utility operates. Understanding this geographic variation is essential for developing a coherent AI infrastructure strategy, because the choice of where to build is increasingly inseparable from the choice of which utility to work with.
Northern Virginia, served primarily by Dominion Energy, represents one end of the spectrum. Dominion has the most data center experience of any utility in the world, having managed the buildout of data center alley for decades. It has invested significantly in the infrastructure and expertise required to serve hyperscale loads, and northern Virginia remains one of the most sought-after locations for AI infrastructure investment globally. However, Dominion’s queue is also the most congested in the country. The seven-year wait times disclosed for loads above 100 megawatts reflect a market where demand has structurally outrun supply.
The Southeast and Beyond
Duke Energy in the Carolinas and Georgia Power in Georgia have both positioned themselves as alternatives to the northern Virginia market. Both have invested in grid capacity and streamlined interconnection processes for large industrial loads. Both are operating in regulatory environments where state economic development priorities are more explicitly supportive of data center investment. The result is that projects defaulting to northern Virginia a few years ago are now evaluating the Carolinas and Georgia as primary options, and the utility relationships in those markets are consequently becoming more valuable and more competitive.
In Texas, ERCOT’s deregulated market structure offers different dynamics. Behind-the-meter generation is more straightforward than in regulated markets, which is why several gigawatt campus projects have gravitated to Texas. Grid interconnection for grid-connected facilities requires navigating ERCOT’s own interconnection process, which has its own queue dynamics. Texas also carries weather-related resilience risks demonstrated by the 2021 winter storm event. Developers choosing Texas are consequently making a different set of trade-offs between regulatory simplicity, behind-the-meter flexibility, and grid resilience compared to those choosing regulated markets in the Southeast or Mid-Atlantic.
The International Comparison
The Gulf states offer a relevant comparison point for understanding what an explicitly utility-enabled AI infrastructure strategy can look like. As we have covered in our analysis of the Gulf’s digital recalibration, the combination of state-directed utility investment, sovereign wealth fund capital, and explicit policy support for AI infrastructure development has allowed Gulf markets to attract data center investment at a pace that competitive US markets are struggling to match on pure timeline grounds.
The lesson for US utility regulators is not that they should replicate state capitalism. It is that the competitive pressure from international markets is real and that the regulatory friction in US interconnection processes is a genuine disadvantage relative to markets where utility investment is directed by national strategy rather than determined by rate case proceedings. That competitive pressure is increasingly visible in economic development policy discussions at the state level, where governors are pushing utilities and public utility commissions to move faster without always having the regulatory authority to make that happen.
How Utilities Are Responding to AI Demand
Utilities are not passive participants in this dynamic. Several are actively competing for AI infrastructure investment in ways that reshape both their own strategic positioning and the competitive landscape for data center development in their territories. The shift from utility-as-vendor to utility-as-strategic-partner is one of the most significant structural changes in the AI infrastructure ecosystem, and it is happening largely outside the attention of the technology press.
Duke Energy in the Carolinas has developed a dedicated data center interconnection programme that provides more predictable timelines, a dedicated team of engineers and relationship managers, and a streamlined permitting process for projects above certain scale thresholds. Georgia Power has invested in transmission capacity upgrades in the corridor north of Atlanta that has become a major data center market, anticipating demand rather than simply responding to it. Both utilities are making explicit bets that competing effectively for AI infrastructure investment will generate long-term load growth and revenue that justifies the upfront capital and process investment.
The Two-Tier Utility Landscape
As a result, a two-tier utility landscape is emerging. Utilities that have invested in interconnection capacity, developed streamlined processes for large industrial loads, and built genuine developer relationships are attracting a disproportionate share of AI infrastructure investment. Utilities that have not are, by contrast, losing projects to competing service territories. That competitive dynamic is creating pressure for regulatory reform at the state level. Governors and economic development agencies are pushing utilities and public utility commissions to accelerate interconnection processes for data center projects, with the economic development argument proving increasingly persuasive in state capitals.
The two-tier dynamic is self-reinforcing. Utilities that attract AI infrastructure investment gain operational experience managing large, complex loads, which makes them better at it, which makes them more attractive to the next round of developers. Over a five-year horizon, that compounding advantage is likely to produce a significant and durable gap between utility service territories that have positioned themselves for AI infrastructure and those that have not.
The regulatory environment is also creating pressure from an unexpected direction. As ratepayer advocates and community groups have become more sophisticated about data center economics, they are asking utilities in rate cases whether other customers are subsidising data center interconnection. Utilities that have developed clear, defensible frameworks for data center cost allocation are better positioned to avoid the regulatory friction that can slow interconnection approvals and complicate relationships with state regulators, whose support is ultimately required for utility capital programmes to proceed.
The Hyperscaler Utility Investment Model
The hyperscalers are responding to the utility leverage dynamic in a way that is itself reshaping the market. Amazon Web Services, Google Cloud, and Microsoft Azure have all expanded their utility engagement teams significantly over the past two years. In some cases, they are committing to fund transmission infrastructure upgrades directly in exchange for priority interconnection access. That willingness to invest in utility infrastructure, rather than simply waiting in queue, reflects a recognition that the traditional procurement model for grid power is no longer adequate at the scale these companies need to grow.
Google’s approach has been particularly notable. The company has publicly committed to funding specific grid infrastructure upgrades in markets where it is building large campuses, effectively accelerating the utility’s capital programme in exchange for queue priority. That model, sometimes called a customer-funded interconnection, is becoming more common as hyperscalers with the financial resources to do so realise that funding grid upgrades themselves is cheaper than waiting in queue for the utility to fund them through its normal capital programme.
What This Means for Data Center Developers and Hyperscalers
The shift in leverage toward utilities has practical implications for how data center developers and hyperscalers should approach site selection, project development, and capacity planning. Those implications are already visible in how the most sophisticated operators are structuring their development processes, and they will become more visible over the next three to five years as the gap between utility-savvy developers and those who have not adapted continues to widen.
Site selection criteria that once focused on land cost and fibre proximity now need to give equal weight to the utility relationship and the realistic interconnection timeline. A site with low land costs and available fibre is worth little if the interconnection queue is seven years. It is worth even less if the utility has limited capacity to accelerate it. The site’s value is consequently a function not just of its physical characteristics but of its position in the utility’s strategic priorities and of the developer’s standing with the utility’s interconnection and key accounts teams.
The Utility Relationship as a Core Competency
Developers with deep utility relationships and a track record of successfully delivering on their interconnection commitments are commanding premium valuations from hyperscalers seeking sites with confirmed or near-confirmed power access. Hyperscalers are willing to pay for that certainty. The premium reflects the option value of knowing that the interconnection timeline is reliable, that the rate structure is unlikely to be adversely revised, and that the utility relationship is robust enough to navigate the operational issues that inevitably arise during large facility commissioning.
For hyperscalers, the utility relationship is increasingly a core competency rather than a procurement function. The questions involved, specifically which markets to enter, which utilities to partner with, what level of transmission investment to underwrite, and how to navigate regulatory proceedings that determine rate structures and interconnection terms, are strategic questions that require executive-level attention and relationships. The utility engagement function, which historically sat within real estate or facilities management, is migrating toward the C-suite in the most sophisticated operators.
The C-Suite Elevation of Utility Engagement
The C-suite elevation of utility engagement reflects a recognition that the decisions made in this domain have consequences that extend across the entire organisation. A hyperscaler that commits to a large AI campus in a specific utility service territory is making a ten-year relationship commitment to that utility. The terms of that relationship, including rate structures, interconnection timelines, and the allocation of transmission infrastructure costs, will affect the economics of the facility for its entire operational life.
The regulatory dimension is particularly important. Public utility commissions in most US states conduct proceedings that determine how costs are allocated between different customer classes, how utilities may invest their capital, and what rate structures apply to large industrial customers. Data center developers and hyperscalers that participate actively in these proceedings, either directly or through industry coalitions, have the ability to shape regulatory outcomes that affect their costs for years into the future. Those that do not participate are subject to outcomes determined by other parties in those proceedings, which typically include residential ratepayer advocates and environmental groups whose interests may not align with AI infrastructure development.
The Competitive Moat That Utility Relationships Create
The utility relationship is not just a prerequisite for development. In the current environment, it creates a competitive advantage that competitors cannot easily replicate. Queue positions in attractive utility service territories are finite and non-transferable. Developers cannot purchase relationships built over years of successful project delivery. They also cannot accelerate the operational history needed to show public utility commissions that their projects deliver promised benefits without creating problems for the grid.
The developers and hyperscalers who have invested most deeply in utility relationships over the past five years are now operating from positions that are structurally advantaged relative to those who have not. They have queue positions that latecomers cannot replicate without waiting years. Relationship capital with utility interconnection teams translates into more collaborative and therefore faster interconnection processes. Their regulatory track records make rate case proceedings go more smoothly.
The Financial Value of Confirmed Power Access
For investors evaluating AI infrastructure companies, the quality of the utility relationship portfolio is becoming a meaningful component of investment analysis. A developer that can demonstrate confirmed interconnection agreements in multiple high-demand utility service territories, with realistic timelines and favourable rate structures, is demonstrably more valuable than one with the same amount of planned capacity but without those confirmations.
The market is beginning to price this distinction, as build-ready sites with confirmed power access now command a valuation premium over sites still waiting in the queue. As we have covered in our analysis of how data centers are becoming power infrastructure companies, investors now incorporate the financial value of confirmed near-term power access into data center valuations at a premium that did not exist before the current power constraint became structurally visible. In other words, the utility relationship has become both a financial asset and an operational one. That is a genuinely new development in the economics of AI infrastructure.
The Political Dimension of the Utility Alliance
The utility relationship dynamic is also shaping how hyperscalers approach the political dimension of AI infrastructure development. Utilities have long-standing relationships with governors, state legislators, and public utility commissioners that technology companies do not. A hyperscaler that works through its utility partner to advocate for regulatory reform is more likely to be heard than one that approaches state government directly as an outside investor seeking preferential treatment.
One utility that is publicly supportive of data center development in its service territory, framing the investment as an economic development win and a grid modernisation opportunity, can provide significant political cover for developers facing community opposition. One that is neutral or ambivalent is of less value. Most harmful of all is a utility that frames data center development as a burden on residential ratepayers, as the political dynamics we have covered in our analysis of the AI data center backlash becoming a swing-state political issue increasingly put utilities in precisely this framing. The quality of the utility relationship has a direct effect on the political environment in which development proceeds.
The Structural Reality Nobody Can Engineer Around
The regulated utility did not ask to become the most powerful player in AI infrastructure. Circumstance and the structural logic of grid economics made it one. It gained that position because power demand grew faster than the infrastructure needed to deliver it, and because regulatory and operational processes control access to grid power in ways that capital alone cannot easily accelerate.
In a market where power is the primary constraint on growth, the entity that controls access to power controls the pace of the entire buildout. The developers and hyperscalers who understood this earliest, and invested accordingly in the utility relationships, regulatory engagement, and long-term power strategies that the current environment requires, are building a sustainable competitive advantage. Those who have not are operating at a structural disadvantage that capital alone cannot overcome.
The market is now beginning to reflect that distinction in valuations, financing costs, and the widening gap between the projects that are commissioning capacity and those still searching for their first confirmed grid connection. In an asset class where power access has become the primary determinant of project viability, the entity that controls power access has become the primary determinant of competitive outcomes. That is the structural reality of AI infrastructure in 2026, and it is one that the most sophisticated operators have already incorporated into how they build, invest, and plan. The rest of the market is catching up, and the cost of doing so is measured in time, capital, and the competitive distance that separates those who moved early from those who did not.
