By Lina Irawan 
At the heart of this unfolding scenario lies a critical and rapidly emerging imbalance: the pace at which leading AI developers and hyperscalers can conceptualize, finance, and construct cutting-edge AI infrastructure far outstrips the U.S.’s capacity to concurrently develop the necessary electricity generation and transmission infrastructure required to power it. This divergence in speed creates a significant chasm between technological ambition and foundational energy realities.
"There is a real disconnect there," Rob Gramlich, president of the consulting firm Grid Strategies, recently told Fortune. He elaborated on the fundamental differences in operational paradigms: "Tech companies are famous for moving fast," he noted, driven by market pressures, venture capital, and the imperative to innovate quickly. In stark contrast, utilities are "notoriously move very slowly," a characteristic born not of inefficiency but of necessity. Their mandate is to ensure the reliability and safety of an intricate, interconnected grid comprising thousands of components, from power plants to distribution lines. This complex undertaking demands "deliberate study and planning" that can span many years, involving extensive engineering, regulatory approvals, environmental impact assessments, and public consultations.
This inherent "mismatch in timing," as Gramlich described it, is precisely why the nation’s electric grid finds itself ill-prepared to meet the escalating electricity demands of the burgeoning data center industry. The projections are staggering: data centers are forecast to consume nearly 12% of all U.S. electricity by 2030. This figure represents an almost six-fold increase from their pre-AI boom share in 2018, according to groundbreaking research by the Lawrence Berkeley National Lab, a federally funded scientific research center renowned for its work on energy systems.
The AI revolution isn’t merely adding to existing demand; it’s accelerating it beyond previous expectations. The North American Electric Reliability Corporation (NERC), a critical grid watchdog, underscored this acceleration in its 2025 assessment of the grid’s reliability. The report projects that North America’s summer peak electricity demand will surge by more than 224 gigawatts (GW) over the next decade. Crucially, this forecast represents a monumental 69% increase above the growth projected just a year earlier, with new AI data centers identified as the primary driver. The impact is particularly pronounced in certain regions; within the Western U.S. grid region, planned data centers alone account for an average of 10% of demand forecasts, a figure that can skyrocket to as high as 40% in specific localized areas, placing immense strain on regional grids.
Even before the advent of the AI boom, the U.S. electric grid was already experiencing significant constraints. Gramlich pointed out that the U.S. electricity industry had "got out of practice on building new infrastructure" following a quarter-century of relatively stagnant power demand. This period of minimal growth, he explained, was a consequence of earlier, massive grid expansions in the 1980s and 1990s, which left utilities with ample excess capacity. The subsequent two decades, from 2000 to 2023, did not witness the same explosive electricity demand growth that characterized previous eras, when the widespread adoption of kitchen appliances and residential air conditioning units dramatically reshaped household energy consumption.
However, the landscape has fundamentally shifted. "But now we have electric vehicles, we have electric space heating, we have new manufacturing, much of which is using electricity more than traditional manufacturing, and then we have data centers," Gramlich elucidated. He stressed the synchronized nature of these new demands: "The growth is about half data centers and half those other new things, but those new quickly expanding uses of electricity are happening all at once." This confluence of electrification trends, from transportation to industry, coupled with the insatiable appetite of AI, presents a cumulative challenge that the grid, designed for a different era, is struggling to accommodate.
Access to Electricity as a Bottleneck to Data Centers
The practical implications of this mismatch are profound, manifesting as significant bottlenecks for data center development. Kathryn Burke, who leads U.S. specialty energy and power growth at the insurance firm Marsh, articulated this pressing issue to Fortune. She stated unequivocally that access to power is "probably the number one, if not top five bottlenecks for data center development in the U.S. right now." The demand, she explained, is not just for more power, but for a faster rate of delivery than the existing energy grid can realistically provide, given its current state and the inherent challenges of upgrading it.
Companies are requesting gigawatts of electricity—a monumental amount that, for context, can power approximately 750,000 U.S. homes—with an expectation of delivery often within two years. Burke highlighted the stark reality: generating and transmitting a gigawatt of electricity can take "a lot more time" than these ambitious timelines suggest. Independent research from Berkeley Lab, focusing on interconnection queues, corroborates this. U.S. power projects that ultimately came online in 2023 spent a median of five years navigating the process from the initial request for a grid connection to achieving commercial operation. This five-year median timeline starkly contrasts with the two-year expectation of many data center developers, creating an unavoidable scheduling conflict.
Compounding the generation challenge is a severe deficit in the country’s electricity transmission infrastructure. The Department of Energy (DOE) underscored this deficiency in July, stating that there is a "pressing need" for significant expansion of transmission capacity. This urgency, the DOE explained, is "due to load growth from data centers" among other factors like a burgeoning economy and the integration of new renewable energy sources often located far from demand centers. Building new transmission lines is notoriously difficult, facing hurdles from local opposition (NIMBYism – Not In My Backyard), complex multi-state regulatory environments, lengthy permitting processes, and substantial capital investment. Without adequate transmission, even if new generation capacity is built, the power cannot reliably reach the data centers where it is desperately needed.
The most immediate and widespread consequence of this grid strain may not necessarily be widespread blackouts across residential areas, though localized outages could certainly occur. Gramlich reassured that utilities, operating under strict reliability mandates, generally do not connect new customers if they cannot guarantee reliable service. Instead, a more probable scenario is that data centers will face years-long waits for full, uninterrupted service. Alternatively, they may be forced to accept provisional connections, under which their electricity supply can be curtailed or interrupted during periods of peak demand or grid stress. This "interruptible" service, while offering some power, introduces significant operational risks and inefficiencies for critical AI workloads that require continuous uptime.
"There is scarcity on the grid," Gramlich affirmed. "Not everybody’s going to get the full level of service that they want, at least until the system can catch up to these new demands." This scarcity translates into concrete financial and logistical challenges for data center developers.
Burke further elaborated on the economic repercussions, noting that utilities are increasingly demanding greater financial commitments from data center developers. This is because utilities must invest substantial capital in new generation capacity and grid upgrades specifically to serve these energy-intensive facilities. They seek assurances that these investments will not be "stranded" if a data center project encounters unforeseen delays or ultimately fails to materialize. These upfront financial requirements add another layer of complexity and cost to data center development. Burke’s assessment paints a challenging picture for the industry: she predicts that around 50% to 60% of planned data center projects will experience delays, consequently failing to be operational within the one-to-two-year timeframe companies are optimistically targeting.
The uncertainty extends beyond mere delays. "It’s hard to predict how many of these data centers are actually going to get built at the end of the day if they get the financing, if they get all of the different constraints, and so that’s an area that is still a question mark, but we are seeing a lot of that slowdown," Burke concluded. This suggests a looming period of recalibration for the AI infrastructure sector, where the sheer pace of technological advancement is now directly confronting the slower, more methodical realities of energy infrastructure development. The future of AI’s expansion in the U.S. will, to a significant extent, hinge on the nation’s ability to bridge this critical power gap, demanding innovative solutions, expedited regulatory processes, and unprecedented collaboration between tech giants and traditional utilities.
