Data center operators are bracing for a critical power shortage around 2027 or 2028 that could dramatically slow AI infrastructure growth. While demand for computing power continues to surge, the availability of electrical power to feed these massive facilities is becoming the primary bottleneck, not the chips themselves. Power demand from data centers is projected to double by 2035 to almost 9% of all U.S. electricity consumption, marking potentially the biggest spike in demand since air conditioning became widespread. The challenge is immediate and concrete. Microsoft's CEO recently framed the main barrier to data center expansion as one of power supply rather than computing capacity, citing more demand than the company's data centers could currently handle. This shift in perspective signals a fundamental change in how the industry thinks about growth constraints. What Exactly Is the "Power Cliff" Experts Are Warning About? Industry insiders describe an anticipated power shortage that will arrive in waves. For the next year or two, growth will continue at an accelerated pace. But once available power runs out in 2027 or 2028, deal flow and new data center development are expected to slow significantly. One executive director at an international bank investing in data centers put it bluntly: "For the next year or two, it's all growth, maybe even a hockey stick curve. Once power runs out in 2027 or 2028, that's where we think deal flow will start to slow down". Quantifying the exact shortage is difficult because it depends on variables like regional power generation capacity, demand forecasts, and infrastructure investments. However, industry stakeholders agree that energy availability and redundancy is "far and away the number one obstacle" to successful data center development, both today and through 2030. Why Is Power Supply Suddenly the Limiting Factor? The scale of modern AI and high-performance computing (HPC) data centers has exploded. New hyperscale facilities are 10 to 50 times larger than data centers built just five to ten years ago, placing enormous stress on electrical infrastructure that was designed for a different era. These massive facilities require unprecedented amounts of continuous power, and the grid in many regions simply cannot accommodate them without significant upgrades. Beyond raw power generation, there are cascading supply chain challenges that compound the problem. Developers and operators report critical shortages across multiple components needed to build and operate these facilities. Steps to Navigate the Power Shortage Challenge - Tariff Relief for Critical Components: Policymakers should reduce tariffs on specialized computing hardware, cooling systems, transformers, and power equipment to accelerate manufacturing and reduce lead times that delay construction schedules. - Tax Incentives for Domestic Manufacturing: Governments can expand domestic production capacity through tax incentives, reducing offshore dependencies and supply chain vulnerabilities that currently constrain data center deployment. - Streamlined Permitting Processes: Regulatory approval remains a major obstacle, with 53% of developers citing permitting as the top barrier to data center development, so expedited review processes could unlock projects that are otherwise ready to proceed. What Supply Chain Bottlenecks Are Slowing Development? The shortage of specialized computing hardware is the most acute problem. A significant majority of providers, 68%, cite specialized computing hardware like graphics processing units (GPUs) and artificial intelligence (AI) chips as the most difficult component to source. Beyond chips, cooling systems present another critical constraint, with 62% of providers reporting sourcing difficulties. Transformers, batteries, rare earth materials, and power equipment like generators and switchgear all face extended lead times that impact construction schedules. Manufacturing capacity constraints are the root cause. The sudden demand for equipment at scales never before required has outpaced supplier capacity. Additionally, many critical components are sourced from outside the United States, making tariff and trade policy confusion a significant complicating factor. How Are Communities Responding to Data Center Growth? Recent community opposition to data center development has created unexpected regulatory friction. Public outcry has caused local governments to reject or delay data centers even in traditionally business-friendly areas, driven by concerns over water consumption, grid reliability, and general opposition to AI expansion. Regulatory and permitting challenges are cited as the phase where deals most commonly break down, with 48% of respondents reporting this issue. "Big tech and data center companies are an easy target for politicians, utilities, and local groups to blame for rising energy prices, even when it is not necessarily true. Navigating this issue and correcting the narrative will therefore become increasingly important," stated Jeff Atkin. Jeff Atkin, Foley & Lardner LLP The reality is more nuanced than public perception suggests. Many operators are actually the largest investors in their host communities, generating boosts to local tax rolls and creating construction jobs. Most developers are also transitioning from water-intensive evaporative cooling to closed-loop systems that use considerably less water. Some newer developments will actually return more water to local communities than they consume. Infrastructure improvements like utility upgrades, substations, and road improvements can extend benefits to previously undeveloped areas. What Does This Mean for the Future of AI Infrastructure? The consensus among industry leaders is that demand for data center capacity will continue to grow. While some consolidation or "right-sizing" may occur, there has never been a period in human history where usage of storage or computing has contracted in any meaningful way. Demand tends to expand to meet supply. Quantum computing, which some feared might disrupt the data center demand model, is ranked very low as a current obstacle at just 6%, suggesting it is not expected to fundamentally change the trajectory in the near term. The immediate challenge remains straightforward: building enough power generation and distribution capacity to support the AI revolution that is already underway. The window between now and 2027 represents a critical period for policymakers, utilities, and developers to coordinate on infrastructure expansion. Without significant action on power generation, transmission, and permitting, the industry's explosive growth could hit a hard ceiling just as AI applications are reaching mainstream adoption.
