One of the largest power grid operators in the U.S. is asking Federal Energy Regulatory Commission (FERC) to only allow large data centers to connect to the grid if there’s enough capacity available to serve them reliably, according to Utility Dive.
The infrastructure underlying big data processing, typically large-scale data centers, is starting to receive pushback as more residents in the U.S. see climbing energy costs.
These facilities, which house servers, storage arrays, and network equipment, require continuous power for consumption and cooling, which contribute significantly to global electricity demand. The U.S. Department of Energy estimates that a single hyperscale data center can consume between 20 and 50 megawatts of power, which is enough to power 30,000–75,000 homes.
Right now, PJM has no authority to stop those connections, and data centers are allowed to connect even if there isn’t enough electricity to go around, according to Tom Rutigliano and Claire Lang-Ree, both from the National Resources Defense Council, wrote in an op-ed to Utility Drive. If nothing changes, by June 2027, the region may fall below reliability standards. That means a greater risk of rolling blackouts, especially during winter storms and summer heat waves. PJM appears to agree: It recently published scenarios that show deep shortages in 2030.
The Harvard Business Review finds that these massive facilities not only strain already stressed power grids but also create air pollution, including fine particulate matter, resulting in significant respiratory-related health consequences that are estimated to cost up to $20 billion per year in the United States by 2028.
Rutigliano and Lang Ree suggest that “data centers should be looking to rapidly obtain firm service by investing in energy storage and being as flexible as possible. Right now, in PJM, a 6-hour battery contributes more to reliability than a gas-fired combustion turbine, thanks to the battery’s better reliability in the winter. This storage also provides a crucial missing link PJM needs to become a low-carbon grid. A modest investment on the scale of data center development can build enough storage to cover the demands of data centers and provide a clear path to zero-carbon operations. Any tech company that’s serious about its climate commitments and speed to market should be looking very seriously at large-scale, grid-connected energy storage.”
Here, experts in the field offer their predictions for what 2026 holds for data centers:
2026’s Data Center Capital Storm: AI and compute drive a new era of investment: The scale of investment required to support AI’s growth is unlike anything we’ve seen before, and 2026 will mark the beginning of an unprecedented capital storm in data center infrastructure. The expected 15 gigawatts of U.S. data center leasing activity in 2025 alone demands roughly $150 billion in infrastructure funding—before even accounting for the $600 billion in chips that will power those facilities. With interest rates expected to move meaningfully lower, a massive wave of capital—both equity and debt—will unlock, accelerating projects across every major hyperscale and colocation market. That’s only the beginning. Leasing is projected to jump another 25% in 2026, pushing total capacity demand toward 19 GW and setting another record. The trajectory of compute demand is staggering as training runs that consume 150–200 MW today could reach multi-gigawatt scales by 2030. Additionally, U.S. AI power requirements will potentially balloon from 5 GW today to 30–70 GW by 2030. Every year, the industry breaks new records, and in 2026, the flood of capital chasing AI infrastructure will redefine the boundaries of scale.—Tom Traugott, SVP of emerging technologies, from EdgeCore Digital Infrastructure
Modularity will be key to scaling liquid cooling in AI data centers: As AI workloads continue to drive power densities ever higher, data center operators will seek out more powerful, modular liquid cooling systems that can be easily deployed and scaled incrementally as thermal regulation needs grow. By late 2026, expect to see skidded, modular units starting at 2MW (and reaching well beyond) become the de facto models for high-density data center builds.—Angela Taylor, chief of staff & head of strategy, LiquidStack
Operational Technology (OT) exploitation will bring physical consequences to the forefront: Operational Technology (OT) and critical infrastructure will move from an under-the-radar concern to a high-impact threat. Nation-state and criminal groups will aggressively exploit the growing number of significant security maturity gaps that still exist between IT and OT systems. The targeting of systems like building management and industrial controls—evidenced by threat actors like Volt Typhoon—will lead to several minor-to-moderate, localized disruptions of essential services that will elevate public awareness of OT risks.—Stephen Boyer, co-founder and chief innovation officer, Bitsight
Power will become more of a commodity for tech companies: The defining bottleneck of 2026 will be power, not GPUs, as data centers hit grid limits, cooling capacity plateaus, and the energy required for inference grows faster than semiconductor supply can accommodate. This power squeeze represents a generational inflection point, analogous not to chip scarcity but to the early electrification era before widespread grid expansion, a fundamental physical limit overtaking raw compute as the barrier to growth—Natalie Hwang, founding managing partner at Apeira Capital
Optimization becomes the next frontier for data centers: Every major technology shift follows the same pattern: invest, then optimize, and finally, harness value. 2025 saw unprecedented levels of data center investment as companies look to scale up AI initiatives, but the tides are beginning to turn. CFOs want to see results. Rather than investing in more hardware to solve compute performance shortcomings, optimization will be the story of 2026, as companies look for ways to harness value from their existing investments.—JG Chirapurath, president at DataPelago
The Infrastructure Integration Imperative: The era of point solutions is ending. In 2026, data centers will increasingly be defined by co-design—where compute, power delivery, and cooling are engineered as a single system rather than assembled as parts. The M&A wave we’re seeing—with hyperscalers acquiring energy assets and power and cooling vendors consolidating—reflects the market’s understanding of this convergence. The focus has shifted from buying components to building AI-ready ecosystems.—Josh Claman, CEO at Accelsius
Use of temporary buildings and modular solutions will be vital to achieve maximum “time to power” for data center deployments: We’ve reached a point where data centers can’t be built fast enough. To jumpstart deployments and ensure available power is utilized immediately, we’ll see more organizations turn to modular data centers and temporary buildings, such as Microsoft's use of tents, for data center operations until a permanent on-site facility is completed.—Kevin Roof, director of offer and capture management, LiquidStack