The North American grid is entering a structurally different phase. From 2000 to 2020, US electricity demand was essentially flat, growing at under 0.5% per year as efficiency gains offset population and economic growth. The system now faces a sharp reversal: concentrated industrial demand driven by AI data centers, electrification, and reshoring is pushing load growth to levels not seen in over 70 years. Interconnection queues are expanding, reserve margins are tightening, and transmission timelines are increasingly incompatible with hyperscaler deployment schedules. This post examines the evolving grid context, the technical pathways to relieve structural constraints, and how Azimuth is positioning within this transition.

Nonlinear Load Growth
AI has broken the 20-year pattern of essentially flat US electricity demand. US data centers consumed approximately 183 TWh of electricity in 2024, roughly 4% of total US consumption, and that figure is projected to more than double to 426 TWh by 2030. Individual campuses now range from hundreds of megawatts to multi-gigawatt clusters: loads equivalent to mid-sized cities. For context, a single hyperscale AI facility can consume as much electricity annually as 100,000 households. Electrification of transport and industry compounds the effect, but AI is the primary accelerant. The economic value of compute has fundamentally changed how large buyers think about power: getting connected to the grid quickly has become more important than minimizing electricity costs, because every month of delay in powering a data center carries competitive and financial consequences that far outweigh marginal differences in electricity pricing.

Transmission as the Binding Constraint
The most immediate bottleneck is not generation capacity but transmission and interconnection. Connecting a new power plant or large load to the grid requires a formal study and approval process that can take five to seven years in some regions. The backlog is enormous: PJM, Pennsylvania-Jersey-Maryland Interconnection, the largest regional transmission organization in the United States, has over 100 GW of proposed generation waiting in its interconnection queue, the vast majority being renewable energy. In Texas, interconnection requests from large loads jumped over three-fold from 63 GW at end-2024 to 230 GW in 2025, driven almost entirely by data center developers. Substation equipment shortages, permitting and land-use opposition compound the queue problem. The result is visible across regions: renewable energy being curtailed in areas with excess generation, and scarcity pricing in areas where demand is outpacing supply. Power cannot move efficiently enough to match supply with demand. Building the infrastructure to fix the imbalance takes years the market does not have.

Reliability as a Premium Product
Large industrial buyers increasingly view grid access as uncertain in both timing and firmness. Hyperscalers are exploring options such as accepting higher cost for firm power, co-locating generation with load, and private or hybrid grid architectures. The grid is no longer the default architecture; it is one option among several.

The constraints described above are not new in type, only in magnitude. What has changed is the pace: demand that grid planners would have expected to materialize over two decades is arriving in two to three years. That compression is what makes the current moment structurally different from prior cycles of grid investment. The technical responses fall into three broad categories: making better use of what already exists, increasing the physical capacity of the network, and building generation that bypasses the grid entirely. None of these is sufficient on its own. The most likely path forward combines near-term digital and operational fixes with longer-horizon infrastructure investments, while a parallel market for behind-the-meter power develops to serve loads that cannot wait.

Smarter Utilization of Existing Assets
Virtual power plants aggregate distributed resources – batteries, EV fleets, flexible industrial loads – into dispatchable capacity, providing peak shaving and reserve margin support without new infrastructure. Grid-enhancing technologies and dynamic line rating can increase effective transmission capacity through real-time monitoring alone. Both are pragmatic and near-term, though insufficient to offset structural load growth on their own. Long-duration storage can serve as capacity insurance and congestion mitigation, with cost and bankability as the gating factors.

Increasing Physical Throughput
Advanced conductors provide incremental line capacity gains but are transitional tools. High Voltage Direct Current (HVDC) transmission enables bulk long-distance power transfer with lower energy losses than conventional alternating current (AC) lines, making it well-suited to moving power from remote generation sources to load centers hundreds of miles away. Permitting timelines and capital intensity remain barriers to rapid deployment.

High-temperature superconductors represent the most significant step-change: transmitting multiples of conventional capacity within the same right-of-way, making them attractive in dense urban corridors where new siting is politically infeasible. The constraints are less in physics than in cryogenic systems, materials durability, and supply chain scale.

Solid-state transformers complement these innovations by enabling voltage flexibility, bidirectional flow, and compatibility with increasingly DC-native industrial loads.

Partial Decoupling: Behind-the-Meter Power
Where timelines are incompatible with grid expansion, large loads are building dedicated behind-the-meter generation—typically gas-fired, often hybridized with renewables—for accelerated time-to-power and greater operational control. The complexity is real: these projects must navigate permitting, fuel supply, emissions compliance, and the capital burden of building generation that would otherwise be purchased. Reliability engineering for islanded or semi-islanded operations is more demanding than grid-served load, and financing markets are still developing appropriate frameworks.

Texas moved first. Senate Bill 6, signed in June 2025, requires large power users above 75 MW to fund a portion of their own interconnection costs upfront, subjects co-location arrangements involving existing power plants to state regulatory approval, and mandates that any new large load connecting after December 31, 2025 install hardware allowing the grid operator to cut their power remotely during emergencies.

Federal regulators followed in December 2025, ordering PJM, the largest grid operator in the US, to establish rules requiring co-located loads to pay their share of grid infrastructure and ancillary service costs, regardless of how much power they actually draw from the grid. Rates are still being set through a regulatory process running into 2026, meaning cost exposure for projects already in development is not yet fully known.

Both actions close the same loophole: large loads using grid infrastructure without paying proportionate costs. Projects in either market that were financially structured around minimal grid cost exposure now carry repricing risk. Georgia, South Carolina, California, and Iowa have introduced similar measures, and this is becoming a national policy direction, not a regional anomaly.

Enoflex and the Superconductor Ecosystem
Enoflex is an Azimuth portfolio company manufacturing pipes from composite materials that find applications in many energy transition technologies. One example of this is high temperature superconducting cables, which depend on reliable cryogenic transport systems. Enoflex’s products have a smooth-bore pipe optimized for liquid nitrogen transport; the geometry reduces pressure loss, lowering pumping requirements and heat generation – factors that directly affect efficiency and operating cost in superconducting systems.

Enoflex currently supplies pipe to leading cable companies and superconductors. Rather than underwriting full cable system risk, Enoflex is anchored in a critical enabling component. If superconducting transmission scales in constrained urban nodes, the cryogenic layer must scale with it.

PureVolt: Modular, Grid-Optional Power
PureVolt has evolved from a grid-connected concept to a grid-agnostic platform built around modular ~500 MW trains. Initial phases deploy smaller turbines for early capacity delivery, and as larger units come online, early turbines transition into backup roles. This sequencing provides the fast time-to-power that hyperscalers require while building toward scale, achieving at least Tier III reliability without full grid support. In markets where interconnection timelines are uncertain, that optionality carries real strategic value for our hyperscaler partners.

Conclusion
The North American grid has led to significant economic losses tied to grid failures from lack of resiliency, and infrastructure that is operating beyond its designated service life. AI-driven load growth has further exposed constraints in transmission, interconnection, and reserve planning, and the response is multi-layered: digital optimization, incremental upgrades, high-density transmission, and parallel behind-the-meter solutions.

The opportunity lies not only in generation but in integration – components and architectures that increase density, flexibility, and speed to power. Azimuth is positioned within that layer through Enoflex’s superconducting enabling infrastructure and PureVolt’s modular firm power platform. In a system built for linear growth, nonlinear demand favors those who solve bottlenecks rather than those who assume the grid will expand on schedule.