The 2026 Grid Capex Supercycle: Wires, Transformers, and the Cost of Connection

The capex step change is real, and it is happening on the wires side #

The Edison Electric Institute Annual Capex Survey shows United States investor-owned utilities deploying roughly USD 50 billion per year of transmission and distribution capital across the early 2010s. By 2024 the same tracker registered above USD 130 billion. Member company guidance and rate case filings now point to USD 170 billion or more for 2026, with several large multi-state holding companies disclosing five-year plans that imply continued mid-teens annual growth in T and D spend through 2029. The composition has tilted decisively toward transmission, away from generation, and toward distribution hardening. Generation capex sits well below its 2008 peak in real terms. T and D capex is at an all-time high in both nominal and real terms.

The shift matters because the regulatory recovery mechanics differ. Generation in restructured markets clears at marginal cost. Transmission and distribution recover under cost-of-service, with FERC formula rates capping authorized return on equity around 10.5 percent for most regions and binding the holding company weighted ROE toward the 1.30 cap that FERC enforces on incentive adders. Utility CFOs are running into a different problem from a decade ago. The capex pipeline is not constrained by regulator willingness. It is constrained by transformer slots, conductor availability, NEPA review timelines, and the human capacity of state public utility commissions to process simultaneous rate cases at higher revenue requirements.

Demand drivers: data centers do most of the work, but not all of it #

The single largest incremental load category is hyperscale and colocation data centers. NERC's 2024 Long-Term Reliability Assessment lifted the ten-year United States peak demand forecast by roughly 15 percent versus the 2022 vintage, with most of the revision concentrated in PJM, ERCOT, MISO, and the Southeast. Cross-tabulating utility integrated resource plans, ERCOT's large flexible load study, and Dominion's PJM filings yields a working estimate of 40 to 60 gigawatts of new data center capacity online in the United States between 2024 and 2030. That sits on top of an ERCOT forecast adding another 30 gigawatts of behind-the-fence flexible loads.

EV adoption has run slower than the 2022 trajectories assumed. United States light-duty EV share of new sales sits in the high teens rather than the mid-twenties planners penciled in three years ago. Building electrification of space and water heat is underperforming. Heat pump shipments have grown but penetration of existing housing stock outside the Pacific Northwest remains in the single digits. Reshoring is partial. Battery, semiconductor, and pharmaceutical fabs add real load, but the broader manufacturing base has not yet driven a national kilowatt-hour reset. Weatherization is the other quiet driver. Texas spent more than USD 12 billion on transmission rebuilds and substation hardening after the February 2021 freeze, and Hurricane Beryl in July 2024 prompted CenterPoint's USD 5 billion Greater Houston resiliency plan. These are recovered through distribution rate cases, not transmission tariffs, which is why distribution capex is now growing faster than transmission for several Gulf and Atlantic utilities.

The interconnection queue and FERC Order 2023 #

Active interconnection queues across MISO, PJM, CAISO, SPP, ERCOT, NYISO, and ISO-NE held more than 2,600 gigawatts of generation and storage requests at the end of 2024, against roughly 1,300 gigawatts of installed United States capacity. Average study-to-commercial-operation time has stretched beyond five years in MISO, PJM, and CAISO. Withdrawal rates exceed 70 percent for projects that entered queues before 2020, reflecting the real economics of network upgrade cost assignment under the prior serial study process. FERC Order 2023, issued in July 2023 and now in implementation across regions, replaces serial review with a cluster study process, imposes withdrawal penalties to deter speculative submissions, and mandates readiness deposits scaled to project size.

Implementation has been uneven. MISO's first compliant cycle cleared in 2024 with a sharply reduced queue and a tightening of the network upgrade socialization formula. PJM filed a transition cycle that effectively paused new study windows through 2025 and prioritized a backlog of 230 gigawatts. CAISO, contending with both transmission scarcity in Southern California and Path 26 constraints, has had to layer Order 2023 on top of its zonal capacity reform. The practical outlook for 2026 is that interconnection cost shares for new projects remain at multiples of pre-2020 norms. A typical PJM solar plus storage project that would have paid USD 50 to 75 per kilowatt of network upgrades in 2018 now faces USD 150 to 300 per kilowatt, occasionally more, because the marginal upgrade now triggers backbone reinforcement that the developer must underwrite up front.

Equipment supply: the transformer backlog is the hard constraint #

Large power transformers, defined as units rated 100 megavolt-amperes and above, have moved from a commodity procurement item to a four to six year scheduled order with milestone deposits. The four global incumbents, Siemens Energy, GE Vernova, Hitachi Energy, and Hyundai Electric, together with Korean second-tier suppliers including HD Hyundai Electric's smaller affiliates, run a combined order book whose 2026 production slots have been sold since mid-2024. Spot price ranges have moved from USD 1.5 million to USD 2 million per unit pre-pandemic to USD 2.5 million to USD 3 million for utility-grade single-phase units, with high-voltage direct current valve hall transformers commanding multiples of that.

Conductor and steel pass-through compounds the issue. Aluminum conductor steel reinforced cable, the workhorse of high-voltage overhead lines, follows London Metal Exchange aluminum and grain-oriented electrical steel for transformer cores. Grain-oriented electrical steel capacity is concentrated in fewer than ten producers globally, dominated by Nippon Steel, JFE, Cleveland-Cliffs, Baowu, and a handful of European mills. Lead times for utility-grade GOES coil have run nine to twelve months on top of transformer fabrication. United States utilities have responded by pooling spares through the Edison Electric Institute Spare Transformer Equipment Program and through bilateral mutual assistance, but the program covers contingency replacement, not capex growth.

The capacity response is real but slow. Hitachi Energy is investing more than USD 6 billion across global transformer and HVDC capacity through 2027. GE Vernova has expanded Pittsburgh and South Carolina lines. Siemens Energy's grid technologies division is scaling Charlotte. Hyundai Electric is tripling its Alabama transformer plant. The combined nameplate increase, on Wood Mackenzie's tracking, lifts global large power transformer output by roughly 35 percent by 2028 against 2023, against demand growth that several BloombergNEF Grid scenarios put at 50 percent or more over the same window. The supply gap closes after 2028 in the base case, not before.

Global parallels: TYNDP, China State Grid, India, and the HVDC pipeline #

The European Union's Ten-Year Network Development Plan 2024, published by ENTSO-E, identifies EUR 600 billion of cross-border and reinforcing transmission investment across 2024 to 2034. The plan is anchored on offshore wind integration in the North Sea and Baltic, north-to-south continental flows, and the Iberian peninsula's interconnection deficit. The Northern Lights interconnector class of projects, including NeuConnect between the United Kingdom and Germany, the LionLink multi-purpose interconnector, and EuroAsia between Greece, Cyprus, and Israel, illustrate the multi-purpose offshore pattern: hybrid offshore platforms that simultaneously evacuate wind generation and trade between national markets. Each of these projects carries multi-billion-euro budgets and multi-year HVDC equipment lock-ins.

China State Grid Corporation announced a record RMB 600 billion capital plan for 2024, with SGCC's 2025 to 2030 envelope tracking near RMB 700 billion annually under National Energy Administration guidance. The Chinese ultra-high-voltage program, including the eight-vertical eight-horizontal AC backbone and the parallel UHVDC corridors moving Sichuan hydro and Xinjiang wind to coastal load centers, is the world's largest single grid investment program in absolute terms and remains the principal source of HVDC manufacturing learning curve. India's Power Grid Corporation is executing a roughly USD 16 billion five-year capex plan to evacuate 500 gigawatts of renewable target capacity, with green corridor phases two and three driving 2026 and 2027 spend.

Globally, the HVDC project pipeline through 2030 has cleared USD 300 billion on BloombergNEF and Hitachi Energy market tracking, more than double the cumulative installed HVDC base of the prior thirty years. The pipeline is concentrated in offshore wind connection, long-distance renewable export, and asynchronous interconnection. The IEA's World Energy Investment 2024 puts global grid investment at USD 400 billion in 2024 against an implied requirement closer to USD 600 billion by 2030 if Paris-aligned electrification is to be physically feasible. The financing gap is real, but the equipment constraint binds first.

Grid edge: smart meters, DERMS, and the distribution hardening case #

Distribution capex has historically been the quieter cousin of transmission. That is changing. AMI 2.0 rollouts, distribution automation, advanced distribution management systems, and distributed energy resource management systems together represent the fastest growing category inside utility capex plans. United States smart meter penetration has cleared 75 percent of meter points, and the next investment wave is grid edge intelligence: feeder-level state estimation, dynamic hosting capacity calculation, and active management of behind-the-meter solar plus storage. Itron, Landis+Gyr, Schneider Electric, and Siemens Grid Software are the principal vendors.

DERMS deployment has accelerated in California, Hawaii, New York, and parts of Texas where rooftop solar and EV penetration are already material. The economic logic is straightforward. Reconductoring a 12.47 kilovolt feeder to absorb additional distributed generation costs USD 250,000 to USD 800,000 per mile. A DERMS-enabled curtailment scheme, layered on top of inverter-based distributed resources, can defer that reconductoring by three to seven years at a fraction of the capital cost. State commissions are beginning to allow utilities to capitalize software platforms alongside hardware, which materially improves the regulated return economics. The risk for the grid edge thesis is double counting. Some hosting capacity gains attributed to DERMS investments depend on deeper transformer and conductor upgrades that the same plans defer.

Implications for utilities, equipment OEMs, developers, and offtakers #

For utility CFOs and treasurers, the binding constraints are equity issuance windows, holding company credit ratings under elevated capex, and the speed at which state commissions process trackers, formula rates, and forward test years. Promethean assumes the share of plans approved through ATC formula rate or equivalent forward-looking constructs continues to grow, because historical test year cases cannot keep up with the run rate of T and D spend. Expect more multi-year rate plans, more rider mechanisms for resilience and large load, and explicit cost allocation fights with hyperscaler tenants reshaping system peaks.

For equipment OEMs, the transformer and HVDC slot premium will persist into 2028. The strategic question is how much of the implied cash flow surge to convert into permanent capacity versus leave as cyclical margin. Hitachi Energy and Siemens Energy have signaled the former. GE Vernova and Hyundai Electric are pursuing a more measured stance. For developers and IPPs, queue position carries option value rivaling project economics. Strategos models that for representative interconnection-cleared projects in PJM and MISO, the premium for a position in front of cluster study can exceed the levelized cost difference between solar and gas peaking. For corporate offtakers, hyperscalers and electrified industrials, a signed power purchase agreement is no longer sufficient. Physical delivery requires queue position, transformer reservation, and substation site control, and the latter two are binding in 2026 and 2027. Argus tracks 14 utility holding companies, eight equipment OEMs, and the principal queue cycles each quarter.

Sources #

• EEI Annual Property Tax and Capex Survey
• FERC Order 2023 interconnection process reform
• ENTSO-E Ten Year Network Development Plan 2024
• IEA World Energy Investment 2024
• BloombergNEF Power Grid Long Term Outlook
• Wood Mackenzie Grid Edge and Transmission services
• S and P Global Market Intelligence US utilities capex
• NERC 2024 Long Term Reliability Assessment
• China NDRC and National Energy Administration grid plans
• Reuters utility and equipment supplier coverage