Hyperscaler Utility Capex 2026: Three Q1 Strategies

AEP CEO Bill Fehrman said the quiet part out loud last week. With 63 GW of new large load contracted by 2030 (about 90 percent data centers) and a five-year capital plan now sitting at $78B, he is reviewing whether AEP’s utilities should remain in PJM and SPP at all. That single Q1 2026 hyperscaler utility capex disclosure reframes how every distribution buyer should read the next RFQ cycle. The complaint is interconnection speed. The implied threat is a multi-decade transmission, generator step-up, and SCADA buildout outside the RTO planning regime.

It was one of four hyperscaler utility capex announcements that landed inside the same 48-hour window. Stack them together and you can see three different theories of how to absorb hyperscale load, plus one cross-cutting reliability mandate that hits every Transmission Owner regardless of strategy. Distribution buyers should decide which of these models their utility will follow before the next RFQ cycle starts.

What the Q1 2026 numbers actually say

The four disclosures are not directly comparable. AEP includes ERCOT (41 GW of the 63), Duke is mostly Carolinas, Entergy is gas-anchored MISO South, and Xcel is a four-state IOU. Net adds and pace tell more than headline totals.

Utility	Contracted hyperscaler load	Capex plan	Q1 2026 net add	Strategy signal
AEP	63 GW by 2030 (90 percent data centers)	$78B (5-yr)	+$6B capex	Threaten PJM/SPP exit, build outside RTO
Entergy	5.2 GW Meta plus 7-12 GW pipeline	$57B (4-yr)	+$14B (33 percent)	Seven new NGCCs, gas-anchored
Duke	7.6 GW total, 5 GW under construction	$103B (5-yr)	+2.7 GW contracts	Steady pace, plan unchanged
Xcel	2 GW under construction, 6 GW target by EOY 2027	(4-state portfolio)	Google deal as template	Productized large-load tariff

The composite read is that hyperscalers are signing whatever generation mix the local utility offers. Procurement timeline beats carbon profile. That has direct consequences for what equipment ships and when.

Three competing strategies

1. AEP: scale by leaving the RTO

The 63 GW figure is the largest single-IOU hyperscaler-load disclosure on record. Geographic split is 41 GW in ERCOT, 16 GW in PJM, 6 GW in SPP. The threat to leave PJM and SPP is procurement-relevant for one reason: even if AEP stays under what Fehrman called “alternative structures,” the threat alone forces both RTOs into queue redesign. Every utility whose generation is queue-stuck behind AEP’s data centers gets pulled along.

If AEP actually departs, it must build replacement transmission and balancing capacity. That is GSU transformers in the 800-plus MVA range, 230 kV and 500 kV breakers, GIS substations at injection points, and a multi-year reconfiguration of every flowgate inside the affected zones. Hitachi Energy, GE Prolec, SPX Transformer Solutions, and Pennsylvania Transformer are the relevant lanes for primary HV iron. Lead times for greater than 500 MVA GSUs sit at 36-48 months as of Q2 2026.

Co-ops and munis sharing AEP service territory across Ohio, Indiana, Michigan, Kentucky, West Virginia, Virginia, and Oklahoma should expect coordination demands and adjacency procurement opportunities on the distribution feeders that connect to data center campuses. The same shop that bids customer-funded substations for an AEP hyperscaler colocation will quote your padmount order three weeks later.

2. Entergy: gas-anchored at the customer

Entergy raised its four-year capital plan by $14B in a single quarter to $57B, anchored by a Meta data center in North Louisiana. The build is seven new natural-gas combined-cycle units totaling more than 5.2 GW, plus 1+ GW of additional industrial load behind the Meta anchor.

Per-unit math: average 750 MW NGCC needs one 800-900 MVA GSU plus auxiliary transformers, switchyard breakers, instrument transformers, and HV interconnect at 230 kV or 500 kV. Ballpark per-unit primary equipment spend is $30M to $80M. Seven units puts $200M to $560M in primary equipment alone, before BOP transformers, MV switchgear, and the long-distance transmission that delivers power to the campus.

Entergy’s strategy survives even if hyperscaler load forecasts get cut by 30 percent, because the underlying industrial demand from Air Liquide, ExxonMobil, and similar legacy customers in the LNG export corridor stays put. That makes it the most resilient procurement target of the four. It also creates rate-case tension on stranded-asset risk by 2027-2030, which Mississippi, Arkansas, and Louisiana co-ops should price into their multi-year plans now.

3. Xcel: productize the tariff, customer pays

The Google deal in Minnesota is a 1 GW data center anchoring a Clean Energy Accelerator Charge structure: 1,400 MW wind plus 200 MW solar plus 300 MW long-duration storage. CEO Bob Frenzel called it the template. Xcel is replicating it in Colorado (already filed), Texas, New Mexico, and Wisconsin.

Tariff terms: data center customer covers full cost of infrastructure, long-term contract with termination fees, credit requirements, incremental cost tests. No socialization to residential ratepayers. From a procurement angle this means standardized substation designs across five states, customer-funded substations using a repeatable spec set, and equipment OEMs sourcing from a hyperscaler procurement spec rather than a Xcel utility spec when the customer pays.

Power transformer demand from each 1 GW campus runs three to five units in the 60-100 MVA range stepping from 230 or 138 kV to MV, plus extensive HV switchgear (likely GIS in the Minneapolis-area sites where space is constrained). Vendors in the lane include Hitachi Energy, Siemens Energy, GE Prolec for power transformers, and ABB, Eaton, Powell, and Schneider Electric for switchgear.

For a municipal utility serving a hyperscaler colocation, the Xcel template is the cleanest model for cost-of-service ratemaking. The credit and termination-fee provisions in particular are the part to copy.

The Duke baseline: steady wins

Duke’s 7.6 GW total with 5 GW under construction is the most operationally advanced of the four. The $103B 2026-2030 capital plan is unchanged from the 2025 disclosure, which tells you the data-center growth was already baked in. Duke is not raising capex; it is executing.

Net add of 2.7 GW in Q1 alone signals momentum. Maintained pace puts Duke past 10 GW by year-end 2026. Suppliers should expect Duke purchasing to be steady but elevated through 2030, with concentrated demand in the Charlotte and Triangle hyperscaler corridors. Historic Duke vendors include Hitachi Energy, Eaton (Cooper Power), Howard Industries, ERMCO, Siemens Energy, GE Prolec, and ABB.

The under-discussed angle: each 250 MW campus pulls 4-6 substation-level transformers in the 60-100 MVA range plus dozens of MV padmounts on the campus side, plus a full distribution-system upgrade in surrounding circuits. That is the part co-ops adjacent to Duke service territory in the Carolinas can position for in adjacency procurement.

The cross-cutting mandate: NERC Level 3 alert

NERC’s May 4 Level 3 alert applies to every Transmission Planner, Planning Coordinator, and Transmission Owner serving computational loads at the 20 MW threshold. That captures essentially every hyperscaler interconnection at all four utilities above. Seven actions are mandatory. The acknowledgment deadline was May 11. Full responses are due August 3, 2026.

Action 6 is the procurement event. NERC mandates installation of dynamic fault recorders (DFRs) and PMU monitoring at substations serving computational loads. SEL, GE Vernova, Qualitrol, and Hitachi Energy are the primary lanes. Expect a procurement spike at hyperscale-adjacent substations through Q3 2026.

Action 3 will trigger re-evaluation of every protection scheme on circuits feeding 20 MW or higher computational loads. Where existing relays cannot meet the new study findings, replacements follow. SEL, ABB, Hitachi Energy, and GE again.

Smaller utilities and co-ops sitting in the same footprint as a hyperscaler colocation are caught in the same mandate without the engineering bench depth to absorb it. The NERC alert is going to drive consulting and pre-engineered commissioning packages through end of 2026 for any TO that has not commissioned a hyperscale load before.

The bigger trajectory is Fault Ride-Through and protection coordination standards on the load side, with FERC submission targeted for end of 2026 and standards likely effective 2027-2028. That is the part procurement teams should plan for in the next two RFP cycles.

What this means for distribution buyers

If your service territory is adjacent to a hyperscaler colocation under any of the four IOUs above, four practical takeaways:

The OEM short list is the same across all four strategies. Hitachi Energy, Siemens Energy, GE Prolec, and Eaton appear in every footprint. Capacity competition is real. Frame contracts signed in 2026-2027 will lock in supplier relationships for the 2028-2037 build cycle.

Tariff structure determines procurement spec ownership. Xcel-style customer-funded substations follow hyperscaler spec, not utility spec. One-off deals like AEP and Entergy follow utility spec. The vendor short list narrows when the customer pays.

NERC Level 3 compliance is non-negotiable on the August 3 deadline. DFR/PMU installation is the immediate equipment ask. Protection-scheme replacements follow over the subsequent 12-18 months.

The 2027-2028 reliability standards on Fault Ride-Through are coming. Specifying for them now in long-lead RFQs is cheaper than retrofitting under compliance pressure.

The full breakdown of which OEMs to short-list by region, lead-time risk by equipment class, and tariff-specific procurement playbooks for muni and co-op buyers sits in our Q2 2026 hyperscaler procurement intelligence brief. If you are sitting on a frame contract decision this quarter, that is the document to read first.

Related reading

• The $1.4 Trillion Utility Capex Wave: What It Means for Transformer Procurement
• Data Center Demand: 50 GW of New Grid Infrastructure Procurement
• FERC Large Load Interconnection Rules: What Changes for Equipment Procurement
• PJM Interconnection Queue: 220 GW, 800 Projects, Equipment Wave
• MISO Data Center Power Demand and Infrastructure Procurement