⚡ THIS WEEK'S SIGNAL
I spent the week inside the transformer data, and one thing reframed the whole interconnection debate for me: the queue everyone argues about is no longer where the time goes. By PJM's own count, a project spends three-plus years reaching an interconnection agreement — and another four years waiting to actually come online after that approval. We've been watching the front door. The jam moved to the loading dock.
━━━━━━━━━━━━━━━━━━━━━━━
📰 MAIN STORY
For years the story was the queue. Get through the studies, sign the agreement, energize. That story is now half the picture. Fresh Wood Mackenzie data, reported last week, puts substation transformer lead times at roughly 140 weeks in 2023, about 150 in 2025, and north of 160 weeks in 2026 — and the binding constraint has quietly migrated from getting approved to getting equipped.
Watch what each player is actually optimizing for. Developers want time to market, and they'll pre-buy delivery slots years out to protect it. Wood Mackenzie's analyst names the real chokepoint plainly — it's a combination of power limits and equipment constraints, not the queue alone. PJM, for its part, is drawing a hard boundary around its own reputation: its spokesperson argues that a project is done with PJM the moment it signs the interconnection service agreement, and the years that follow shouldn't be charged to the queue. Everyone is pointing at the part of the timeline that isn't theirs.
Here's the contrast that does the work. A hyperscaler with a balance sheet can lock manufacturing slots in advance and, failing that, go behind the meter entirely. The smaller developer, the rural co-op, the utility trying to serve everyone else — they're all standing in the same equipment line. Wood Mackenzie projects data centers could absorb up to 40% of total U.S. electrical-equipment demand by 2030, up from a low single-digit share, with that buying power concentrated in a handful of hyperscale names. The transformer that energizes a $40 billion campus and the one that backfills an aging neighborhood substation come off the same constrained line.
The numbers worth screenshotting: Wood Mackenzie clocked average transformer lead times rising from roughly 50 weeks in 2021 to about 120 in 2024, with some large substation and generator step-up units running 80 to 210 weeks. As of January 2026, PJM reported more than 21 GW of projects in engineering-procurement status and another 8.2 GW under construction, and its milestone-delay data attributed 29% of change requests to permitting, 28% to "other," and 23% to supply chain. The equipment market itself is forecast to scale from about $20 billion in 2026 to $65 billion by 2030 as U.S. data center capacity grows from roughly 24 GW to 100 GW.
Two roads from here. Scenario A: the ~$2 billion already committed to North American transformer production — Hitachi Energy, Siemens Energy, others — comes online around 2028, lead times start compressing, and the queue reforms (FERC Order 1920, the looming RM26-4 action) finally translate into faster energization. Equipment stops being the binding constraint, and the grid keeps the load. Scenario B: lead times keep climbing — Wood Mackenzie's own read is that the pressures are amplifying, not easing — and behind-the-meter generation accelerates as the rational workaround. The load that utilities planned transmission around goes off-grid, and that planned capacity becomes a stranded-asset conversation.
One thing to watch: whether FERC's end-of-June RM26-4 order touches post-agreement energization at all — because a faster front door doesn't move a transformer one week sooner.
━━━━━━━━━━━━━━━━━━━━━━━
PRESENTED BY [SPONSOR]
[Sponsor copy — one-line value proposition, two sentences of context, CTA.]
━━━━━━━━━━━━━━━━━━━━━━━
⚡ QUICK HITS
FERC's large-load order is due any day: FERC has set an end-of-June 2026 deadline to act in Docket RM26-4-000, the rulemaking governing how massive new loads — data centers included — connect to the interstate transmission system. How it draws the federal-state line on cost allocation could reset underwriting models across every ISO queue.
VoltaGrid's $1B raise builds the picks-and-shovels layer: Blackstone and Halliburton's billion-dollar investment values the behind-the-meter generation firm above $10 billion, against a 2.3 GW Oracle contract and a Propell acquisition that vertically integrates manufacturing. It's the clearest sign yet that private capital is treating off-grid power as an asset class, not a stopgap.
TMI's restart hits the same wall as everyone else: In a March 31 FERC filing (Docket ER26-2028), Constellation said restarting Three Mile Island Unit 1 needs significant transmission work before the plant is fully deliverable. Even a marquee 800-MW nuclear restart can't escape the post-approval transmission queue — the bottleneck is structural, not project-specific.
The states are writing the pledge into law: Mandatory utility fixed-charge measures and large-load cost-allocation rules are now moving across 27 states, with New York's A10141/S9144 proposing a multi-year pause on new data center construction pending ratepayer-protection rules. The voluntary White House pledge now has statutory competition with actual enforcement teeth.
Meta's 6.6 GW nuclear stack sets the template: Meta's agreements with TerraPower, Oklo, and Vistra blend existing baseload in Ohio and Pennsylvania with long-dated SMR optionality targeted for 2032. It's a multi-vendor portfolio model other hyperscalers are now racing to copy — and a reminder that nuclear timelines and AI timelines still don't line up.
━━━━━━━━━━━━━━━━━━━━━━━
🔧 TOOL / RESOURCE OF THE WEEK
LBNL "Queued Up" (Lawrence Berkeley National Laboratory): the annual primary-source dataset on U.S. interconnection-queue volume and timelines, drawn from seven ISOs/RTOs and dozens of utilities. Right now, it's the cleanest way to separate real queue-processing time from the post-agreement equipment delays this issue is about — useful before you underwrite any project timeline on a vendor's stated date. → U.S. Interconnection Queue Data Through 2025: Complete Interconnection Request Dataset and Summarized Data Workbook | LBL ETA Publications
━━━━━━━━━━━━━━━━━━━━━━━
💬 CLOSING THOUGHT
If the equipment line is the real constraint this cycle, then the queue reforms everyone's bracing for might be solving last year's problem. I'm cautiously optimistic about the 2028 manufacturing buildout — but "cautiously" is doing real work there, because committed capacity and delivered transformers are not the same thing, and the people I trust on this aren't ready to call the peak. So, I'll ask the room: has anyone actually seen a substation transformer lead times shorten this year? And if it did — what changed to make it move?
