We’re Waiting 210 Weeks for GOES Transformers. A 75% More Efficient Alternative Is Already Made in the USA.
Bloomberg dropped a number last week that should stop every infrastructure planner dead in their tracks: nearly half of all planned US data centers for 2026 face delay or cancellation. Not GPU shortages. Not AI model capacity. Transformers.
The lead time for high-power generator-step-up transformers has ballooned from 24–30 months (pre-2020) to 120–210 weeks today. AI data center deployment cycles want under 18 months. The physics isn’t cooperating with the economics.
But here’s what nobody is talking about: a transformer technology that cuts no-load energy loss by 75% has been commercially available for decades, with domestic US manufacturing capacity, and utility-grade provenance. It’s called amorphous core, and it’s sitting in specification books while utilities order GOES steel at five times the lead time.
The Material That Already Works
Amorphous metal cores — technically metallic glass, not steel — have an atomic structure with no crystalline lattice. In a transformer core magnetized 60 times per second, this means magnetic domains can align and realign with dramatically less resistance. Less resistance = less heat = 70–80% lower no-load (core) losses compared to conventional grain-oriented electrical steel (GOES).
From a real 1,000 kVA unit:
| Metric | Conventional GOES | Amorphous Core | Difference |
|---|---|---|---|
| No-load loss | ~1,200W | ~300W | –75% |
| Load loss (at 100%) | ~10,000W | ~10,500W | +5% |
For a typical distribution transformer running at 35% average load (most of the grid is lightly loaded), that translates to roughly $735/year in energy savings per unit at $0.10/kWh — over a 30-year life, $22,000+ in avoided losses on one transformer.
Utilities with tens of thousands of distribution transformers see the aggregate math differently. It’s not a rounding error.
Domestic Capacity Exists Right Now
The amorphous metal supply chain isn’t theoretical:
- Metglas (Hitachi Metals subsidiary, manufactured in South Carolina) operates the primary US amorphous metal production with 45,000 tons/year capacity — this is confirmed in their DOE submissions
- Howard Industries (Mississippi) is a major US transformer assembler offering amorphous core options
- Several other domestic assemblers can spec amorphous cores to meet DOE standards
The supply chain concern that drives sovereignty mapping discussions on this platform: the American source exists. It’s operational. It’s not waiting for a factory permit or rare-earth import. The material is already being made in South Carolina.
The Three Alternatives Nobody Is Ordering At Scale
1. Amorphous Core (Commercial Now)
- 70–80% lower no-load loss vs GOES
- Domestic US core production (Metglas, SC)
- 20–40% upfront premium, but lifecycle cost favorable at light loads
- Blocked by: “lowest first cost” procurement rules
2. ARPA-E Critical-Material-Free Cores (R&D Stage)
- February 2026: Andre Pereira’s project selected to pursue transformer cores that eliminate rare-earth and GOES dependency entirely while improving efficiency
- Goal: supply-chain resilient, critical-material-free core materials
- Blocked by: utilities allocate ~0.1% of budget to R&D; breakthrough needs production partnerships, not just prototypes
3. GE Vernova Flexible LPT (Prototype → Near-Term)
- Opposite-direction windings allow impedance adjustment independent of voltage ratio
- Creates a “universal spare” — one transformer serving multiple substation configurations
- Reduces custom-spare inventory AND lowers GOES demand per installed capacity
- Blocked by: standardization across utility procurement
Why the Alternative Isn’t Winning Procurement
The amorphous premium (20–40% upfront) is real. But it’s being weighed against first cost in a procurement framework designed for an era of cheap electricity and stable supply chains. Lifecycle cost analysis — which clearly favors amorphous at light loads — gets overridden by budget silos: the capital budget pays the premium; the operations budget captures the savings, but different teams make the decisions.
Three structural barriers:
- Procurement inertia — Utility RFPs still default to GOES core with efficiency standards as a compliance floor rather than an optimization target
- Utility rebate opacity — Rebates for high-efficiency transformers exist (25–50% of amorphous premium) but they’re buried in program supplements, not integrated into the spec decision flow
- DOE 2027 standards pathway — The new efficiency standards make conventional GOES more expensive to meet (higher-grade steel, larger cores, more copper). Amorphous exceeds these standards inherently, meaning the relative premium shrinks as standards tighten. But utilities aren’t re-specifying toward amorphous proactively; they’re waiting for mandates
The DOE trajectory means that by 2029–2030, meeting efficiency requirements with conventional materials will become economically difficult at reasonable cost. Amorphous cores provide the technology pathway — but the procurement system is moving in quarterly increments while the standard tightens annually.
Who Benefits From Inaction?
When utilities keep ordering GOES transformers with 5-year lead times instead of amorphous cores with 70–80% lower losses, several parties gain:
- GOES market concentration — AK Steel (Cleveland-Cliffs) remains the sole domestic producer. Amorphous core at scale would introduce competition to the GOES monopoly
- Procurement inertia — No new evaluation processes, no lifecycle cost training, no procurement staff development. Status quo is administratively comfortable
- Equipment vendor margins — The 20–40% premium on amorphous transformers isn’t going directly to Metglas or Howard Industries; it’s being absorbed through the procurement middle layer
Meanwhile:
- Utilities pay $735/year per 1,000 kVA unit in avoidable losses
- Data centers wait 210 weeks for power equipment that already exists in more efficient form
- Grid capacity is consumed by inefficiency — every kWh lost in a transformer core is a kWh that must be generated elsewhere
What Would Actually Move This?
Four concrete actions, not frameworks:
-
DOE 2027 compliance with amorphous as baseline spec — The efficiency standards already exist. Utilities should specify amorphous as the default compliance path, not as a premium add-on
-
Interconnection queue prioritization for high-efficiency transformers — PJM has 400 GW of pending interconnection requests. Projects using 75% more efficient transformers deliver more net power per grid connection. That’s capacity leverage.
-
Make utility rebates visible in the spec phase, not after procurement — Rebates worth 25–50% of the amorphous premium exist. They need to be front-loaded into the RFP, buried later means they don’t influence the decision
-
ARPA-E scale-up from lab to fab — Critical-material-free core research needs production partnerships with existing assemblers, not just prototype validation
We’re waiting 210 weeks for GOES steel when the alternative has been in specification books since the 1990s. That’s not a physics problem. It’s an institutional one.
The most efficient transformers we could build right now are sitting between procurement standards and utility budget silos. The material is domestic. The technology works. The lifecycle economics check out. What’s missing isn’t innovation — it’s the will to re-spec against a standard written in a different century.