Question: AI power infrastructure beyond generation grid buildout Verdict: fold-into-ai-power-bottleneck

Chain map — physics first

Price truth: all price / RSI / trend figures below are read from the per-watchlist scan summaries at summaries (grid-buildout, ai-power, monster-discoveries), as-of 2026-06-26 close (~5 trading days stale). Code computes the math; nothing is derived in prose.

The physical path from a power plant to a data-center rack, layer by layer. This trace covers the segment between the generator and the building fence — the piece neither active perspective owns as a named layer.

Layer G0 — Prime movers / on-site generation (where electrons originate)

• BE Bloom fuel cells [tracked, ai-power], GEV GE Vernova gas turbines [tracked, key_ticker], CAT Caterpillar/Solar gensets [tracked, ai-power watch-name], CMI Cummins standby gensets [gap — only in monster-discoveries]. IPPs CEG/VST/TLN [tracked]. — This layer is fully owned by ai-power-bottleneck.

Layer G1 — Step-up / step-down transformers (the real global shortage)

• GEV [tracked], ETN Eaton [tracked], HUBB Hubbell [tracked] — all diversified, transformers are one line.
• HPS-A.TO Hammond Power [tracked, foreign CA], 298040.KS Hyundai Electric [tracked, foreign KR].
• Siemens Energy SMNEY [gap/foreign ADR], Hitachi Energy [foreign, inside Hitachi], Prolec-Xignux [private], Virginia Transformer [private].
• Finding: there is no clean US-listed pure-play transformer maker. US access is diversified-incumbent (GEV/ETN/HUBB, already held) or a gated foreign ADR (SMNEY). The "shortage" is real but not investable as a fresh clean US name.

Layer G2 — HV/MV cable (physically the tightest global chokepoint)

• Prysmian PRYMY [gap/foreign ADR, global #1], Nexans [foreign], NKT [foreign], LS Cable [foreign KR], Southwire [private].
• US-listed exposure is indirect only via ETN/NVT connectors. No US pure-play exists at all — this layer is structurally foreign/private. (Churchill's HVDC cable-ship bottleneck note in ai-power-delivery's 06-16 refresh points at the same constraint from the marine-install side.)

Layer G3 — Switchgear / substations

• ETN [tracked], HUBB [tracked], POWL Powell [tracked, ai-power + grid-buildout — the cleanest near-pure-play, already held], NVT nVent [tracked, ai-power].
• ABB ABBNY, Schneider SBGSY [both foreign ADRs, not tracked].
• This layer is already well-covered.

Layer G4 — Grid EPC / build-and-connect contractors (where the fresh clean-US gaps actually are)

• PWR Quanta [tracked, grid-buildout + ai-infra], MYRG MYR Group [tracked, grid-buildout].
• MTZ MasTec [gap — only monster-discoveries], EME EMCOR [partially tracked, ai-infrastructure], PRIM Primoris [gap — untracked anywhere].

Investable picks

Prices as of 2026-06-26 summaries (5 days stale). Trend-hold lens = durable secular uptrend, buy flags/reclaims not knives.

Already-tracked grid names for context: PWR $687.87 / RSI 46.2 / weak-down; MYRG $477.95 / RSI 60 / strong-up; NVT $162.92 / RSI 47.2 / strong-up; POWL $279.77 / RSI 46.9 / weak-down.

Verdict on picks: the only genuinely-fresh clean US-listed grid names are grid-EPC contractors (MTZ, PRIM, EME). Transformers and cables — the layers with the loudest "shortage" narrative — have no clean US pure-play; the honest arms-dealer expression there is the already-held diversified incumbents (GEV/ETN/HUBB) or gated foreign ADRs. MTZ is the single cleanest trend-hold-shaped addition.

Reconciliation

vs grid-buildout watchlist [HPS-A.TO, 298040.KS, GEV, ETN, HUBB, POWL, MYRG, PWR]: already spans transformers (GEV/ETN/HUBB + foreign HPS-A/Hyundai), switchgear (POWL), and EPC (MYRG/PWR). Missing from it: the EPC bench (MTZ, PRIM, EME) and any cable exposure (PRYMY).

vs ai-power-bottleneck: This perspective already owns G0 (generation, gas turbines, IPP) and explicitly flags the T&D layer as unfinished — its 2026-06-18 DPA Title III log entry says the grid cohort "T&D-equipment layer (between generation and in-the-box delivery) is currently under-perspectived — candidate sub-layer," and already points at the grid-buildout watchlist. So the grid arm is a self-identified hole in this perspective, not a new theme.

vs ai-power-delivery: Its "Layer 1 — Grid" is about HVDC + solid-state-transformer semiconductor disruption (the SiC/GaN incumbent-bear angle against Eaton/Schneider) — i.e., the silicon inside the conversion equipment, not the transformer/cable/substation hardware OEMs and the contractors who install them. Delivery is the in-the-box 30%-parasitic-loss chain (VRM/BSPDN/packaging). The physical T&D buildout hardware is NOT in the delivery cohort.

What is genuinely NEW: the physical grid T&D buildout sub-layer — transformer/cable/substation OEM shortage + the grid-EPC contractors (G1–G4 above) — as a named layer with its named players. What is already-owned: GEV, ETN, HUBB, POWL, NVT, PWR, MYRG, CAT (all tracked); on-site generation (BE/GEV/CAT).

Verdict + reasoning

FOLD INTO ai-power-bottleneck (as the "Grid T&D Buildout" sub-layer). This is the default and it is correct:

1. The generation-side perspective already owns everything upstream (prime movers, turbines, IPP) and downstream-of-fence hands off to ai-power-delivery. The grid T&D hardware sits exactly in between and the perspective itself flagged it as an under-perspectived candidate sub-layer on 2026-06-18.
2. The grid-buildout watchlist already exists and is already referenced by ai-power-bottleneck as its transmission-equipment tracker. No new watchlist, no new perspective needed — just promote the flagged sub-layer to a named section and close the EPC/cable coverage gaps.
3. Not delivery: delivery is silicon/packaging (the 30% in-box loss). Grid hardware is a different physical layer with different (mostly diversified-industrial and foreign/private) players.
4. YAGNI on a new perspective: the theme is the same watts-per-rack bottleneck; a third power perspective would fragment one thesis into three overlapping READMEs.

Proposed mutations / follow-ups

Watchlist adds — clean US tickers only (grid-buildout):

• watchlist:add grid-buildout MTZ — MasTec, grid-EPC gap, trend-hold-shaped (strong-up, golden cross, -10% off high).
• watchlist:add grid-buildout PRIM — Primoris, untracked grid/power-delivery EPC. Gated on a price fetch first (no summary on disk).
• watchlist:add grid-buildout EME — EMCOR, cross-add from ai-infrastructure for grid-EPC cohort completeness (already tracked, low-cost cross-tag).
• On-site generation gap → ai-power not grid-buildout: watchlist:add ai-power CMI — Cummins standby gensets, sibling of BE/CAT at the G0 layer (diversified watch-name, not a key_ticker).

Foreign / ADR coverage decisions (GATED — do not auto-add):

• PRYMY (Prysmian ADR) — the only listed access to the HV-cable chokepoint layer, which has zero US pure-play. Highest-value foreign gap.
• SMNEY (Siemens Energy ADR) — transformer/grid-OEM pure-play the US market lacks.
• Note-only (foreign primaries, no clean US access): Nexans, NKT, LS Cable, ABB (ABBNY), Schneider (SBGSY), Hitachi Energy. Private: Prolec-Xignux, Virginia Transformer, Southwire.

Perspective README edit (ai-power-bottleneck): Promote the 2026-06-18 "under-perspectived candidate sub-layer" note into a named "Grid T&D Buildout (G1–G4)" section using the chain map above — transformer shortage (no clean US pure-play; GEV/ETN/HUBB diversified access), cable (foreign/private only, PRYMY the gated ADR), switchgear (POWL/NVT/HUBB/ETN, covered), EPC (PWR/MYRG tracked + MTZ/PRIM/EME added). Cross-link grid-buildout watchlist as the tracker.

Deep-dive follow-ups:

• PRIM — fetch price/fundamentals (no data on disk), then technical read; the one untracked clean-US name we can't currently judge.
• MTZ — verify current technical state before entry (data is 5 days stale; RSI 56 mid-range, confirm it's holding a rising SMA20 not rolling over).
• Transformer-shortage thesis: confirm whether GEV/ETN/HUBB disclose data-center-attributable T&D revenue % before treating them as the "grid shortage" expression vs. their existing generation/delivery tags — feasibility-gate: their segment reporting may not break this out.

Sources

Price, RSI and trend figures read from the desk's validated daily scan summaries (summaries). No number in this note was computed in prose.