This replaces the prior 5-name semiconductor-only portfolio summary. The cohort formally extends to the chip-to-grid value chain — semiconductors plus AI power infrastructure. The cohort architecture decisions, concentration rules, and open questions below reflect the completed C-COHERENCE pass (2026-05-04) and all 9 deep-dive theses.
§ 01Section 1 — Executive Summary
The cohort expanded from 5 to 9 names on 2026-05-04 by executing the deferred next step identified in prior portfolio-summary Open Question §4: running Vertiv, Eaton, Navitas, and Texas Instruments as full deep-dives. The expansion is not a bolt-on. It is a structural decision to hold the cohort's analytical frame at the chip-to-grid level — the continuous voltage path from the substation to the transistor — rather than at the semiconductor level alone. Two of the four new names (ETN and VRT) are not semiconductor companies. They sit in this cohort because the dollar-weighted center of mass of the AI-DC build sits at G1-G4 (electrical infrastructure) not at L8b/L8c (power semis): ETN and VRT each capture $700K–$2.8M per MW of addressable content versus roughly 20-60x less for the power-semi BOM layer. This is the load-bearing reason for the scope expansion, not a thematic extension.
The cohort is architecturally split into two independent layers: chip-layer names (NVTS, TXN at L8b/L8c) and grid/system-layer names (ETN, VRT at G1-G4). These are held as independent positions. Box-builders are vendor-agnostic on power-semi supplier by commercial design — ETN customer-analyst confirmed this explicitly (per refinement-log Finding 12): Eaton purchases IGBTs/GaN/SiC on a competitively sourced basis and does not name power-semi suppliers. The chip-to-grid frame explains why these names are researched together, not because their P&Ls mechanically chain. A strong ETN order book does not automatically translate into NVTS or TXN revenue. This distinction must be preserved in every sizing and pair-trade decision.
The headline conviction call entering May 2026 is a net long-bias book of 7 longs versus 2 shorts, with conviction anchored at the silicon tier (TSM at 5/5, NVDA and AVGO at 4/5) and progressively scaling down toward the power-infrastructure completion tier (ETN and VRT at 3/5) and the probe tier (NVTS at 2/5). The two shorts (WOLF at 4/5, INTC at 3/5) hedge distinct structural risks — SiC substrate share collapse and foundry-leadership binary — neither of which is a Taiwan-tail hedge. The cohort's Taiwan-tail profile materially improved with the expansion: TXN (~10-15% Taiwan exposure), ETN (minimal), and VRT (zero Taiwan production) provide a partial structural hedge not available in the 5-name cohort.
The single most important post-expansion finding for a PM to internalize: The expansion meaningfully changed the cohort's risk profile in two ways simultaneously. Taiwan-tail concentration is materially reduced — three of the four new names hedge it structurally. But AI-capex single-factor concentration is materially amplified — ETN (16-19% AI-DC revenue mix, growing toward 22-26% post-Mobility spin) and VRT (65-80% AI-DC mix) are both AI-capex amplifiers, and holding both alongside NVDA, TSM, and AVGO means that a sustained hyperscaler capex pause would hit the cohort from the top (accelerator revenue) and the bottom (infrastructure revenue) simultaneously. This is why the ETN+VRT combined cap rule is binding and non-negotiable: combined ETN+VRT must not exceed 1.5-2.0x the TXN NAV position. If TXN is at 3% NAV, ETN+VRT combined cannot exceed 4.5-6.0% NAV. This rule is the portfolio-construction mechanism that prevents the AI-capex amplification from becoming a single-factor bet wearing two tickers.
§ 02Section 2 — Position Sizing and Conviction Tiers
Binding Constraints (Read Before the Table)
- ETN+VRT combined cap: Combined ETN+VRT position cannot exceed 1.5-2.0x TXN NAV position size. This is a non-negotiable concentration discipline rule per refinement-log Finding 16.
- TXN-NVTS pair structure: TXN long at 2.5% NAV / NVTS long at 0.75% NAV, ratio 1:3 in TXN's favor. This pair is already finalized. Pair breaks at TXN $210.
- VRT sizing is probe-only at current prices ($330): Scale to medium only on correction to $230-250 (35-38x EV/EBITDA). Buy-on-weakness discipline is not optional — the asymmetry does not clear 2:1 at $330.
- ETN buy-on-weakness discipline: Initiate 1/3 at current levels ($424), reserve 2/3 below $380 / below $360. The asymmetric entry was at $250-290 in mid-2024 and that is gone. Full size only at $360 or below.
Position Sizing Table
| Ticker | Direction | Conviction (1-5) | Recommended Sizing (% NAV) | Cohort Architecture Role | Where to Add | Where to Trim |
|---|---|---|---|---|---|---|
| TSM | Long | 5 | 4.0-5.0% | Core | Add 30% reserve on Section 232 adverse surprise or Taiwan-posture-driven drawdown; target <$300 ADR (forward P/E ~15x) for oversized add | Trim if Samsung SF2P yields hold 70%+ for 2+ quarters AND Qualcomm/AMD announce >20% volume shift to Samsung 2nm; or if FY26 GM falls below 56% outside Arizona-ramp explanation |
| NVDA | Long | 4 | 3.5-4.5%* | Core | 40% of intended position in reserve; add on Section 232 drawdown, Q1/Q2 FY27 inventory headline, or AI Diffusion Rule surprise | Trim if gross margin compresses below 70% (absent one-time) or NVDA share of incremental hyperscaler accelerator capex falls below 55% in any 2-quarter rolling window |
| AVGO | Long | 4 | 2.0-3.0% | Core | Reserve ~30% for Section 232 / EC VMware shock creating 10-15% drawdown without changing AI ASIC fact pattern | Trim if Hock Tan's $100B 2027 AI number misses by >25% in H2 2026 print, or SBC climbs while buybacks pull back materially |
| TXN | Long | 3 | 2.5% | Core + Hedge | Add aggressively into sub-$240 reset; September 2026 MOFCOM print is the likely entry opportunity | Trim if MOFCOM imposes >20% retroactive duty; if GM stuck below 59% through FY27 despite +15% revenue growth; if industrial growth stalls below +10% YoY |
| ETN | Long | 3 | 1.5-2.0% (buy-on-weakness; initiate 1/3 now at ~$424) | Completion | Add to 2.0% on weakness below $380; full 2.0-2.5% below $360 | Trim if Electrical Americas DC-specific orders decelerate below +30% YoY for 2 consecutive quarters; or Boyd integration flags cost overruns/customer retention failure in Q3 2026 |
| VRT | Long | 3 | 1.0-1.5% (probe only at $330) | Completion | Scale to medium (2.0-2.5%) only on correction to $230-250 (35-38x EV/EBITDA) | Trim immediately if EMEA prints three consecutive negative quarters; or if VRT cooling market-share erodes 5+ ppts in quarterly CDU bookings data by 2027 |
| NVTS | Long | 2 | 0.5-0.75% (probe) | Probe | Do not add — probe only. If NVTS lands a named Kyber-class reference design at OCP/Computex 2026, re-evaluate to 1.0% | Exit if PSMC 650V qualification slips beyond Q4 2026, or GF Burlington tape-out slips beyond Q1 2027, or FY26 10-K discloses zero named AI-DC customers above 10% |
| WOLF | Short | 4 | 1.0-1.5% (medium; pair preferred) | Short | Add on CHIPS PMT headline pop (dry powder opportunity) | Cover immediately if federal equity stake or CHIPS PMT delivers in full; or if Renesas announces strategic sale at >50% premium to current |
| INTC | Short | 3 | 0.5-0.75% (small probe; pair preferred) | Short | Add 1/3 on next regulatory non-event (no federal equity announcement at next Treasury/Commerce quarterly update); add 1/3 if H2 2026 14A external customer commitment fails | Cover if two named external 14A customers commit by year-end 2026 at meaningful volume (beyond Apple entry-tier), or if Trump administration announces federal equity stake |
*NVDA: 40% of intended position held in reserve per thesis.md entry discipline.
NAV Math — What the Cohort Actually Looks Like at Sane Sizing
At base sizing (midpoint of ranges above):
| Layer | Names | Approx. NAV % |
|---|---|---|
| Core longs | TSM 4.5%, NVDA 4.0%, AVGO 2.5%, TXN 2.5% | ~13.5% |
| Completion longs | ETN 1.75%, VRT 1.25% | ~3.0% |
| Probe long | NVTS 0.65% | ~0.65% |
| Shorts | WOLF -1.25%, INTC -0.65% | ~-1.9% gross short exposure |
| Cohort net long exposure | ~15.25% NAV (net) | |
| Cohort gross long exposure | ~17.15% NAV |
ETN+VRT cap check: ETN 1.75% + VRT 1.25% = 3.0% combined. TXN at 2.5%. Ratio: 3.0/2.5 = 1.2x. This clears the 1.5-2.0x cap rule comfortably at base sizing. If TXN is scaled to 3% NAV, ETN+VRT ceiling becomes 4.5-6.0% NAV — providing room to scale ETN to 2.5% and VRT to 2.0% at the right entry points.
AI-capex amplification check: At base sizing, NVDA (4.0%), ETN (1.75%), VRT (1.25%) represent approximately 7.0% NAV directly exposed to "AI capex pause." AVGO adds another 2.5% with meaningful AI-DC exposure. Total AI-capex-sensitive long exposure is approximately 9.5% NAV at base. In an AI-capex pause scenario where all four names drawdown 30-45%, gross portfolio impact is approximately -2.9% to -4.3% NAV — manageable but not trivial. This is the quantification the ETN+VRT cap rule is designed to contain.
§ 03Section 3 — Layer Coverage Map
The chip-to-grid stack runs from G0 (power generation) through L0 (materials). The table below shows cohort coverage by layer, gaps, and the bench names that would close them.
| Layer | Layer Name | Cohort Name(s) | Coverage | Implication |
|---|---|---|---|---|
| G0 | Power generation / grid | — | Uncovered | Out of scope for this cohort by design; tracked via Crucible privates (Shatterdome, Oklo, Bloom) |
| G1 | Site / utility interface (switchgear, MV transformers, ATS) | ETN, VRT (UPS/ATS) | Covered | ETN is primary; FERC Order 1920 pipeline addressable |
| G2 | DC power backbone (UPS, BBUs, busways, distribution) | ETN, VRT | Covered | Overlapping coverage; ETN = lower-beta, VRT = purer AI-DC |
| G2T | Cooling / thermal (CDUs, cold plates, immersion) | VRT (incumbent, ~22% CDU share), ETN/Boyd (challenger, $1.7B closed Mar 2026), Ecolab/CoolIT (bench watch) | Covered — three-player, contested | Single most important competitive dynamics monitor in 2026-27; no longer VRT-unchallenged |
| G3 | 800V transition layer (SSTs, 800V PSUs, supercaps) | ETN, VRT | Covered | 800V product portfolios from both; VRT H2 2026 launch is primary catalyst |
| G4 | Rack / row power (PDUs, busbars, power shelves) | VRT | Covered (partial) | ETN has some overlap; connector layer (TE Connectivity, Amphenol) is uncovered |
| L8a | SiC (HV front-end rectification) | WOLF (short) | Covered (short) | The cautionary tale; ON is on bench as the paired long |
| L8b | GaN power semis (high-density conversion, 800V) | NVTS (density), TXN (system-BOM vertical) | Covered — two expressions of the four-way race | Innoscience (#1 globally at ~30% share) and Infineon (scale) are uncovered bench; IFX is the natural next add |
| L8c | Board-level power (VRM, BCM, eFuses) | TXN (analog/embedded power) | Covered (partial) | MPWR is the uncovered bench name at L8c; KLA-level importance understated by mention count |
| L8d | MOCVD / GaN-SiC equipment | — | Uncovered | AIXTRON and Veeco are watch items; not in cohort |
| L7 | Networking / optics (NVLink, Ethernet switch, CPO) | NVDA (NVLink, Mellanox IB), AVGO (Tomahawk) | Covered | CPO (Lightmatter, Ayar Labs) is the watch-tier disruption risk; not a cohort position |
| L6 | Accelerators / processors (GPU, ASIC, CPU) | NVDA (merchant GPU), AVGO (ASIC design partner), INTC (short; foundry pivot) | Covered | AMD is the uncovered bench name at this layer; tracked but not deep-dived |
| L5 | Advanced packaging / OSAT (CoWoS, SoIC, EMIB) | TSM (CoWoS-L), INTC (EMIB/Foveros, short) | Covered | BESI and Shibaura are the uncovered tool names; watch-tier |
| L4 | Memory (HBM, DRAM, NAND) | — None — | Under-covered — gap | SK Hynix (>50% HBM bits) is bench #2; next deep-dive cycle should close this. Memory is Bottleneck #2 in the three-bottleneck frame; it has no cohort coverage |
| L3 | Foundry / process (N3, N2, A16, 18A) | TSM (90% leading-edge), INTC (foundry pivot, short) | Covered | Samsung Foundry is the watch-tier risk; SMIC is the China parallel stack item |
| L2 | EDA (design software) | — | Uncovered | Cadence and Synopsys are watch-tier; no deep-dive planned |
| L1 | WFE (lithography, etch, deposition) | — None — | Uncovered — gap | ASML is bench #1 (conviction 5 in prior version; excluded only due to budget allocation). Next deep-dive cycle: ASML closes L1. This is the most important cohort gap |
| L0 | Materials / gases | — | Uncovered | Shin-Etsu, SUMCO, SUPRA (Crucible) are watch-tier |
Two explicit gaps from B1 watchlist refresh:
- L4 (memory): SK Hynix would close this gap. >50% HBM bits, HBM4 ramping into Rubin/MI450X/TPUv7. Memory is Bottleneck #2 and has zero cohort coverage.
- L1 (litho/WFE): ASML would close this gap. The "only irreplaceable company in chips" carries conviction 5 in prior work and is the cleanest excluded name in the bench. ASML + SK Hynix are the highest-priority additions for the next deep-dive cycle.
§ 04Section 4 — Pair Structures and Cohort Trades
Pair 1: TXN long / NVTS long (1:3 in TXN's favor) — Finalized
Structure: TXN 2.5% NAV / NVTS 0.75% NAV. Ratio 1:3 TXN-to-NVTS. This pair is finalized and not subject to re-sizing without a thesis event.
Directional view: Both names are expressions of the GaN four-way race, but at different layers of the bet. TXN bets on system-BOM vertical integration at L8b-L8c — capturing the 1,500+ analog SKUs per rack around the headline GaN socket. NVTS bets on single-stage 800V-to-6V density at L8b — a purer architectural bet on the socket itself. The pair captures both the conservative-and-broad (TXN) and the speculative-and-concentrated (NVTS) expressions of the same thesis without doubling down on either.
Pair mechanics: From TXN financial.md stress test — at TXN $280/NVTS $16.77: TXN bull case $360-435 + NVTS bull case $35-40 delivers pair return of +32-55% on a weighted basis. TXN bear case $200-225 + NVTS bear case $4-8 delivers pair return of approximately -26-36%. Ratio approximately 1.4:1 — adequate for the cohort-hedge value TXN provides.
Pair breaks at TXN $210. Below $210 (below multi-year support and below the MOFCOM worst-case scenario + cycle-stall combo), the TXN anchor of the pair is impaired. Exit the pair, not just NVTS, if TXN breaks $210 on closing basis.
Pair 2: ETN long / VRT long — Concentration Trade, Not Hedge
Structure: ETN medium (~2.0% NAV max) / VRT small (1.0-1.5% NAV). Combined cap: ≤1.5-2.0x TXN NAV.
Directional view: This is a concentration trade on AI-DC infrastructure, not a hedge pair. Both names move directionally together on AI-capex. The distinction is risk-adjusted return: ETN is the lower-beta, higher-margin-of-safety expression (utility T&D + industrial + aerospace diversification, Boyd Thermal closing the CDU gap, 27.9x EV/EBITDA); VRT is the purer-play with the best business quality (32% ROIC, 2.9x book-to-bill, 53.4x EV/EBITDA). In a VRT multiple-compression scenario, ETN drawdown is approximately -27% vs VRT drawdown of -42-54% — ETN provides meaningful downside protection per dollar of AI-DC exposure.
Multi-cycle cushion vs. pure-play: ETN's Electrical Americas segment (utility T&D, industrial power quality, aerospace) provides a floor that VRT's pure AI-DC mix does not. In a mild AI-capex moderation (not pause), ETN's non-AI segments sustain earnings while VRT's multiple compresses faster. This is why ETN receives the larger allocation within the combined cap.
Combined cap at base sizing check: ETN 1.75% + VRT 1.25% = 3.0%. TXN 2.5%. Ratio 1.2x. Within the 1.5-2.0x ceiling. At full-size entries (ETN 2.5%, VRT 2.0%, TXN 3.0%): combined 4.5%, ratio 1.5x — exactly at the ceiling. Do not exceed these without re-examining the AI-capex concentration risk explicitly.
Pair 3: VRT long / Schneider Electric (SU.PA) relative-value consideration
Directional view: VRT at 53.4x EV/EBITDA versus Schneider Electric at approximately 22x EV/EBITDA for comparable AI-DC infrastructure exposure. The valuation gap is 143% in VRT's favor. For a user wanting AI-DC infrastructure exposure without VRT's full multiple, SU.PA offers:
- Explicit 800V product portfolio with planned 2028-2030 real revenue ramp
- Record backlog confirming the same AI-DC order dynamic
- Schneider home-field advantage in EMEA — where VRT is showing Q1 2026 -20% YoY
- ~22x EV/EBITDA provides 60% better downside margin-of-safety vs VRT
Should you consider SU.PA instead of (or in addition to) VRT? For AI-DC infrastructure exposure with lower multiple risk, SU.PA is defensible as the Bench long entry point before VRT corrects. SU.PA is currently on the bench (watchlist §1); the 2028-2030 revenue ramp framing means near-term catalyst clarity is lower than VRT's $15B backlog picture. The trade-off: SU.PA offers better risk-adjusted entry but lower near-term earnings momentum. The practical answer is to take a probe in VRT now for the Kyber cycle option, and consider SU.PA as a relative-value add if VRT reaches full sizing without SU.PA entering the deep-dive.
Not recommended as an explicit pair-trade (long SU.PA / short VRT) because the EU F-Gas implementing act (June 2026) could move both names together, and the EMEA competition dynamic creates correlation risk in a single-name relative trade.
Pair 4: WOLF short / ON (or NVTS) small long — SiC structural break with selective long
Structure: WOLF short 1.0-1.5% NAV, paired with onsemi (ON) long at ~1.5-2.0x WOLF gross. ON is on the bench (not deep-dived); the pair is approximated.
Directional view: WOLF substrate share has collapsed from >60% to 33.7% monotonically over 36 months with no inflection signal. ON (onsemi) holds the SiC vertical-integration thesis in the same end-market (EV + AI-DC) with a strengthening competitive position via the Qorvo SiC JFET acquisition and >$250M AI-DC revenue in 2025. The pair captures the "structural transitions don't guarantee equal outcomes" theme — same end-demand, opposite competitive trajectory.
NVTS is a much weaker pair leg against WOLF because NVTS is GaN (L8b) not SiC (L8a) — different layers of the power-semi stack. The thematic link exists (both are power-semi 800V beneficiaries) but the direct competitive overlap that makes a pair structurally clean is absent. ON is the correct pair leg.
Current state: WOLF is sized at medium short (~1.25% NAV); ON is on the bench without a deep-dive position. The pair should be approximated at 1:1.5 gross WOLF:ON until an ON deep-dive is completed.
Pair 5: INTC short / TXN long — Foundry Leadership + Embedded Analog
Structure: INTC small short (~0.65% NAV) / TXN long (2.5% NAV). Asymmetric size favors TXN.
Directional view: Both are bets on US fab buildout outcomes — in opposite directions. TXN is the vertical-integration-with-chips-subsidy thesis: ~$6.6B PV from CHIPS + §48D + state subsidies, internal manufacturing at 95%+ by 2030, no foundry ambiguity. INTC is the structurally-impaired-IDM-betting-on-14A binary: -$48B cumulative FCF, ROIC ~0.7% vs WACC 12.7%, foundry external book press-release-grade. The pair captures US domestic fab thesis in one direction (TXN: working) and one direction (INTC: broken). The sizing asymmetry reflects the conviction asymmetry: TXN is a medium core position; INTC is a probe-sized short with asymmetry that barely clears 1.3:1.
Note: Section 232 is an ambiguous catalyst for this pair. A punitive Section 232 on advanced-node imports hurts TSM/NVDA margins but is theoretically supportive of INTC's domestic-foundry narrative. If Section 232 lands as a broad tariff, the INTC short leg should be reconsidered. The TSM long / INTC short pair (documented in the prior portfolio summary) remains the cleaner expression of foundry leadership divergence for that reason — it is purely process-node, not tariff-mediated.
Pair 6: Cohort vs. SOX Hedge — Market-Neutral Expression
Directional view: The cohort currently has ~15% NAV net long exposure in a sector (semiconductors + AI infrastructure) that is itself correlated with the SOX index. For a user wanting market-neutral expression of the chip-to-grid thesis, the cohort could be expressed as long-cohort / short-SMH (or short-SOX).
Recommendation: Not recommended as a primary expression. The cohort's differentiated alpha (vs. passive SOX exposure) sits in:
- The VRT/ETN non-semi positions that SOX does not capture
- The WOLF structural-break short that outperforms both the SOX and the sector
- The TXN-NVTS pair's GaN architecture bet that is idiosyncratic to the 800V transition timing
A passive SOX hedge would neutralize the non-semi alpha while leaving the GaN/cooling alpha intact — but at the cost of additional short exposure in a structurally bullish sector. Better to run the book with modest gross long exposure and manage the AI-capex concentration risk through the ETN+VRT cap rule rather than a market-neutral overlay.
§ 05Section 5 — Cohort Concentration Risks (Honest)
Risk 1: AI-Capex Single-Factor — Now the Cohort's Largest Residual Risk
What it is: A sustained hyperscaler AI-capex pause (defined as Big 5 hyperscalers collectively cutting 2027 datacenter capex budgets by 20%+ relative to 2026 published guidance) would hit the cohort simultaneously at multiple layers. NVDA (direct accelerator revenue), AVGO (ASIC design-services revenue), ETN (order intake deceleration), and VRT (both multiple compression and eventual backlog depletion) all move negatively in a pause scenario. TXN provides partial offset through its diversified industrial base, but only partial — its AI-DC data center revenue (+90% YoY Q1 2026) is still 9% of total revenue and would decelerate.
Quantified NAV exposure at base sizing: Approximately 9.5% NAV in names where AI-capex pause is the dominant bear-case mechanism (NVDA 4.0%, VRT 1.25%, ETN 1.75%, AVGO 2.5%). In a -35% drawdown scenario on this group (consistent with a 20-25% hyperscaler capex cut): portfolio impact approximately -3.3% NAV. In a -50% drawdown scenario: -4.75% NAV. These numbers assume TXN (+0.5%), WOLF short (+0.4%), and INTC short (+0.25%) provide partial offsets in a downturn.
The ETN+VRT cap rule is the portfolio-construction mechanism that keeps this risk contained. Without it, a user tempted to size ETN to 3% and VRT to 2.5% would have 5.5% combined AI-infrastructure amplification on top of the NVDA/AVGO silicon exposure, concentrating a 15%+ NAV position in a single macro factor. Do not let the cohort's analytical coherence rationalize sizing discipline away.
Risk 2: Cooling/Thermal Three-Player Consolidation by 2027
What it is: Until March 2026, VRT held unchallenged CDU market leadership (~22% share). Two simultaneous M&A events changed the structure permanently: Eaton acquired Boyd Thermal ($9.5B, $1.7B revenue at 80%+ DC mix, closed March 2026) and Ecolab acquired CoolIT ($4.75B, March 2026). The liquid cooling market is now a credible three-player race (VRT / ETN-Boyd / Ecolab-CoolIT) entering the highest-demand window of the AI-DC cycle (2026-2028). This matters because cooling is VRT's fastest-growing, highest-margin product line — "liquid cooling revenue more than doubled in Q1 2025" per competitor.md — and margin expansion from cooling penetration growth is the mechanism that justifies VRT's premium multiple.
Monitoring requirement: Watch CDU market share quarterly from Q3 2026 onward. The observable signal is VRT cooling-segment revenue growth relative to Boyd's disclosed integration revenue and any CoolIT revenue break-outs in Ecolab reporting. A 5+ ppt CDU share loss by VRT over two consecutive quarters is the trigger to reassess the probe sizing — the cooling moat is the load-bearing support for the 53.4x EV/EBITDA multiple.
Risk 3: Calendar Mismatch on Backlog Conversion
What it is: Both ETN ($19.6B backlog, DC orders +200% YoY Q4 2025) and VRT ($15B backlog, 2.9x book-to-bill) carry enormous order books that are real but are gated by tier-2 chokepoints that are independent of demand. For ETN: GOES (grain-oriented electrical steel) with 128-144 week lead times, Cleveland-Cliffs as the sole US domestic producer, gates power-transformer production at the core of ETN's Electrical Americas backlog. For VRT: HFO refrigerants (R1234ze), with Honeywell and Chemours controlling >75% of global production, gates chiller/CRAC production at the moment of maximum backlog ramp in 2026-2027. These two chokepoints mechanistically explain why Schneider Electric's management uses a "2028-2030 real market impact" framing for 800V revenue — not because demand is uncertain, but because conversion of orders to revenue is physically gated by supply chains with 2-3 year lead times.
Portfolio implication: A 2026 earnings disappointment at ETN or VRT driven by chokepoint-limited production — not demand-limited orders — is the most likely near-term catalyst that does not invalidate the thesis but would compress the multiple. The correct response to this scenario is to add on weakness, not exit. The kill signal is demand-side cancellations, not production-side delays. Monitoring requirement: Watch ETN Electrical Americas revenue growth vs order growth gap; watch VRT deferred revenue balance ($1.8B, +71% YoY) as the indicator that hyperscalers are still pre-paying (demand intact) even if conversion lags.
Secondary Risks
Mexico tariff exposure (ETN and VRT): ETN's Juarez/Monterrey switchgear manufacturing and VRT's Mexicali power-equipment plant face potential 25% Section 232 tariff exposure. At adverse scenario, ETN COGS impact estimated at $150-250M annually with only partial pass-through. VRT's Mexicali expansion was partly scheduled to reduce this risk via capacity flexibility. The Mexico manufacturing tariff is a cohort-specific risk not captured in the prior Taiwan/China geographic risk matrix.
MOFCOM antidumping final determination on TXN (September 13, 2026): Bernstein sizes 11.4% of TXN revenue exposed; our estimate is 3-6% of FY26 revenue at meaningful adverse-outcome risk. This is the single largest asymmetric downside in any long thesis in the cohort on a dated binary basis. TXN position should be held at 1/2-to-2/3 size through Q3 2026 to defend through the binary; scale back to full only after the September 13 determination lands.
EU EMEA Schneider home-field advantage / VRT EMEA crack: VRT EMEA Q1 2026 -20% YoY is the first geographic fault line in the AI-DC order story. If two more consecutive EMEA quarters print negative, it confirms structural Schneider share loss in VRT's second-largest market — not a thesis-killer but a multiple-compressor that accelerates the need to correct to sub-$250 entry.
§ 06Section 6 — Top Catalysts on the Cohort Calendar (Next 12 Months)
Ranked by expected impact on cohort positioning, not by date:
| Rank | Date | Event | Direction | Affected Names | Source |
|---|---|---|---|---|---|
| 1 | April 2, 2026 | ITC Final Determination — Infineon v. Innoscience (337-TA-1407) | Binary for GaN IP regime | NVTS, TXN, ETN, VRT (indirectly) | competitor.md, regulatory.md (NVTS, TXN) |
| 2 | September 13, 2026 | MOFCOM antidumping Final Determination on TXN | Binary bearish for TXN specifically | TXN, cohort GaN read | customer.md, regulatory.md (TXN) |
| 3 | Q2 2026 (August earnings) | Eaton Q2 2026 earnings — Boyd integration revenue run-rate vs $1.7B target; post-H1 margin recovery; GOES impact disclosure | Binary (conviction-shaper for ETN long) | ETN | financial.md, supply-chain.md (ETN) |
| 4 | Q2 2026 (~July) | Vertiv Q2 2026 EMEA print — second geographic data point after Q1 2026 -20% YoY; confirms or refutes structural Schneider share loss | Binary (EMEA-crack confirm or deny) | VRT | competitor.md, market.md (VRT) |
| 5 | June 2026 | EU F-Gas Regulation implementing act — refrigerant transition timelines, R1234ze production requirements | Bearish for VRT cooling-ramp timeline | VRT | regulatory.md (VRT) |
| 6 | H2 2026 | Section 232 presidential decision — advanced-node semiconductor tariffs | Binary (positive for INTC domestic narrative; modal 200-400bp GM hit for NVDA/TSM/AVGO absorbed via pricing power) | NVDA, TSM, AVGO, INTC | regulatory.md (all) |
| 7 | H2 2026 | VRT H2 2026 800V product portfolio launch — first-to-market advantage 18-24 months ahead of Schneider; per-MW content uplift 70-120% | Bullish for VRT; catalyst for Kyber era revenue re-rating | VRT | market.md, competitor.md (VRT) |
| 8 | H2 2026 | 14A external customer commitment decision — two prospective customers under evaluation | Binary (bear thesis retire trigger if positive; accelerate if negative) | INTC (short) | supply-chain.md, customer.md (INTC) |
| 9 | Q1 2027 | ETN Mobility segment spin-off — strips ~$3B low-margin Vehicle/eMobility; reshapes ETN into $24B Electrical + Aerospace + Thermal at ~26% segment margins | Bullish (multiple re-rating catalyst; most important structural event in ETN thesis) | ETN | competitor.md, market.md (ETN) |
| 10 | 2027 | NVIDIA Vera Rubin Ultra deployment timing — 800V Kyber rack at 600kW-1MW per rack; per-MW content inflection for VRT/ETN | Bullish for VRT, ETN; validates Kyber-era revenue uplift thesis | VRT, ETN | synthesis.md, market.md (VRT, ETN) |
Additional dated catalysts not ranked above but important:
| Date | Event | Direction | Source |
|---|---|---|---|
| May-July 2026 | WOLF Q3 FY26 + $575M refi + CHIPS PMT resolution stack | Bearish (short-confirming) | financial.md, regulatory.md (WOLF) |
| H2 2026 | AMD MI450X / Helios benchmarks at OpenAI 1 GW initial deployment | Binary (NVDA moat test) | competitor.md (NVDA) |
| H2 2026 | TSMC Section 232 presidential decision + A16 ramp | Bullish (modal) | regulatory.md, market.md (TSM) |
| Q2-Q3 2026 | FERC Order 1920 RTO/ISO compliance filings | Bullish for ETN (crystallizes PJM pipeline) | regulatory.md (ETN) |
| Q1 2026 | NVTS PSMC 100V production start | Bullish if confirmed on-schedule; bearish if delayed | supply-chain.md (NVTS) |
| April 2026 | Japan M7.7 quake TOK Koriyama resist supply impact on TSMC Q2/Q3 prints | Binary (Q2-Q3 TSMC earnings test) | supply-chain.md (TSM) |
§ 07Section 7 — Open Questions for the User
These are the five questions the user should think through before sizing positions. They are not analytical gaps — the analyst work is complete. They are PM-level judgments that only the user can make.
1. Cohort net long-bias at 7L/2S — comfort level?
The cohort is structurally long-biased by design, reflecting the user's bullish worldview on the chip-to-grid structural buildout. Taiwan-tail is materially hedged by TXN/ETN/VRT (three of the four expansion names have low-to-zero Taiwan exposure). AI-capex single-factor concentration is the dominant residual risk, managed through the ETN+VRT cap rule but not eliminated. The user should explicitly decide: at what trigger do you reduce cohort gross exposure? Suggested answer: if three or more Big 5 hyperscalers revise H2 2026 datacenter capex guidance down more than 15% in Q1/Q2 2026 earnings calls, reduce the cohort's core long exposure by 30-40% across the board before the order-book-to-earnings transmission fully shows up.
2. Cooling consolidation watch-list — when do you trim VRT?
The three-player liquid cooling market (VRT / ETN-Boyd / Ecolab-CoolIT) created by March 2026 M&A requires active monitoring through 2027. The user should set a specific monitoring protocol: quarterly review of VRT cooling-segment revenue growth relative to Boyd integration disclosures. Define trim trigger: if VRT discloses two consecutive quarters of cooling-segment revenue growth below 20% YoY while ETN discloses Boyd cooling revenue on plan, interpret as VRT share loss and trim to sub-1% NAV pending re-evaluation at a more attractive entry.
3. Bench-to-core promotion — ASML and SK Hynix timeline
The cohort has two explicit gaps: L1 (WFE/litho) and L4 (memory). ASML is conviction 5 and was excluded from this deep-dive cycle only because the analytical budget went to chip-to-grid expansion. SK Hynix is the dominant HBM player in a market with Bottleneck #2 status in the three-bottleneck frame. The user should decide: does the next deep-dive cycle (after this one concludes) run ASML + SK Hynix as the L1/L4 gap-closers? Both would add meaningfully to cohort coverage without overlapping existing names.
4. Pair execution — explicit pairs vs. independent positions?
The three finalized pairs (TXN/NVTS, ETN/VRT, WOLF/ON) can be managed as named pairs with explicit ratio discipline, or as independent positions with individual stop-loss and add disciplines. The TXN/NVTS pair is already finalized with explicit ratio (1:3) and a named break price ($210). ETN/VRT is a concentration trade with a combined cap rule. WOLF/ON is approximated without a deep-dived ON thesis. The user should decide whether to (a) run explicit pairs with ratio rebalancing at defined triggers, or (b) manage each position independently. Recommendation: run TXN/NVTS as an explicit pair; manage ETN/VRT independently against the combined cap; hold WOLF as a standalone short until ON is deep-dived.
5. Recession scenario — cohort gross exposure trigger?
The cohort at base sizing carries ~17% NAV gross long exposure in a sector with high PMI sensitivity (embedded analog, industrial power, AI capex). In a recession scenario (US GDP -1% over two quarters, PMI below 47 for three consecutive months), the analog cycle stalls (TXN bear case), ETN's industrial segment decelerates, and the AI-capex pause scenario becomes higher-probability. The user should define a recession trigger at which the cohort gross long exposure is cut to 10-12% NAV. Suggested: maintain current sizing through mid-2026 catalyst cluster (Sections 232, MOFCOM, WOLF May-July window); if post-cluster macro deteriorates, cut 30-40% of gross long exposure across core names before year-end 2026.
§ 08Conviction Distribution Summary
| Ticker | Direction | PM Conviction | Architecture Role | Primary Thesis Risk |
|---|---|---|---|---|
| TSM | Long | 5 | Core | Taiwan tail (binary, uninsurable) |
| NVDA | Long | 4 | Core | Hyperscaler counter-leverage + rack moat retest 2027-28 |
| AVGO | Long | 4 | Core | Valuation (already prices 22% CAGR; no margin of safety) |
| TXN | Long | 3 | Core + hedge | MOFCOM binary September 13, 2026 |
| WOLF | Short | 4 | Short | Takeover optionality (Renesas exit-seller at premium) |
| ETN | Long | 3 | Completion | Boyd integration execution + GOES-gated backlog conversion |
| VRT | Long | 3 | Completion | Multiple compression (53.4x EV/EBITDA) on any AI-capex moderation |
| INTC | Short | 3 | Short | Federal equity / CHIPS loosening flattening asymmetry |
| NVTS | Long | 2 | Probe | Foundry triple-transition + Innoscience cost-floor risk |
§ 09Cross-References
- Cohort synthesis (v3 C-COHERENCE): ./synthesis.md
- Company universe (v3):
./companies.json - Refinement log (full audit trail, Findings 1-16 + C-COHERENCE):
./refinement-log.md - Watchlist (B1 refresh): ./watchlist.md
- Per-ticker theses:
- Core: ./NVDA/thesis.md, ./TSM/thesis.md, ./AVGO/thesis.md, ./TXN/thesis.md
- Completion: ./ETN/thesis.md, ./VRT/thesis.md
- Probe: ./NVTS/thesis.md
- Shorts: ./WOLF/thesis.md, ./INTC/thesis.md