AI infrastructure: power and cooling, full category review

Date: 2026-11-02. Research only; no trades or policy changes. Horizon 10–20 years.

Result

These are analytical judgments, not measured probabilities. All 13 categories have been assessed for the first three; data limitations are scored conservatively. A complete category assessment does not mean every future acquisition outcome or unknown ownership metric has been established. No numerical score is invented for GEV. No candidate passes 100 on this calibration. VRT is near the threshold; a one-point difference should not be treated as economically precise. A reasonable adverse judgment on one or two categories could move any total several points.

Full matrix and justification

The first 12 categories have maximum 10 each. Valuation has maximum 5. No valuation bonus is awarded. Current short interest and complete beneficial-ownership analysis are not verified; ownership is capped at 7. Culture is conservatively assessed from operating/engineering execution, not an employee survey.

Vertiv: financial evidence and risks

Q3 release: sales $4.313bn, +61% reported / +55% organic; GAAP operating profit $984mn, implying 22.8% margin; adjusted FCF $652mn. FY2026 adjusted EPS guide $7.10–7.20 is an estimate, not realized owner earnings. October 28 release.

The 10-Q gives YTD OCF $2.661bn, cash capex plus capitalized software $430.3mn and a $2.313bn deferred-revenue cash inflow. Net working capital contributes $540.5mn after inventory/receivable investment. Simply subtracting every advance would overstate the deterioration, because advances support fulfillment costs, but annualizing this cash conversion would be aggressive. Cash plus short-term investments $3.712bn versus debt $2.929bn is pre-UIG. UIG closed October 26 for $1.45bn cash plus up to $1.15bn earnout; this implies approximately $0.667bn net debt after the initial payment, holding other movements constant. September net cash is not the current post-acquisition state. SBC $50.5mn YTD; quarterly diluted shares 392.682mn versus 390.929mn. Q3 10-Q.

Inference: installed-base service and integration create switching friction, but competing engineering suppliers and hyperscaler procurement prevent a monopoly rating. ROE is flattered by the equity base; it is not a substitute for normalized acquisition-inclusive ROIC. Persistent 20%+ growth would require sustained AI economics at customers. Lower future rack-power spending, cancellations, margin competition or failed integration would break the bullish case.

Potential upgrade: stronger recurring conversion through a full investment cycle, demonstrated UIG returns, and a valuation with an adequate downside margin could justify reconsideration. One outstanding quarter alone does not close those checks. Relative to historical local 98, this assessment is 99; neither is a pass.

Eaton: financial evidence and risks

Q2 sales $8.531bn, +21% reported / +14% organic; segment margin 23.1%, down 80bps YoY. GAAP EPS $2.11 versus adjusted $3.15; intangible amortization, transaction and restructuring charges explain the difference. FY2026 EPS guide $10.36–10.56 GAAP / $13.40–13.60 adjusted. Segment margin is not consolidated GAAP operating margin. July 31 release.

H1 OCF $1.634bn minus $446mn capex gives $1.188bn cash FCF, before disposals. Net income $1.690bn. Working-capital use $833mn is less adverse than prior year; this is part of the cash-flow improvement. Debt totals $20.611bn versus cash/short-term investments $695mn. Acquisition cash spending $11.079bn includes Boyd Thermal $9.55bn and Ultra PCS $1.53bn. Diluted quarterly shares 389.5mn versus 391.4mn. Deferred revenue increases from $923mn to $1.159bn, including $51mn acquisition effect; it is not all organic cash. Receivables and unbilled balances have grown. Q2 10-Q.

Inference: product qualification, safety, specification and channels support strong pricing power, but acquired cooling growth must compensate for financing costs. A mechanical annualized H1 net-income / average-equity proxy is roughly 17%, not clean ROIC: the timing of Boyd and acquisition accounting makes historical per-share returns a poor estimate of prospective incremental returns. Price paid for the asset, debt service and margins matter more than backlog growth alone.

Next catalyst: November 3 earnings, confirmed by issuer calendar. No Q3 results have been assumed. Stronger organic profit conversion, deleveraging and acquisition returns would be necessary to raise the quality score materially.

Schneider Electric: financial evidence and risks

Q3 sales update reports +17% organic and €11.5bn revenue, with Energy Management +19% and Industrial Automation +8%. It does not provide a fresh Q3 balance sheet or cash-flow statement. October 16 revenue update / issuer results page.

H1 adjusted EBITA €4.093bn / 19.3% margin, net income €2.488bn and adjusted EPS €4.79. Issuer-defined OCF €3.755bn is before working-capital movements; subtracting only capex €669mn would overstate free cash. After trade/nontrade working-capital uses €1.208bn / €247mn, FCF is €1.631bn. Net debt €15.360bn. Cash conversion 66% is seasonal, with prior-year factoring and a legal fine affecting comparability; development capitalization also matters. July 30 H1 release, including cash bridge.

PTC agreement has approximately €22bn cash consideration, intended funding €5–6bn equity and €16–17bn debt. Closing expected Q3 2027. Management expects transaction ROCE above WACC by year five including synergies; that is a forecast, not proof of accretion to economic value. Cognite remains another integration/funding commitment. October 5 original agreement release.

Inference: legacy electrification and controls quality is high. The all-in investment changes when the buyer adds material debt and shares. High installed-base loyalty does not make transaction synergies certain. Neither current standalone P/E nor historic ROCE can establish post-deal per-share value. The score discounts this directly rather than assuming acquisition-adjusted EPS represents the complete owner cost. Full standalone SBC and a fully diluted post-financing capitalization have not been independently reconstructed; no bonus is awarded for either.

Explicit valuation sensitivity

Common reference date: October 30 closing prices. VRT $232.50 reference; ETN $369.97 dated history; SU €237.50, independently consistent with the previous-close field on its November 2 page reference. Secondary quote sources can differ slightly. No live broker quote is claimed.

Use ten-year terminal price = starting EPS × (1 + per-share EPS growth)^10 × exit P/E. Price CAGR = (terminal/current price)^(1/10) − 1. Dividends, FX, Danish tax, brokerage costs and interim drawdowns are excluded. These are unweighted illustrative scenarios, not probabilities or fair-value targets. Growth is after dilution and financing effects; for Schneider, the assumed future per-share path must absorb its capital increase.

Starting EPS anchors: VRT $7.15 and ETN $13.50 are current-year adjusted guide midpoints. SU €9.58 is twice H1 adjusted EPS, explicitly a rough normalized working anchor, not company guidance; H2 seasonality and transaction effects make it uncertain. Adjusted EPS can exceed durable owner earnings. The model therefore does not establish undervaluation.

Rationale: VRT bull requires a decade of exceptional compound growth beyond the present orders surge; base allows maturation. ETN/SU base reflects diversified industrial growth with electrification support, while bear reflects slower demand and multiple normalization. A truly severe downside can be much worse than the illustrated bear. A further 20% reduction to the EPS anchor reduces ten-year price CAGR by roughly 2.2 percentage points, all else equal. Dividends can add return, but do not solve weak acquisition economics.

At these assumptions none warrants a valuation-mismatch bonus. A stock can have attractive products and still deliver mediocre long-term returns if growth slows or its multiple falls.

GEV screen, not full score

Q3 revenue $12.642bn, GAAP net margin 10.2% and FCF $5.4bn show progress. Q3 release. However YTD OCF $16.7bn includes $17.0bn working-capital inflow, driven by $19.5bn contract-liability/deferred-income inflow. The €8bn Trench acquisition is planned half cash / half debt, expected closing late 2027. Q3 filing. Wind liabilities, advance obligations and service margins need normalization before annualizing cash or computing a full score. No eligibility claimed.

Portfolio comparison and action

TSM's existing 110 and GOOGL's existing 96 are November 1 internal analytical scores, not rescored here. All three power/cooling candidates diversify business exposure away from chip fabrication, but still share AI spending risk with TSM. None is demonstrated clearly superior to TSM. VRT is the highest-priority follow-up, but 99 does not pass the current rule. ETN and Schneider have interesting franchises, but financing/capital-allocation costs are too material to treat as automatic GOOGL replacements. GOOGL being below gate does not make any lower-scoring alternative eligible.

No new allocation, DCA multiplier or trading instruction follows from this note. Current strategy remains unchanged. The owner's 39,800 DKK cash balance does not justify relaxing the 100-point requirement.