The Electrification Investment Paradox
GE Vernova builds some of the most important machines in the electricity system. At the end of June 2026 its remaining performance obligations, the contracted work it has yet to deliver, stood at USD 176 billion, and second-quarter orders rose 88% on an organic basis, according to its second-quarter results filing. At its 15 October 2026 close of USD 926.44, it also trades at roughly 78 times my estimate of its normalized earnings, or roughly 61 times the consensus estimate I could retrieve.
I do not read that as a prediction that GE Vernova will stumble. It is arithmetic. To earn an 8% annualized return over ten years from that price, my model says GE Vernova’s earnings per share would have to grow about 21% a year, every year, for a decade. My own base case is already generous, with earnings per share growing 16.6% a year, and it still produces a return of only 4.0% a year, because the multiple I assume compresses from 78 times earnings to 24 times by 2036.
That gap between a superb operating business and a mediocre investment is the subject of this article. Electricity demand is growing, grids are being rebuilt and the equipment is being ordered. What is not obvious is how much of that is already in the price.
So I set out to compare eleven publicly traded companies that sell into this build-out: ABB, Eaton, Schneider Electric, Siemens, GE Vernova, Siemens Energy, Hubbell, nVent, Prysmian, Quanta Services and Legrand. I wanted to know which are the strongest businesses regardless of price, which offer the best expected return at today’s prices, and how a Canadian can actually own them without tripping over currencies, depositary receipts, registered-account rules and foreign withholding. For each I built a ten-year operating model with bear, base and bull cases, starting from the company’s own 2026 guidance where it gives earnings guidance, and from my own derivation where it does not, rather than from a price multiple, and I ran it backward to ask what each share price already assumes.
A few conventions apply throughout, and they matter. All prices are closing prices on 15 October 2026. All returns are nominal, annualized over ten years, before Canadian taxes and before currency movements, and are quoted in each stock’s trading currency. They are model outputs under stated assumptions, not forecasts, and nobody should read a decimal place as precision. Where a company reported something I link to its release or filing, and where I estimated something I say so.
I should also say where I am coming from. I work for ABB. The views here are my own and not ABB’s, and everything I say about ABB comes from public information. ABB, Siemens and Eaton are the companies I follow most closely as a research interest, and I applied the same rule to all eleven, so that familiarity does not set any rating. The full disclosure is at the end of the article.
Where I come out surprised me a little. The best business in the group by returns on capital and balance sheet is ABB, and at its current price it is not the best stock. The three stocks with base-case returns above 9% are Hubbell, Legrand and Schneider Electric, and none is the fastest grower. Two of the most exciting names, GE Vernova and Siemens Energy, are the two I would be most cautious about buying today. The rest of the article explains why, starting with the theme itself.
Table 1. The eleven companies and how each participates in electrification
| Company | Primary listing and trading currency | Main electrification exposure | Main investment characteristic |
|---|---|---|---|
| ABB | SIX Swiss Exchange, CHF (reports in USD) | Electrification products, industrial automation, motion | Strongest returns on capital and balance sheet; expensive on earnings |
| Eaton | NYSE, USD (Irish-incorporated) | Electrical power distribution and protection, data-centre power | Excellent franchise priced at the 8% hurdle |
| Schneider Electric | Euronext Paris, EUR | Energy management, distribution, automation, software | Cheaper than peers, but PTC acquisition raises leverage |
| Siemens | Xetra, EUR | Smart Infrastructure inside a diversified industrial group | Highest dividend yield; about 29% of orders came from Smart Infrastructure |
| GE Vernova | NYSE, USD | Gas turbines, grid equipment, wind | Best-known name; most demanding valuation |
| Siemens Energy | Xetra, EUR | Gas turbines, grid technology, wind | Operating recovery largely reflected in the price |
| Hubbell | NYSE, USD | Utility and electrical products | Lowest multiple of the eleven; highest base-case return |
| nVent | NYSE, USD (Irish-incorporated, UK-managed) | Enclosures, cooling, connections for data centres and industry | Fast growth, high AI exposure, weak downside case |
| Prysmian | Euronext Milan, EUR | Power transmission and distribution cables | Margin gains from scarcity; cyclical |
| Quanta Services | NYSE, USD | Construction of power lines, substations and renewables | Large backlog, thin margins, high multiple |
| Legrand | Euronext Paris, EUR | Building electrical systems and data-centre products | High quality, 32% data-centre exposure, acquisition-led |
Source for the table: company disclosures summarized in my research; table is the author’s classification.
Why Electrification Is a Long-Term Investment Theme
The evidence for electrification is unusually good. The International Energy Agency’s Electricity 2026 report has global electricity demand growing about 3% in 2025 and about 3.6% a year from 2026 to 2030, with emerging economies supplying nearly 80% of the added consumption and data centres accounting for about half of the increase in the United States.
Demand growth is only the first link, though, and this is where investors get careless. Electricity demand, in terawatt-hours, is the energy a system must supply over a year. Peak load, in gigawatts, is the capacity it must be able to meet, and it determines how much generation, wiring and switchgear has to exist. Grid capital expenditure is what utilities actually spend to build and replace that equipment, equipment orders are what they commit to buy, and revenue is what manufacturers recognize as they deliver. Earnings and free cash flow are what is left after costs, tax and reinvestment, and shareholder returns are what you earn after paying a particular price for those cash flows.
Each step down that chain loses something. A utility’s capital spending becomes revenue for a switchgear maker or a line contractor, but not an equivalent amount of earnings growth, because input costs, competition, labour, capacity additions and price sit in between. My own base cases show the effect. Across the eleven companies, modelled revenue grows between 5.0% and 10.0% a year, earnings per share between 6.9% and 16.6% a year, and the resulting shareholder returns only between 4.0% and 10.2% a year. Electricity demand growing at 3.6% a year is the most certain number in that sequence, and the one an investor is least able to buy directly.
What makes the demand real, rather than merely forecast, is that it is hitting physical constraints now. According to pv magazine’s report on the US transformer market, lead times for high-capacity units have stretched to as long as four years, a figure it attributes to PwC analysts, and prices have jumped roughly 80% over five years, a figure it attributes only to unnamed industry experts. That is a secondary source, so I treat the figures as indicative. The IEA separately reports more than 2,500 gigawatts of projects stalled in grid connection queues worldwide. The pool includes renewables, storage and large loads such as data centres, not only generation, and the IEA’s own estimate is that measures short of full grid expansion could unlock 1,200 to 1,600 gigawatts of advanced-stage projects. The wires, not the willingness to build, are a limit. Skilled labour and permitting also slow projects, but I found no sourced figures for them and will not make any up.
Bottlenecks help suppliers while they last. Prysmian’s Transmission segment, which makes high-voltage cable, earned an adjusted EBITDA margin of 21.2% in the second quarter of 2026, up from 17.1% a year earlier, according to the company’s guidance upgrade release. Scarcity pays. But scarcity also invites capacity, and capacity decided today arrives in two to four years. That timing is my judgement rather than a fact, yet it matters because a stock priced as though today’s margins are permanent is a bet that the bottleneck never loosens. The bear cases in my model assume it does.
Replacement demand and new-build demand also behave differently. Replacing aged transformers, switchgear and cables is steady, recurring and not very sensitive to the economic cycle. Utilities are the customers for much of it, and I have written separately about the best Canadian utility stocks for investors who prefer to own the customers. New generation and new data-centre capacity are lumpier, depend on financing and policy, and can be paused. In my judgement Hubbell, with its utility base, leans toward the steadier side, and nVent, with its data-centre growth, leans toward the build wave. Their valuations reflect that difference only partly.
Where Electrification Spending Actually Goes
Follow a dollar of spending through the chain. At one end is generation: gas turbines, wind turbines, and the balance of plant. In the middle is the grid: high-voltage transmission equipment, transformers, switchgear and cable, then the lower-voltage distribution equipment that gets power to a neighbourhood or a factory. Beyond the meter are the industrial electrical systems and automation that run a plant, the building systems that distribute power inside a structure, and, increasingly, the electrical and cooling infrastructure of data centres. Running through all of it is construction, because someone has to install the equipment, string the lines and build the substations.
The eleven companies sit at different points, and the differences are economic as well as geographic. GE Vernova and Siemens Energy are the generation and heavy-grid names. In the second quarter of 2026 GE Vernova’s Power segment earned an adjusted EBITDA margin of 18.8% and its Electrification segment 18.4%, while Wind lost money, a margin of about negative 13.6%. Siemens Energy shows the same pattern: in its fiscal third quarter, Gas Services earned a profit margin before special items of 17.3%, Grid Technologies 19.9% and Siemens Gamesa, its wind business, 2.7%, according to the segment tables in its quarterly earnings release. Gas turbines and grid equipment are where the money is; wind has been where it goes missing.
Hubbell, Eaton, Schneider Electric and ABB make the equipment in the middle and at the edges of the grid. Legrand and nVent are closer to the building and the data hall. Prysmian makes the cable that connects them. Siemens straddles several layers, and I treat it separately because only part of the group is electrification at all. Quanta Services is a contractor: it builds and installs rather than manufactures. The distinction matters because a manufacturer with a proprietary product and an installed base can earn adjusted operating margins of roughly 21% to 24%, as Legrand and Hubbell do, while a contractor earns an adjusted EBITDA margin closer to 10%.
Capital intensity and cyclicality follow the same lines. A cable maker or a turbine builder must invest in plant and takes metal and project risk. A contractor invests little in plant but takes execution risk and competes on bids. Pricing power is highest where there is a physical bottleneck or a standard that is expensive to switch.
Backlog deserves a careful sentence, because it is the number most often used to justify a valuation. GE Vernova’s USD 176 billion is about 3.8 times the middle of its 2026 revenue guidance of USD 45.5 billion to USD 46.5 billion (my arithmetic). Siemens Energy’s EUR 162 billion is more than three years of its current quarterly revenue run rate (my arithmetic). Quanta’s backlog is USD 53.4 billion, of which USD 32.3 billion is expected to convert within twelve months, according to its second-quarter filing. Those are real contracts and they give visibility into volume. They do not tell you the margin on that volume, which depends on how the contracts were priced, what costs have done since, and whether the company executes. Siemens Gamesa’s book-to-bill of 0.38 in the latest quarter shows that a large backlog can coexist with a shrinking business. Visibility is not the same as profitability, or as a good return from today’s price.
A few companies sit just outside the eleven, and I want to say why. Hitachi Energy is a major grid-equipment business, but it is not independently investable. ABB announced that it completed the sale of its remaining 19.9% stake in the venture to Hitachi on 28 December 2022, so ABB no longer holds any of it and the only way to own the business is through Hitachi Ltd, a conglomerate whose other businesses dominate the share price. I cover Hitachi’s transformation in my article on investing in Japan from Canada. Vertiv, the data-centre power and cooling company, reported second-quarter sales of USD 3,274.3 million, up 24%, with an adjusted operating margin of 22.6%, according to its earnings release, but it is a pure data-centre play and would have turned this into an AI article. I could not verify the filings of the Korean makers HD Hyundai Electric and LS Electric, so I give no figures for them. For what buying Korean shares involves, see my piece on investing in South Korea from Canada, which focuses on Samsung and SK hynix and covers the access questions.
The Eleven Companies
I take them roughly in order of how much of the electrical system they touch, from the diversified majors to the specialists and the contractor. Table 2 at the end of this section sets the margins, returns on capital and balance sheets side by side. “Model EPS” is my own base-year estimate, explained in the next section. It is not company guidance, and where a company does guide earnings I show the difference.
ABB
ABB makes electrification products, industrial automation and motion equipment, from power distribution in buildings to the drives that keep factories running. In the second quarter of 2026, according to its results release, revenue was USD 9,475 million, up 12% on a comparable basis, orders were USD 12,042 million, up 30%, and the operational EBITA margin was 20.2%, up from 19.3% a year earlier.
ABB reports in US dollars and trades in Swiss francs. At CHF 76.20 on 15 October, the shares equal about USD 91.96 at the Bank of Canada-derived rate of USD 1.2069 per franc, and my model EPS of USD 2.75 puts the multiple at 33.4 times. That EPS is for continuing operations and excludes the gain expected when the sale of Robotics to SoftBank, at an enterprise value of USD 5.375 billion, closes in the second half of 2026. ABB has also made a recommended cash offer of 503 pence a share for Rotork, about USD 5.5 billion, which it expects to close in the first half of 2027. It is an offer, not a completed deal, and my model treats it as a 2027 cash purchase earning an assumed 3.7% after tax.
The valuation is the issue. In my base case ABB’s earnings per share grow 8.4% a year, but earning 8% from this price needs 11.3%, and the multiple I assume for 2036 is 21 times against 33 today. The base return is 5.2%, with negative 3.0% in the bear case and 9.4% in the bull case. That is a Hold. ABB reports third-quarter results on 20 October 2026, after the date of this analysis. I have neither used nor tried to anticipate them, and the analysis should be refreshed once they are out.
Eaton
Eaton makes switchgear, busway and uninterruptible power supplies, one of the most direct industrial ways to own the data-centre build-out. Its second-quarter results showed sales of USD 8,531 million, up 21% and 14% organically, a segment margin of 23.1%, electrical backlog up 43% and a book-to-bill ratio of 1.2. Eaton bought Boyd Thermal for USD 9.55 billion in March 2026, which pushed net debt to about USD 19.9 billion by my calculation from the reported balance sheet, and it has signed, but not completed, a plan to combine its Mobility business with Dana, which I have not modelled.
I could not obtain a line-by-line breakdown of the USD 3.04 gap between Eaton’s GAAP and adjusted guidance, so I charged USD 0.60 of the adjusted midpoint of USD 13.50 as recurring cost, which is my judgement, and arrived at model EPS of USD 12.90 (Table 3). At USD 402.48 the shares trade at 31.2 times my figure, or 29.8 times the unadjusted guidance midpoint.
Eaton is a first-rate franchise priced almost exactly at my 8% hurdle. The base case produces 8.0% because earnings growth of 10.4% a year is just what is needed, which leaves no cushion. It is incorporated in Ireland despite its New York listing, according to its annual report on Form 10-K, which raises a Canadian dividend-tax question I address later. I rate it a Hold because the price leaves no room for the data-centre cycle to disappoint. The rating does not depend on my EPS charge: it stays a Hold anywhere between USD 12.46 and USD 13.50.
Schneider Electric
Schneider sells energy management and electrical distribution, industrial automation and a growing software business. Its third-quarter release, published on 16 October, one day after the pricing date used in this article, showed revenue of EUR 11,460 million, up 16.9% organically. It raised 2026 targets to organic revenue growth of 12% to 14% and adjusted EBITA growth of 18% to 21%, a margin improvement of 100 to 120 basis points on 18.7%.
On 5 October Schneider announced it would buy PTC, the software company, for USD 205 a share in cash, an equity value of USD 22.6 billion, and its shares fell 9.97% that day, a fall that coincided with the announcement. They closed at EUR 258.00 on 15 October. The company expects the deal to close by the third quarter of 2027, to be accretive to adjusted EPS, and to pause buybacks in 2027 and 2028.
Schneider does not guide earnings per share, so the base year is my extrapolation of its guided EBITA growth. I estimate adjusted EPS of about EUR 10.20 (consensus is EUR 10.45) and deduct EUR 0.50 for restructuring costs that recur every year, giving model EPS of EUR 9.70. Both steps are my judgement.
The part I cannot confirm is the financing. Press reports describe up to EUR 17 billion of new debt and up to EUR 6 billion of new shares, but the company’s release does not state the split. My model assumes EUR 6 billion of equity, a reading of the press reports, and calibrates the acquired earnings to guided accretion of about 3% in 2028. Net debt then rises to roughly EUR 28 billion in 2027. Varying the equity assumption moves the return by about 0.1 percentage point, so the financing mix does not decide the rating. At 26.6 times model EPS, Schneider is the cheapest of the three electrical majors. The base return is 9.4%, with 0.8% in the bear case and 13.3% in the bull case. That is a Modest Buy, but a narrow one: the weighted return clears the 8% threshold by only 0.2 percentage points, and at model EPS of EUR 9.45 the rating would be a Hold. I hold it with less conviction than the numbers suggest.
Siemens
Siemens is the odd one out: a diversified group with automation, rail, a medical-technology subsidiary and an electrification business, and only the last is the subject here. In its fiscal third quarter, which ended in June 2026, Siemens reported orders of EUR 27.9 billion, of which Smart Infrastructure was EUR 8.0 billion, about 29%, up 42% on a comparable basis, with a profit margin of 20.0%. Fiscal 2026 EPS guidance, before purchase-price amortization, is EUR 11.20 to EUR 11.50, and my model EPS is the EUR 11.35 midpoint, so for Siemens the model figure is the company’s own pre-amortization guide. The fiscal year ends on 30 September. Siemens also said it can proceed with the planned spin-off of Siemens Healthineers now that it has binding tax decisions.
The S&P return on invested capital is 6.9%, the lowest of the group. At EUR 270.65 the shares trade at 23.8 times model EPS, with the highest dividend yield of the eleven at 1.98%. The base return is 6.1%, with a bear case of negative 1.2% and a bull case of 9.6%. Siemens is a Hold, and I return to its Healthineers complication when I compare it with ABB.
GE Vernova
GE Vernova sells gas turbines and services, grid equipment through its Electrification segment, and wind turbines. In the second quarter of 2026 revenue was USD 11,104 million, up 22%, orders were USD 24.2 billion, and Electrification revenue rose 68% to USD 3,637 million. It bought out the remaining half of the transformer maker Prolec GE in February 2026 for USD 5,254 million in cash. First-half free cash flow was USD 9,897 million, much of it customer deposits and working capital that I would expect to unwind, against a full-year guide of USD 11.5 billion to USD 12.5 billion. Net cash was about USD 10.3 billion, partly customers’ money.
The earnings are where careless analysis goes wrong, because first-half GAAP EPS of USD 19.96 includes large one-time gains. The company gives no EPS guidance, only an adjusted EBITDA margin of 12% to 14% on revenue of USD 45.5 billion to USD 46.5 billion. I took the middle of that range, applied the roughly 53% of adjusted EBITDA that reached net income in the second quarter, and divided by 267 million shares to get model EPS of USD 11.90. That is my construction, not company guidance, and it is below the lowest analyst estimate I found, USD 12.42, so I show both it and the consensus of USD 15.17.
GE Vernova trades at 77.9 times model EPS and 61.1 times consensus. The model has operating margin rising by seven points over two years, share count falling about 12% through buybacks, and a terminal multiple of 24 times. That gives 2036 earnings of USD 55.16 and a share price of about USD 1,324 against USD 926.44 today, a total return of 4.0%. The bear case is negative 7.9% and the bull case, at 30 times, is 11.1%, for a weighted return of 2.8%. It is an Avoid, but the most fragile rating in the group. At consensus EPS the base return would be about 5.9% and the weighted return 4.7%, which is a Hold. The Avoid flips at roughly USD 14.50, which would take about 65% of adjusted EBITDA converting to net income, against the 53% seen in the second quarter. It is not a poor company. It is a company whose price already contains the good news.
Siemens Energy
Siemens Energy competes with GE Vernova in turbines and grid equipment and has come a long way in a short time. In its fiscal third quarter it reported orders of EUR 17.9 billion, revenue of EUR 11.4 billion, up 18.5%, backlog of EUR 162 billion and profit before special items of EUR 1,623 million, a margin of 14.2%. The company guides the full-year margin to between 10% and 12%, towards the upper end. Adjusted net cash was EUR 7,719 million at 30 June 2026. Fiscal 2025 earnings were EUR 1.60 a share and earlier years included heavy losses, so the market is looking at a recovery from a low base.
The recovery is credible, but much of the margin gain is already at the top of what management guided for the full year. Profit margin before special items, the company’s own measure, was 17.3% for Gas Services and 19.9% for Grid Technologies in the quarter, against full-year guidance of 14% to 16% and 18% to 20%. What remains is operating leverage and a Siemens Gamesa that, at 2.7%, has only just returned to a positive margin. On a model EPS of EUR 4.35, the shares at EUR 132.36 trade at 30.4 times, and I think much of the improvement is in the price. To earn 8% a year, EPS would need to grow 14.1% a year against 11.8% in my base case. The base return is 5.8%, but the bear case is negative 8.2%, the deepest of the group, so the weighted return is 3.8%, an Avoid under my rule.
Hubbell
Hubbell is the least glamorous company on the list and, to my eye, the most interesting. It makes products utilities use to build and maintain networks, plus electrical products for commercial and industrial buildings. In the second quarter of 2026 sales were USD 1,711.8 million, up 10.2% organically, and adjusted operating margin was 23.9%, according to its results release. Model EPS is USD 20.15, which starts from the USD 20.25 to USD 20.55 adjusted guidance, excludes about USD 2.50 of non-cash acquisition amortization, and charges half of the recurring transaction costs of about USD 0.50 (Table 3). The company’s GAAP guidance, which includes both items, is USD 17.25 to USD 17.55.
Hubbell has the most leverage of the leaders after buying NSI Industries for about USD 3.0 billion with debt. Net debt is USD 4.9 billion, 52% of total capital, so only half of surplus cash goes to buybacks in the model. At USD 464.71 the shares trade at 23.1 times model EPS, the lowest multiple of the eleven, and 26.7 times GAAP guidance, almost exactly their own five-year GAAP median of 26.6. The yield is 1.22%.
The base case is 10.2% a year, the best of the eleven, with a bear case of positive 2.5%, also the best, and a bull case of 13.9%. It needs only 7.6% annual EPS growth to earn 8%, against 9.8% in the base case. What would hurt is a stumble in utility capital plans, a pause in data-centre demand, or a multiple that falls to 16. Hubbell is not cheap, but it is not asking for extraordinary performance.
nVent
nVent makes enclosures, electrical connections, protection and cooling products for data centres and industrial settings, and it is the most direct beneficiary of AI infrastructure spending among the eleven. Second-quarter 2026 sales were USD 1,471.3 million, up 53% and 46.9% organically, according to its results release. Adjusted EPS was USD 1.45. Comparisons are distorted by the 2025 sale of Thermal Management and purchase of Avail Electrical Products Group.
Model EPS is USD 4.95, the adjusted guidance midpoint of USD 5.05 less USD 0.10 of my own charge, and at USD 145.73 the stock trades at 29.4 times. My base case assumes organic growth of 12% in the early years falling to 5% by the end of the decade, since the 32% to 34% organic growth guided for 2026 is not a run rate. The S&P return on invested capital is only 9.3%, a figure I hold loosely because those deals distort 2025. The base return is 7.6%, but the bear case, at negative 4.1%, assumes growth of 2% and margins three points lower, which is what a data-centre digestion period would look like. That is why I rate nVent a Hold despite the growth.
Legrand
Legrand makes the switches, sockets, cable management, power distribution and data-centre products inside buildings. In the first half of 2026, according to its half-year results published through the French market regulator’s open data, sales were EUR 5,400.3 million, up 13.1%, with organic growth of 9.8%, acquisitions adding 6.9% and currency subtracting 3.7%. The adjusted operating margin was 20.8%. Data centres were 32% of revenue, with organic growth above 30%, up from 26% of sales in 2025. Net debt rose to EUR 5,755.3 million from EUR 3,294.0 million a year earlier as the company made seven acquisitions.
Legrand is a compounder in the plain sense: sales have grown at roughly 7.9% a year from 2021 to 2025, it pays out half its earnings, and it converts nearly all of net income to free cash. I use a reported-basis EPS of EUR 5.45, below the adjusted consensus of EUR 6.01, so the figure is my normalization and not company guidance. It is a stricter basis than the amortization add-back I allow for Eaton, Hubbell and Quanta, and on consensus the multiple would be 22.2 times. At EUR 133.70 the shares trade at 24.5 times that, about 14% above the company’s five-year median multiple of 21.5. The yield is 1.78%. With a third of sales growing above 30%, some of that premium is the market paying for data centres, and a return to the median would cost about 12% of the share price before earnings growth.
The base return is 10.0%, with 2.2% in the bear case and 13.9% in the bull case, a Modest Buy. The risks are a pause in data-centre spending that hits a third of sales, rising acquisition prices or Europe’s weakness persisting. Legrand reports third-quarter results on 5 November 2026.
Prysmian
Prysmian makes the power cables that connect generation, grids and buildings, and it is the clearest example in the group of a scarcity margin. Second-quarter 2026 revenue was EUR 6,021 million, up 9.4% organically, and adjusted EBITDA was EUR 730 million, a margin of 15.4% at standard metal prices. Its Transmission segment, which makes high-voltage cable, earned 21.2%, against 17.1% a year earlier. Guidance was raised to adjusted EBITDA of EUR 2.8 billion to EUR 2.9 billion.
The company does not guide earnings per share, so I built model EPS from consensus adjusted EPS of EUR 4.78 less about EUR 0.23 for recurring items, which is my judgement, giving EUR 4.55. At EUR 109.80 that is 24.1 times, above Prysmian’s own GAAP median of 21.2, though at 12.5 times EBITDA it is at its median on that measure. A cable maker’s earnings follow metal prices, capacity and project timing, so the terminal multiple is 15. The base return is 6.2%, the bear case negative 5.9% and the bull case 11.6%, for a weighted return of 4.5%. It is a Hold, with the weakest Canadian access of the group.
Quanta Services
Quanta is the contractor. It builds transmission and distribution lines, substations and the infrastructure around power generation for utilities and developers. Its second-quarter results show revenue of USD 9,557 million and a record backlog of USD 53.4 billion.
Deducting about USD 1.90 of stock compensation and integration costs from the USD 16.70 adjusted midpoint gives model EPS of USD 14.80 (Table 3), my own adjustment and not company guidance, and at USD 642.21 Quanta trades at 43.4 times that, or 38.5 times the adjusted midpoint. That is a very high multiple for a business with a 10.4% adjusted EBITDA margin and a 9.1% S&P return on invested capital. Even with earnings per share compounding 13.0% a year in my base case, a terminal multiple of 22 times produces a return of 5.7%, with negative 6.2% in the bear case and 12.1% in the bull case. The yield is 0.07%. Quanta is a Hold, and it is the clearest case here of a contractor being priced like a manufacturer with proprietary products. The rating holds anywhere from USD 14.80 up to the unadjusted USD 16.70, but at USD 13.90 it would become an Avoid.
Table 2. Business quality at a glance
| Company | Latest margin (company measure) | Return on invested capital, 2025 (S&P, secondary) | Balance sheet | Growth or backlog evidence |
|---|---|---|---|---|
| ABB | Operational EBITA 20.2% (Q2 2026) | 24.2% | Net debt 0.3 times EBITDA | Orders up 30% (Q2) |
| Eaton | Segment margin 23.1% (Q2) | 15.8% | Net debt about USD 19.9bn after Boyd (my calculation) | Electrical backlog up 43% |
| Schneider Electric | Adjusted EBITA 19.3% (H1) | 13.0% (company ROCE 15.1%) | Net debt EUR 15.4bn before PTC | Q3 organic growth 16.9% |
| Siemens | Industrial Business profit margin 17.3% (Q3 FY2026) | 6.9% | Industrial net debt EUR 12.2bn, 0.9x (Sept 2025) | Backlog EUR 132bn |
| GE Vernova | Adjusted EBITDA 11.3% (Q2) | Not meaningful (S&P shows 50.7%, distorted) | Net cash about USD 10.3bn, partly deposits | Backlog USD 176bn |
| Siemens Energy | Profit margin before special items 14.2% (Q3 FY2026) | Not meaningful (S&P shows 19.5%, distorted) | Net cash EUR 7.7bn | Backlog EUR 162bn |
| Hubbell | Adjusted operating margin 23.9% (Q2) | 18.1% | Net debt USD 4.9bn, 52% of capital | Organic growth 10.2% (Q2) |
| nVent | Return on sales 23.2% and 27.3% (two segments, Q2) | 9.3% | Net debt about USD 1.2bn (my calculation) | Organic growth 46.9% (Q2) |
| Prysmian | Adjusted EBITDA 15.4% at standard metal prices (Q2) | 15.4% | Net financial debt EUR 4.1bn | Organic growth 9.4% (Q2) |
| Quanta Services | Adjusted EBITDA about 10.4% (guidance) | 9.1% | Net debt about USD 5.6bn | Backlog USD 53.4bn |
| Legrand | Adjusted operating margin 20.8% (H1) | 12.1% | Net debt EUR 5.8bn | Organic growth 9.8% (H1) |
Margin definitions differ by company and are not directly comparable. Returns on capital are secondary S&P figures for fiscal 2025. Balance-sheet and growth figures are company-reported for the latest period shown unless marked as my calculation.
The Valuation Problem
Before comparing prices I need to compare earnings, and that is harder than it looks. Reported earnings follow accounting standards and include everything. Adjusted earnings are what a company presents after removing items it considers unusual or non-cash. Normalized earnings are my own attempt to describe an ordinary year, which means deciding for myself which adjustments I accept. Each gives a different multiple for the same stock.
Some adjustments are reasonable. Amortization of intangible assets acquired in a purchase is a non-cash charge, and the cash cost of the acquisition was already counted when the money went out, so I add it back. Others look temporary but are not. Restructuring, integration costs, transaction costs and stock-based compensation recur in companies that restructure, acquire and pay staff in shares every year, so I treat part or all of them as real costs. In the second quarter Quanta Services’ adjusted EPS of USD 4.24 exceeded GAAP EPS of USD 2.96 by USD 1.28, built from USD 1.03 of amortization, USD 0.42 of stock-based compensation and USD 0.19 of acquisition and integration costs, less tax effects. I accept the first and not the other two.
Unusual gains cut the other way. GE Vernova’s first-half GAAP EPS of USD 19.96 includes a USD 3,992 million pre-tax gain from remeasuring its Prolec GE stake when it bought the rest of the company, plus a USD 330 million gain on a software sale. Doubling the first-half figure would put the shares near 23 times earnings and make them look reasonable, which they are not.
Table 3. Reported, adjusted and model earnings per share for fiscal 2026
| Company | Reported (guidance midpoint) | Company-adjusted | Model EPS | What separates them |
|---|---|---|---|---|
| Eaton | USD 10.46 | USD 13.50 | USD 12.90 | Amortization excluded; USD 0.60 of restructuring and integration charged (my estimate) |
| Hubbell | USD 17.40 | USD 20.40 | USD 20.15 | About USD 2.50 of amortization excluded; half of about USD 0.50 of transaction costs charged |
| Quanta Services | USD 11.67 | USD 16.70 | USD 14.80 | Amortization excluded; about USD 1.90 of stock compensation and integration charged (my estimate) |
| nVent | USD 4.34 | USD 5.05 | USD 4.95 | USD 0.10 of recurring costs charged (my estimate) |
| GE Vernova | Not meaningful (first half USD 19.96) | Not guided | USD 11.90 | Prolec and software gains excluded; my derivation from guided EBITDA; consensus USD 15.17 is non-GAAP |
| Schneider Electric | About EUR 9.4 (my estimate) | EUR 10.20 (my extrapolation of guided EBITA growth; consensus 10.45) | EUR 9.70 | EUR 0.50 of recurring restructuring charged (my estimate) |
| Legrand | EUR 5.45 (my reported-basis estimate) | Consensus EUR 6.01 | EUR 5.45 | Adjusted consensus is higher; I did not reconcile the gap line by line |
Midpoints of company guidance except where marked. Model EPS is the author’s estimate and is lower than consensus wherever I retrieved consensus. It is not company guidance for any company in this table.
My model EPS sits below consensus for every company where I found a consensus figure. For fiscal 2027 it is lower by 2% for Hubbell, 5% for Quanta, 6% for Siemens, 8% for Schneider, 11% for Prysmian, 13% for nVent, 16% for Siemens Energy and 26% for GE Vernova, though that last consensus number I could not confirm as to year or basis. The gap is explained mostly by what I choose to count as a cost, and I would rather err toward paying for recurring charges than ignore them. I did not change any estimate merely because it differs from consensus, and where a rating threshold depends on my figure, the profile says so.
Table 4. Valuation at the 15 October 2026 close
| Company | Price, 15 Oct 2026 close | Model EPS, fiscal 2026 | P/E on model EPS | Dividend yield | Qualification |
|---|---|---|---|---|---|
| ABB | CHF 76.20 | USD 2.75 | 33.4x | 1.23% | Priced in CHF, earnings in USD (USD 1.2069 per franc); excludes Robotics sale gain |
| Eaton | USD 402.48 | USD 12.90 | 31.2x | 1.09% | 29.8x on company-adjusted EPS |
| Schneider Electric | EUR 258.00 | EUR 9.70 | 26.6x | 1.63% | Before any new shares issued for PTC |
| Siemens | EUR 270.65 | EUR 11.35 | 23.8x | 1.98% | Pre-PPA guidance; year ends 30 September |
| GE Vernova | USD 926.44 | USD 11.90 | 77.9x | 0.22% | 61.1x on consensus USD 15.17; no company EPS guidance |
| Siemens Energy | EUR 132.36 | EUR 4.35 | 30.4x | 0.53% | 29.3x on consensus; year ends 30 September |
| Hubbell | USD 464.71 | USD 20.15 | 23.1x | 1.22% | 26.7x on GAAP guidance |
| nVent | USD 145.73 | USD 4.95 | 29.4x | 0.58% | 28.9x on company-adjusted EPS |
| Prysmian | EUR 109.80 | EUR 4.55 | 24.1x | 0.82% | Built from consensus; no company EPS guidance |
| Quanta Services | USD 642.21 | USD 14.80 | 43.4x | 0.07% | 38.5x on adjusted EPS; about 55x on GAAP guidance |
| Legrand | EUR 133.70 | EUR 5.45 | 24.5x | 1.78% | Reported-basis EPS; 22.2x on adjusted consensus |
Prices are 15 October 2026 closes from a market-data aggregator’s daily histories (secondary), checked against a second provider where one was available: Eaton, Hubbell and Quanta (a second aggregator), Siemens (Stuttgart EUR 270.50 against the Xetra EUR 270.65), Schneider (a German exchange quote of EUR 257.75), Legrand (an Euronext Paris close of EUR 133.70 on a second market-data site), Siemens Energy (Xetra), and GE Vernova and nVent (second market-data sites). These are secondary-source checks and not exchange confirmations. ABB’s CHF 76.20 rests on one source. For Prysmian a Stuttgart quote showed EUR 110.80 against the Milan EUR 109.80, a gap of about 0.9% that I attribute to venue and timing; it would lower the base return by roughly 0.1 percentage point and would not change the rating. Each multiple is in the company’s own reporting currency. Dividend yields use the latest annual or annualized dividend divided by the trading-currency price. ABB’s price is converted at a rate derived from the Bank of Canada’s published CHF and USD rates for 15 October.
The valuation date deserves a warning. Siemens Energy fell 5.4% and then 2.2% on 14 and 15 October, and Eaton, GE Vernova, Quanta, nVent and Prysmian fell with it. I did not find a cause and will not invent one. Prices moved again on 16 October (GE Vernova closed at USD 921.20 and Eaton at USD 391.25), and I have kept 15 October throughout so that every figure shares one date. Moves of several percent in two days can cross a rating threshold, which is why I show how close each rating sits to its line.
What the price already assumes
A multiple tells you what you are paying, not what you are paying for. The more useful question is the reverse one: given today’s price, what has to happen for me to earn 8% a year, my hurdle for owning a business over a decade? I solved the model backward for the earnings growth that delivers it, holding the base-case terminal multiple, payout and acquisitions constant.
Table 5. Reverse valuation: growth required to earn 8% a year over ten years
| Company | Model EPS growth, base case | EPS growth needed for 8% a year | Average organic revenue growth needed for 8% | Average organic revenue growth, base case |
|---|---|---|---|---|
| Hubbell | 9.8% | 7.6% | 3.7% | 5.7% |
| Legrand | 9.3% | 7.3% | 3.7% | 5.5% |
| Schneider Electric | 9.2% | 7.8% | 4.2% | 5.5% |
| Eaton | 10.4% | 10.4% | 6.5% | 6.5% |
| nVent | 11.6% | 12.0% | 8.3% | 7.9% |
| Prysmian | 10.2% | 12.1% | 6.7% | 5.0% |
| Siemens | 6.9% | 8.8% | 5.8% | 4.0% |
| Siemens Energy | 11.8% | 14.1% | 9.1% | 6.9% |
| Quanta Services | 13.0% | 15.5% | 9.3% | 7.0% |
| ABB | 8.4% | 11.3% | 8.2% | 5.5% |
| GE Vernova | 16.6% | 21.0% | 12.4% | 8.3% |
Nominal, pre-tax, in the trading currency, from the author’s operating model. “Needed” is the growth that, with everything else in the base case unchanged, returns 8% a year. Organic growth figures are ten-year averages.
Three companies have a cushion: Hubbell, Legrand and Schneider Electric could grow earnings roughly 1.4 to 2.2 percentage points a year more slowly than my base case and still earn 8%. Eaton sits on the line. Everyone else needs more than the base case supplies, and the shortfalls are large where the multiples are large. GE Vernova would need 21.0% against a base case of 16.6% that is already remarkable, one in which operating margin rises seven points in two years.
What Could These Stocks Return Over Ten Years?
For each company I built the same chain. Revenue grows organically at a rate that changes over three phases, plus bolt-on acquisitions bought at stated multiples of sales. Operating margin moves from the base-year level to a terminal level, interest is charged on net debt, and net income converts to free cash flow at a stated percentage. Dividends are a share of the prior year’s earnings, and a fraction of surplus cash buys back stock at a price path that glides from today’s multiple to the terminal one. At the end of ten years the terminal share price is a terminal P/E multiplied by 2036 earnings per share, and total return reinvests dividends along the way.
Three large corporate actions are inside the model and three are not. Schneider’s purchase of PTC is in, with EUR 6 billion of new shares assumed, acquired earnings calibrated to guided accretion, synergies phased in over three years and buybacks paused for two. ABB’s offer for Rotork is in as a USD 5.5 billion cash purchase at an assumed 3.7% after-tax yield, which is my assumption. Eaton’s Boyd Thermal purchase is already in the base-year net debt. Not in the model are the Robotics sale gain at ABB, the Mobility combination with Dana at Eaton, and the planned separation of Siemens Healthineers, which I handle only by treating the group as consolidated.
Table 6. Ten-year return scenarios, annualized
| Company | Bear | Base | Bull | Weighted (25/50/25) | Rating |
|---|---|---|---|---|---|
| Hubbell | 2.5% | 10.2% | 13.9% | 9.2% | Modest Buy |
| Legrand | 2.2% | 10.0% | 13.9% | 9.0% | Modest Buy |
| Schneider Electric | 0.8% | 9.4% | 13.3% | 8.2% | Modest Buy |
| Eaton | -1.3% | 8.0% | 12.4% | 6.8% | Hold |
| nVent | -4.1% | 7.6% | 13.3% | 6.1% | Hold |
| Prysmian | -5.9% | 6.2% | 11.6% | 4.5% | Hold |
| Siemens | -1.2% | 6.1% | 9.6% | 5.2% | Hold |
| Siemens Energy | -8.2% | 5.8% | 11.7% | 3.8% | Avoid |
| Quanta Services | -6.2% | 5.7% | 12.1% | 4.3% | Hold |
| ABB | -3.0% | 5.2% | 9.4% | 4.2% | Hold |
| GE Vernova | -7.9% | 4.0% | 11.1% | 2.8% | Avoid |
Nominal, before tax and before Canadian-dollar effects, in each stock’s trading currency, from the author’s operating model. ABB is modelled in US dollars at a constant USD 1.2069 per Swiss franc. Weights are the author’s judgement and not probabilities. Rating rule: Strong Buy needs a base return of at least 14% and a bear return of at least 3%; Buy needs 11% and a weighted return of 9%; Modest Buy needs 9% and 8%; Avoid is a base return below 5% or a weighted return below 4%; everything else is Hold.
A rating depends on the base case and on a weighted average that gives the bear and bull cases a quarter each. That is why Siemens Energy is an Avoid with a base return of 5.8%: its bear case of negative 8.2% pulls the weighted figure down to 3.8%, below my 4% line. ABB and Quanta, at weighted returns of 4.2% and 4.3%, clear that line by hardly any margin. No company reaches the 11% base return a full Buy requires. The weights are my judgement, and changing them would shuffle the borderline cases.
The point of the exercise is the way earnings growth and returns separate. GE Vernova’s earnings per share rise from USD 11.90 to USD 55.16 in my base case, yet the 2036 share price of USD 1,324 is only 43% above USD 926.44, because the multiple falls from 78 times to 24. Hubbell’s earnings grow more slowly, at 9.8%, but its multiple only falls from 23.1 to 21, so the price roughly doubles to about USD 1,077, and with about USD 91 a share of cumulative dividends the return is 10.2%. Fast growth does not rescue a multiple that has to shrink.
In my model one extra percentage point of organic growth in every year adds roughly 1.0 to 1.1 points to the annualized return, and one extra turn of terminal P/E adds roughly 0.4 to 0.6 points. Hubbell’s base case would return 6.9% a year if it ended at 15 times earnings rather than 21, and 12.7% if it ended at 27. GE Vernova would return 6.2% at a terminal multiple of 30 times with base-case growth, still short of 8%, and ABB 7.5% at 27 times. These are sensitivities and not forecasts.
The terminal multiple is the largest single judgement in the model, and it is a judgement and not arithmetic. I set it against each company’s five-year median forward multiple, a secondary figure on consensus-adjusted earnings. For eight of the eleven my base terminal multiple is at or below that median (ABB is within 0.1 of a turn), and for Siemens, nVent and Legrand it is higher by 0.8 to 1.0 of a turn. The weakest test is GE Vernova, whose median of 55.8 times covers only two years of listed history. I think a mature power-equipment company cannot hold a multiple like that for a decade, and I use 24 times. At 55.8 times the same model would return about 12.6% a year, so the Avoid rests heavily on that view, and a reader who disagrees should reach a different conclusion.
Three ratings are also sensitive to my earnings estimates. GE Vernova becomes a Hold at consensus EPS, Schneider’s Modest Buy becomes a Hold at EUR 9.45, and Quanta’s Hold becomes an Avoid at USD 13.90. Each is discussed in its profile. Eaton’s Hold is stable across the range of EPS I considered.
What these numbers cannot do is predict. A 0.2 percentage point difference between Hubbell’s 10.2% and Legrand’s 10.0% is about what changing a terminal multiple by half a turn would do, and I do not treat it as a meaningful ranking. They are a disciplined way of asking what each price assumes, and they are most useful at the extremes, where the answers are not close.
How Much of This Investment Theme Is Really Artificial Intelligence?
It would be easy to write this as an AI article, and I think that would be a mistake. Data centres are one source of demand among several, and the companies differ a great deal in how much of their results depend on them. The IEA puts data centres at about half of the expected increase in US electricity demand, but nearly 80% of the added demand globally is in emerging economies, where the drivers are cooling, industry and household electrification. Grid replacement, industrial electrification, vehicle charging and gas generation would all be growing without a single new data hall.
Where the exposure can be measured, it is large. Legrand reports that data centres were 32% of its first-half 2026 revenue and growing above 30% organically. Siemens’ call commentary, as reported in secondary coverage that I could not match to its release, put data-centre orders at about EUR 6 billion over the first nine months of its fiscal year. nVent’s organic growth of 46.9% in the second quarter is, in my judgement, largely a data-centre story, though it does not disclose the share. Only Legrand’s figure is verified from a company document. For the others the ranking is my judgement, from most to least exposed: nVent, Legrand, Eaton, Schneider, Siemens, Quanta, GE Vernova, Hubbell, ABB, Prysmian and Siemens Energy.
A diversified supplier such as Schneider or Siemens can see its data-centre business slow and still grow elsewhere. A company like nVent or Legrand has less to fall back on, and the model’s bear cases reflect that: nVent’s is negative 4.1% and Legrand’s, though better at positive 2.2%, depends on a business where a third of revenue can pause. There is also a valuation channel: when sentiment shifts, these stocks tend to move together whatever each one’s AI revenue. The group-wide fall on 14 and 15 October looked like that, but I could not establish its cause, so I am describing a pattern and not explaining an event.
The companies whose results depend least on continued AI capital spending are, in my judgement, Hubbell, ABB, Prysmian and Siemens Energy, whose customers are utilities, process industries and the grid. They are not immune, but the dependence is indirect and spread across more customers.
Five Head-to-Head Comparisons
The profiles describe each business on its own. These comparisons set companies side by side where the contrast teaches something about the gap between business quality and shareholder return.
ABB vs Eaton vs Schneider Electric
These three sell the most similar products in the group, and their quality is closer than their prices. Their margins (Table 2) are within a few points of each other. On the S&P measure of return on capital ABB leads at 24.2% against 15.8% for Eaton and 13.0% for Schneider, and ABB has little net debt while Eaton carries about USD 19.9 billion and Schneider’s EUR 15.4 billion is about to rise.
On price the order reverses. ABB trades at 33.4 times model earnings, Eaton at 31.2 and Schneider at 26.6, and the base-case returns are 5.2%, 8.0% and 9.4%. ABB can be the strongest business and still the least attractive stock because the quality is already in the price: it needs 11.3% annual EPS growth to earn 8% and the base case supplies 8.4%, while Schneider needs 7.8% and has 9.2%.
Schneider’s cushion comes with a catch. PTC is a large cash deal that pauses buybacks and lifts modelled net debt to about EUR 28 billion, so the larger risk is that the acquired earnings fall short of guided accretion. I would not call Schneider a bargain. I would call it the one among the three where the price leaves some room for the deal to go less than perfectly.
GE Vernova vs Siemens Energy
Both build gas turbines and grid equipment, both have very large backlogs (USD 176 billion of remaining performance obligations at GE Vernova, EUR 162 billion at Siemens Energy), both are net cash, and both have a wind business that has dragged on results. GE Vernova’s adjusted EBITDA margin was 11.3% in the second quarter and Siemens Energy’s profit margin before special items was 14.2% in its fiscal third quarter. The two measures are defined differently, but the story is similar: both have recovered from weaker margins, and the market has noticed.
What differs is what is priced. At 77.9 times model earnings GE Vernova assumes margin expansion continues for years and a multiple far above the 24 times I judge a mature power-equipment company can sustain. At 30.4 times Siemens Energy assumes the recovery is permanent, a more modest assumption but not a free one. The bear cases are similarly deep, negative 7.9% and negative 8.2%.
The mistake to avoid is treating operating recovery as an entry price. GE Vernova is the better business and Siemens Energy the less demanding valuation. Neither makes either one a buy, and GE Vernova’s rating is the less secure of the two, because at consensus earnings per share it would be a Hold.
Hubbell vs nVent vs Legrand
Hubbell sells to utilities and building customers, nVent sells enclosures, connections and cooling mostly into data centres and industry, and Legrand sells building electrical systems with a third of its revenue now in data centres. Growth runs the other way from returns: nVent grew organically 46.9% in the second quarter, Hubbell 10.2% and Legrand 9.8% in the first half. The multiples are 23.1, 29.4 and 24.5 times, and the bear cases are positive 2.5%, negative 4.1% and positive 2.2%.
Hubbell and Legrand come out ahead for reasons that have little to do with growth. Both earn adjusted operating margins above 20%, both are priced at multiples that do not have to shrink much to reach my terminal assumption, and both have bear cases that stay positive, as does Schneider’s and no other company’s. nVent has the highest growth and the weakest margin of safety. It needs average organic growth of 8.3% a year against a base case of 7.9% to earn 8%, so the data-centre run has to last longer than I assume.
Between Hubbell and Legrand the numbers do not decide, because 0.2 percentage points is within the model’s error. Hubbell is anchored in utility spending, carries more debt after the NSI deal and trades on the NYSE in US dollars as a US corporation. Legrand converts nearly all net income to free cash, but 32% of sales sit in data centres, its leverage rises with each acquisition, and a Canadian has to buy it in Paris or through an over-the-counter receipt that is unsponsored according to the filings I found.
Prysmian vs Quanta Services
Prysmian manufactures cable and Quanta installs it, along with much else, and the economics differ. A cable maker must build plants and carry metal inventory, and its adjusted EBITDA margin was 15.4% in the second quarter. A contractor must staff, equip and win each job in competition, and Quanta’s is about 10.4%, with an S&P return on capital of 9.1% against Prysmian’s 15.4%.
Quanta has the more visible backlog at USD 53.4 billion. Prysmian trades at 24.1 times model earnings and 12.5 times EBITDA, Quanta at 43.4 times and 24.9 times, and their base returns are 6.2% and 5.7%. Prysmian is cyclical, tied to metal and capacity, and has the weakest Canadian access. This pair illustrates the difference between an essential service and a good investment. Nobody doubts that cables must be laid and lines strung. The question is what the buyer pays for owning the people who do it, and 43 times earnings is a lot.
Siemens vs ABB
Siemens and ABB compete in similar markets but are built differently. ABB is focused on electrification, motion and automation, with Robotics being sold. Siemens is a diversified group of automation, rail, electrification and a majority stake in Siemens Healthineers, the medical-technology business, in which it held about 69% at 30 September 2025. Smart Infrastructure, about 29% of third-quarter orders at a margin of 20.0%, is the Siemens business most directly relevant to this comparison, so Siemens is not a pure play.
Siemens trades at 23.8 times model earnings and ABB at 33.4, and Siemens yields 1.98% against 1.23%. Its return on capital is far lower, 6.9% against 24.2%, though the two S&P figures are built on different structures. The model returns are 6.1% and 5.2%, and Siemens has the milder bear case, negative 1.2% against negative 3.0%.
The Healthineers separation complicates any like-for-like comparison. Siemens said in August that binding tax decisions allow it to proceed with the spin-off as planned, and a shareholder would receive shares in a separate listed company. The release does not give the size or date of the distribution, so my model treats Siemens on a look-through basis, and the return on Siemens shares alone would differ after it. If you want electrification, ABB gives you more of it. Siemens gives you a lower price and a larger set of other businesses.
What Canadian Investors Need to Know Before Buying
Everything above is quoted in each stock’s trading currency and before Canadian tax. For a Canadian the practical questions are where to buy, in which currency, in which account, and how much of the dividend a foreign government takes first. These can move a net return by amounts that are not small beside the gaps between my ratings. Table 7 is the country-by-country reference; the text covers the exceptions that matter and says plainly what I could not verify, because confident guesses in this area cost people money.
A US listing is not a US company
Dividend withholding generally depends on the paying company’s tax residence and the applicable source-country rules, not the exchange where its shares trade. My guide to dividend taxation in Canada covers the US rules and the foreign tax credit, and my comparison of the TSX, NYSE and Nasdaq notes that treatment depends on the issuer’s jurisdiction as well as the account.
Hubbell, GE Vernova and Quanta Services are US corporations. Under Articles X and XXI of the Canada-US tax treaty their dividends bear the 15% treaty rate in a non-registered account and in a TFSA, and no US withholding in an RRSP. My guide to TFSA mechanics explains why the TFSA’s 15% is unrecoverable. The amounts are small: GE Vernova pays USD 2.00 a share on a USD 926.44 price, and Quanta pays USD 0.44.
Eaton and nVent trade in New York and report in US dollars, but both are incorporated in Ireland, as Eaton’s annual report on Form 10-K confirms for Eaton. Incorporation, tax residence and actual withholding are three separate questions. Ireland’s 25% dividend withholding tax, described by Revenue, applies to dividends paid by Irish-resident companies, and I did not establish that Eaton’s dividends bear it. Revenue’s page on exemptions for non-residents says any exemption is not automatic: a treaty-country resident completes Form V2A, certified by the tax authority of the country of residence, and gives it to the company or its agent. nVent’s filing says it is tax resident in the United Kingdom, so its Irish incorporation does not tell me which country, if any, withholds. For both, the practical result for a Canadian holding through a broker is unverified. nVent’s dividend is small, about 0.6%, so the cost is low either way, but the answer should come from the company or broker in writing, and I would not assume US treatment for either, above all inside an RRSP.
The six European names
ABB, Schneider Electric, Siemens, Siemens Energy, Prysmian and Legrand are foreign issuers whose real markets are Zurich, Paris, Frankfurt and Milan. All six also have over-the-counter depositary receipts, and their terms differ more than the tickers suggest. ABB left the NYSE in 2023, and its only US line is a sponsored Level 1 programme on the over-the-counter market (ABBNY, one ADR for one share, Deutsche Bank Trust Company Americas as depositary), according to ABB’s investor page. Siemens Energy’s SMERY is a sponsored Level 1 programme with Citibank, announced in February 2026, and a 2026 SEC filing sets each ADS at one fifth of an ordinary share; the filing does not give the effective date of that ratio, so I do not state one. Siemens describes a Level I programme with Deutsche Bank in which two ADRs equal one share, with no ticker given, and I could not confirm SIEGY. The Schneider, Legrand and Prysmian receipts (SBGSY, LGRDY and PRYMY) are unsponsored, meaning a depositary bank created them without the issuer’s participation. The filings I found are old, and I could not confirm current ratios.
An ADR is a bank-issued certificate representing shares on deposit. It trades in US dollars, but the economics stay in the home currency and the dividend stays foreign-source: a receipt priced in US dollars does not turn a Swiss or French dividend into a US one. Ratios can change, over-the-counter lines are usually thinner with wider spreads, and depositary fees come out of dividends. My guides to over-the-counter stocks and ADRs and investing in global stock markets from Canada cover the structural risks and compare brokers, using ABB’s Swiss listing as an example. My default is the home-market share if the broker supports it, and the ADR only when it does not and the spread and fees look acceptable.
Registered accounts: legal eligibility versus broker acceptance
Whether a security is a qualified investment for an RRSP or TFSA is a legal question. Whether your broker will hold it there is an operational one, and the answers can differ.
On the legal side, the Finance Canada list of designated stock exchanges, last modified in March 2026, includes the SIX Swiss Exchange, Euronext Paris, the Frankfurt Stock Exchange, Borsa Italiana in Milan, the NYSE and Nasdaq. Over-the-counter markets are not on it. The definition of qualified investment in section 204 of the Income Tax Act includes securities listed on a designated stock exchange, and subsection 146(1) for RRSPs and subsection 207.01(1) for TFSAs both incorporate it. On my reading, shares of a company listed on a designated exchange are legally eligible for both. Paragraph 4900(1)(w) of the Income Tax Regulations prescribes “an American Depositary Receipt where the property represented by the receipt is listed on a designated stock exchange”, which on its face covers an ADR over a share listed in Zurich, Paris, Frankfurt or Milan. That is my reading and not a CRA ruling, and I did not confirm how the CRA treats ordinary shares bought over the counter. Holding a non-qualified investment carries tax and penalties, so get written confirmation from the broker first.
On the practical side the documented facts are few. Scotia iTRADE’s help centre lists OTCBB, Pink Sheets and CNQ securities among those it does not accept in TFSA, RIF or RRSP accounts, and names “foreign listed certificates”, so a Pink-market ADR is likely excluded and I could not tell whether the last category reaches other ADRs. Interactive Brokers Canada’s RRSP page and its TFSA page both restrict stocks to designated US and Canadian exchanges and cash to Canadian and US dollars, so no euro or Swiss franc holdings there. TD Direct Investing lists Paris, Frankfurt and Milan among its global markets but not Zurich, and does not say whether registered accounts can hold them. For every other combination, including RBC, CIBC and BMO, broker confirmation is required, and I cannot tell you whether yours will hold ABB in your TFSA.
Withholding tax on the European names
Table 7 summarizes the position by company. The statutory rate is the source country’s; the 15% treaty rate for Canadian portfolio investors comes from secondary sources, because I did not open the treaty texts. What is actually withheld I could not verify for any of the six, so assume the statutory rate comes out until the broker says otherwise in writing.
Table 7. Canadian access and dividend withholding by company. Rates are for dividends paid to Canadian-resident individuals. “Broker check” means I could not verify the point.
| Company | Issuer domicile | Where to buy (currency) | Source-country rate (statutory / treaty) | RRSP and TFSA | Open points |
|---|---|---|---|---|---|
| Hubbell | United States | NYSE (USD) | US 30% / 15% | RRSP: no US withholding. TFSA: 15%, unrecoverable | None of substance |
| GE Vernova | United States | NYSE (USD) | US 30% / 15% | As Hubbell | Dividend very small |
| Quanta Services | United States | NYSE (USD) | US 30% / 15% | As Hubbell | Dividend very small |
| Eaton | Ireland (incorporation) | NYSE (USD) | Ireland 25% applies to Irish-resident payers; exemption by declaration, not automatic | Not the US rule; actual withholding unverified | Tax residence, payer practice and broker route unconfirmed |
| nVent | Ireland (incorporation), UK tax resident per filing | NYSE (USD) | Unresolved (Ireland or UK) | Issuer confirmation | Which country withholds |
| ABB | Switzerland | SIX Zurich (CHF); ABBNY ADR (USD) | Switzerland 35% / 15% | Foreign tax is a cost, not a credit | Zurich access; ADR in a registered account is a broker check |
| Schneider Electric | France | Euronext Paris (EUR); SBGSY ADR (USD) | France 12.8% for documented individuals, 25% otherwise / 15% | Foreign tax is a cost; plan-trust rate unconfirmed | ADR unsponsored; current ratio unconfirmed |
| Legrand | France | Euronext Paris (EUR); LGRDY ADR (USD) | France 12.8% for documented individuals, 25% otherwise / 15% | As Schneider | ADR unsponsored; ratio unconfirmed |
| Siemens | Germany | Xetra Frankfurt (EUR); OTC Level I ADR (USD) | Germany 26.375% / 15% | Foreign tax is a cost | ADR ticker unconfirmed; excess reclaimable from the German federal tax office |
| Siemens Energy | Germany | Xetra Frankfurt (EUR); SMERY ADR (USD) | Germany 26.375% / 15% | Foreign tax is a cost | One ADS is one fifth of a share |
| Prysmian | Italy | Euronext Milan (EUR); PRYMY ADR (USD) | Italy 26% / 15% | Foreign tax is a cost | Unsponsored ADR; thinnest access; refund procedure unconfirmed |
Two consequences follow. First, inside an RRSP or TFSA foreign dividend tax is a cost, not a Canadian foreign tax credit: the CRA’s folio on the foreign tax credit gives no credit for foreign tax on income earned inside those plans, so the withholding is a permanent drag unless someone reclaims the excess. Second, in a non-registered account eligible foreign tax may qualify for a Canadian foreign tax credit, subject to the Canadian tax otherwise payable on the foreign income and other limits; withholding above the treaty rate may be reclaimable from the source country. A legal right to a refund is not the same as being able to obtain one, and I would not count on recovering anything until a broker confirms the procedure. The Swiss tax administration says a claim must be made within three years after the end of the calendar year in which the dividend fell due. Clearing-house notices, which are secondary and for France date from 2018 and 2022, say a French reclaim must be filed within two years after the year-end and a German treaty refund is a standard claim with no relief at source. I did not establish the Italian procedure.
The drag is modest because yields are low. ABB yields about 1.2%, so a 35% Swiss withholding costs about 0.4 percentage points a year before any refund and a 15% rate about 0.2, and for low-yield names I would care more about currency and account.
Currency and capital gains
A 10% stronger Canadian dollar over ten years would subtract roughly 1.0 percentage point a year from a foreign-currency return, and a 10% weaker one would add roughly the same. On that arithmetic Hubbell’s 10.2% base case becomes 9.1% or 11.2% in Canadian dollars. I make no exchange-rate forecast, and the Canadian dollar may behave differently against the euro and Swiss franc than against the US dollar.
Gains in a non-registered account are taxed at a 50% inclusion rate for 2026, as set out in my capital gains guide. Because most of the modelled return is price appreciation rather than dividends, capital gains tax on a successful holding can easily exceed the withholding drag, so the account is part of the decision. My guide to RRSP versus TFSA covers how foreign withholding bears on which account holds which asset. If a registered account is right and your broker will not hold the stock, a fund may be the alternative; my comparison of ETFs, mutual funds and individual stocks sets out the trade-offs.
What I Would Actually Do
Three decisions tend to get run together: which businesses I admire, which I would be willing to own for a decade, and which I would add to at the prices of 15 October 2026. They produce different lists. A compelling industry thesis is not a reason to pay any price for any company that touches it, and the model’s main finding is that in this group the thesis is mostly in the price.
On the model’s numbers, Hubbell and Legrand are the strongest candidates. Hubbell has the highest base return, 10.2%, and the best bear case among the eleven at 2.5%, and a Canadian can buy it on the NYSE in US dollars with ordinary US treaty treatment. Legrand’s base case is 10.0%, with a quality profile I like: a 50% payout, free cash flow close to net income and a broad low-voltage product range. The gap between them is inside the error of the model, so I would choose on practical grounds. Hubbell is simpler for a Canadian. Legrand trades in euros, depends on data centres for about a third of sales, and needs a broker check before it goes into a registered account. Neither is cheap, and I would watch Hubbell’s leverage after NSI and Legrand’s data-centre exposure.
Schneider Electric is interesting and less certain: a narrow Modest Buy, with the PTC acquisition materially changing its balance-sheet risk profile. ABB is the case where the distinction matters most, the best business in the group and a Hold at 33 times model earnings. The third-quarter results on 20 October, the Robotics sale and the Rotork outcome could change the picture, but the price would still need to fall or earnings rise for the arithmetic to work. GE Vernova and Siemens Energy are the two where I am most cautious, and I would own either only after a price decline, not on the assumption that the theme will carry them.
As a sequence of decisions, it would be this.
- Decide how much of your portfolio you want exposed to electrification at all, because the eleven companies share demand drivers and owning five is not five independent bets.
- Start with the businesses where the price does not ask for perfection: Hubbell and Legrand, and Schneider once the financing is clearer.
- Treat ABB, Eaton and Siemens as holds. If you already own them, there is no reason to sell on valuation alone, but there is also no reason to add at today’s multiples.
- Treat GE Vernova and Siemens Energy (both Avoids) and Quanta (a Hold at 43 times model earnings) as businesses to revisit after a price decline, and decide in advance what price would make the arithmetic work. The reverse-valuation table is the place to start.
- Before buying any foreign name, confirm in writing with your broker that it will hold the security in the account you intend to use, and find out what withholding tax will be applied.
Owning a company and buying more of it today are separate decisions. A Hold rating is no instruction to sell, and an attractive industry is no reason to start a position. The decision about additional capital is a decision about price, and price is where this group is least generous.
I applied the same lens of business quality against price paid to two railroads in my comparison of CN versus CPKC. The strongest opportunity is not necessarily the fastest grower. It is the business whose future cash flows justify the price paid today.
Conclusion
Electrification is likely to reshape infrastructure investment over many years, and the companies that make the equipment are, for the most part, good businesses. Business quality, growth expectations and purchase price are three separate questions, and this article is an attempt to keep them apart.
On my model, the most attractive opportunities are Hubbell, Legrand and Schneider Electric, and even they are Modest Buys and not Buys. The main valuation risks sit at the other end of the table, where GE Vernova needs 21% annual earnings growth, Quanta 15.5% and Siemens Energy 14.1%, all against base cases that are already high. The risk in all eleven is the same: that margins from scarcity pricing fade as capacity arrives, while the multiples assume they do not.
The model has limits that I tried to state throughout. Earnings are my own normalized estimates, terminal multiples are judgements, and several ratings, notably GE Vernova’s, move with those assumptions. ABB’s third-quarter results on 20 October 2026 are not reflected, nor are the Robotics gain, the Eaton and Dana Mobility transaction or any Siemens Healthineers distribution. A Canadian buyer faces account and withholding questions that I could not fully verify and that a broker must answer. What I am confident about is the shape of the result: demand growth does not translate automatically into shareholder return.
Investment risk and disclosure. This article is my personal analysis for general information and is not personalized investment, tax or legal advice. The returns shown are model outputs under stated assumptions, not forecasts or promises, and actual results may differ materially, including losses of capital. Prices are closing prices on 15 October 2026 and will have changed. Company information comes from public disclosures and, where marked, secondary sources. I work for ABB. The views in this article are my own and do not represent ABB’s, the article is independent of ABB, and it uses only public information. My professional familiarity with ABB, and my closer attention to ABB, Siemens and Eaton, did not determine any rating. Tax treatment depends on your circumstances, and the points on registered accounts, qualified investments and foreign withholding should be confirmed with your broker, the CRA or a qualified adviser before you act.