GE Vernova

IThe business

GE Vernova is the power half of the old General Electric. It was spun off on 2 April 2024, leaving GE Aerospace behind.⁷ Its customers’ machines generate about a quarter of the world’s electricity.⁷ Scott Strazik puts it higher at home: over half of the electrons in the United States on a given day, and a third of those outside China.¹³ It has about 85,000 people, in about 100 countries, after buying the rest of Prolec GE.²

It reports three segments. In the first nine months of 2026, Power was $16.5B of revenue, Electrification $10.5B and Wind $6.2B.³ Segment EBITDA, a rough cash profit before interest, tax, depreciation and amortization, was $3.1B, $1.9B and a loss of $0.6B.³ So two of the three make all the money, and the third is a problem being managed down. The 2026 guide is $46.5B to $47.5B of revenue, an adjusted EBITDA margin of 13.5% to 14.5%, and $16B to $17B of free cash flow.²

Power

Power is gas first. Gas Power was $13.6B of the segment’s $16.5B in nine months, Nuclear $2.3B and Hydro $0.6B.³ A gas turbine is a jet engine’s heavier cousin bolted to the ground. Heavy-duty units, including the large HA class, run for decades in big plants. Aeroderivatives are smaller, start fast, and come from aircraft engine designs. In the third quarter GE Vernova took orders for 55 heavy-duty units, 21 of them HA, and 57 aeroderivatives.²

The money in this business has always come after the sale. Services were $9.8B of Power’s nine-month revenue, nearly 60%, against $6.7B of equipment.³ A turbine has to be opened up and refitted on a schedule for its whole life. Management says each gigawatt of HA turbines brings in $0.5B of high-margin service revenue over its first 20 years, about $25M a year, and it expects at least 400 HA units running or on contract by the end of 2027.⁴ At the December 2025 update it set out Power services growing from about $12B to $22B by 2035. Strazik now says that will likely prove conservative.¹⁵⁴

Gigawatts are the number to watch. A gigawatt is about what a large nuclear reactor or a mid-sized city needs. Under contract means firm orders plus slot reservation agreements, where a customer pays to hold a place in the factory queue before a full order. That total rose from 83 GW at the start of the year to 122 GW at 30 September: 64 GW in backlog and 58 GW in reservations.²⁴ Management expects at least 130 GW by year end. In the third quarter it sold slots for delivery in 2032.⁴

The price of those slots is what has changed. Full-year 2026 gas orders are running about 50% above 2025’s on a dollars-per-kilowatt basis.⁴ Only a small part of that has reached the income statement. Strazik says the third quarter was the first with a full quarter of better-priced shipments, at the 5 GW quarterly pace.⁴ In January he said most of the higher-margin orders from 2024 and 2025 won’t ship until 2027 and beyond.¹⁶ That gap, between what has been sold and what has been earned, is why this stock has been so strong.

Electrification

Electrification sells what moves power once it’s made: transformers, switchgear, high-voltage direct current links, substations and the devices that keep the grid stable. It had $46.1B of backlog at 30 September, up from about $7B at the start of 2022.³⁴ Revenue was $3.9B in the third quarter, 51% higher, or 16% without Prolec and other changes. The margin was 19.1%.²

Two deals are reshaping it. In February 2026 GE Vernova paid $5.2B for the half of Prolec GE, a North American transformer maker, that it didn’t own.³ On 25 October it agreed to buy Trench Group from Triton Partners for about €8.0B, $9.0B at its stated rate, with close expected in late 2027.³² Trench makes bushings, instrument transformers and coils. Those sound small. A bushing is the insulated part that lets a live conductor pass through a transformer’s grounded case, and bushings are 5% to 8% of a transformer’s bill of materials.⁴ Management says it today buys most of its bushings from others. Data centers are growing fast here too: over $6B of segment orders so far in 2026.⁴

Wind

Wind is the part nobody bids the stock up for. It lost $585M of segment EBITDA in nine months, and the 2026 guide is a loss of $400M to $500M.² It made $72M in the third quarter, helped by tariff recoveries.² Onshore has been the steady part. Offshore has been a costly lesson: the Vineyard Wind dispute settled in the third quarter, and Dogger Bank has about a year to run.³⁴ US orders are thin. Strazik says the company has been about half of a North American onshore market that is now very small, and that he wants to see where it goes before deciding how to create value in this business.⁴ I take that as an open door to a sale or a shrinking, though he didn’t say so.

Where the demand comes from

This is less of a pure AI story than the stock trades like. On gas, Strazik says the 122 GW spans about 100 customers in 25 countries, about 80% traditional utilities and power producers and about 20% for data centers.⁴ US revenue was 46% of the total in 2025 C (calculated).⁷ Asia is contracting for plants that will start in 2032 and 2033.¹⁷ That diversity is a real comfort. It is also a little misleading, because the utilities are ordering partly for the same data-center load, one step removed. I can’t split that out, and I don’t think anyone can.

IIThe moat

Two businesses share this section, and they deserve different grades. Services are protected. Equipment, today, is scarce. Those are different things, and the price treats them as one.

Start with services. GE Vernova has about 7,000 gas turbines in its installed base. About 1,800 are on long-term service agreements with an average of about 10 years left.⁷ Service backlog is $92.7B, of which 16% falls due within a year, 54% within five, and 78% within ten.³ That is about two years of company revenue, contracted for maintenance on machines already sold or being built. Customers can go elsewhere for parts and repairs, and some do. But an owner with a billion-dollar plant doesn’t lightly risk it on a third-party hot section. The company’s own forecast has this business growing by about $10B in ten years, and that comes from the fleet being doubled, from about 200 GW running baseload to about 400 GW.¹⁵ This stream is the closest thing here to an annuity. It is also what survived both earlier collapses.

Equipment protection is the shortage

Now the equipment. Only three makers can supply large gas turbines at scale. That has been true for decades, and it didn’t stop a $22.0B goodwill charge at GE in 2018.⁶ So structure alone does not protect profits here. Demand against capacity does. When GE wrote in its 2018 annual filing that heavy-duty orders would run at 25 to 30 gigawatts “for 2019 and the foreseeable future,” the industry was fighting over too few machines.⁶ Today the head of Siemens Energy, as Utility Dive reports him, says the market could reach 120 GW a year.⁸ The same three firms, more or less the same machines. Margins moved because the gap moved.

Evidence it is widening

Customers now pay up front, and pay more. Contract liabilities, the cash taken in ahead of delivery, rose $19.8B in nine months, in part from Prolec.³ The company also says the 2026 increase in equipment margin in backlog will be substantially larger than the roughly $8B added in 2025.⁴ It didn’t give the rate, and promises one in January. Services pricing is rising too, with a lag.¹³⁴ And management has so far added capacity carefully: 20 GW a year today, 24 GW in 2028, 30 GW in 2030, mostly inside existing factories, without a new greenfield plant.¹⁹ Discipline of that kind is what the industry lacked last time.

Evidence it is narrowing

Discipline is easier to show when you’re sold out. Siemens Energy could ship 15 to 16 GW this year, and is adding about 15 large and 20 medium units of annual capacity by 2028.⁸ Mitsubishi plans to double production from its 2024 level.⁹ GE Vernova’s own 30 GW in 2030 would be about the size of the whole heavy-duty market GE was planning for in 2020.²⁰ Strazik says that on heavy-duty, he has good views of industry capacity for six years and feels balanced. He is less sure how much smaller machines will add, from makers trying to take advantage of the gap.¹⁹ That is the right place to worry. High prices invite substitutes: reciprocating engines, fuel cells, batteries, and small turbines repurposed for data centers.

In the grid the walls are lower. An analyst on the December call pointed out Chinese, Korean and Brazilian firms adding transformer capacity in the United States.¹³ Margins in Electrification are good and rising, but this is a crowded field: the 10-K lists Hitachi Energy, Siemens Energy, Siemens, Schneider Electric, Mitsubishi Electric and ABB.⁷

My judgement: a strong, durable service business sits under a very profitable equipment boom. I would pay a high multiple for the first. I can’t say how long the second lasts, and neither can the company. Its own risk factor says slot reservations may not result in orders.⁷

IIIThe competitive advantage

Within the industry, GE Vernova is the largest and arguably the best placed. The closest listed comparison is Siemens Energy, which sells gas turbines, grid gear and wind turbines in nearly the same mix, with the same wind problem. The table compares the latest quarter of each. Siemens Energy’s fiscal year ends in September, so its figures are a quarter older, and it reports full-year results after this page’s date.

The last column stopped me. On one quarter’s profit, annualized, the market pays more than twice as much for GE Vernova. Part of that is real: a larger fleet and more gigawatts under contract. Part may be a US listing. Segment margins won’t settle it: Power’s 21.4% is before depreciation and Siemens Energy’s Gas Services’ 17.3% is after it, so the two aren’t comparable.²²¹ I can’t split it. Either way, a buyer who wants this boom has a much cheaper way in on this measure, with a Wind unit that is also just turning positive.²¹

Where GE Vernova wins

Scale of fleet first. Strazik’s point about capacity shifting to services in the 2030s only works for the largest fleet: by then the factories that build new units will also make the parts for their first outages.¹⁹ That gives the company a reason to build capacity that a smaller rival doesn’t have. Scope second. It can sell a hyperscaler turbines, transformers, switchgear and software together, and it uses scarce aeroderivatives as a lever: Strazik said on the third-quarter call that the company uses them in partnerships to secure heavy-duty and Electrification business as well.⁴ US manufacturing third. Greenville, South Carolina, builds the heavy-duty machines, and Charleroi, Pennsylvania, makes circuit breakers, at a time when governments care where these things come from.¹⁵

Where it loses

Wind, plainly. It absorbs management time and shows no return. In the grid it is one strong firm among several, with no fleet to lean on the way Power has. And in nuclear it has a design and a long way to go: the small reactor project in Ontario is under way and TVA’s Clinch River has its construction permit, but neither is a profit yet.⁴ I give small reactors no specific value in my base. If they work, that is a free option, and a costly one to develop.

IVManagement

Scott Strazik, 47, has run these businesses since November 2021 and became president as well in July 2025. Before that he ran GE Gas Power from 2018 to 2021 and Power Services before that.¹⁰ This matters more than it first appears. He took over gas at the bottom, in the wreck of the 2018 charge, and spent three years sizing the business for a market of 25 to 30 GW.⁶²⁰ Few chief executives in this boom have seen what the other side looks like from the inside.

The team is turning over at a good moment, with two of the top jobs changing while results are strong. The new CFO comes from an electric-truck maker, a very different problem from having more cash than you can use. Her offer letter carries a $14.5M make-whole stock award and a $5.0M cash signing payment, on top of a $5.2M yearly stock target.²³ That’s to replace what she forfeits. I note it, and I’ll judge her by the first capital decision.

How they are paid, and what that pays them to do

Follow the incentive. Strazik’s reported pay for 2025 was $18.0M, 282 times the median employee: $1.5M of salary, a $4.5M bonus and $10.7M of stock and options.¹⁰ The bonus plan weighs free cash flow 40%, adjusted EBITDA 40% and organic revenue growth 20%. The long-term performance shares weigh three-year cumulative free cash flow 50% and adjusted EBITDA 50%, and a relative share return test moves the result up or down by 20%.¹⁰

Now look at what that measures. The 2025 free cash flow target was $2.5B and the maximum $3.5B. The result was $3.71B, which paid 200% on that line.¹⁰ In that year contract liabilities rose $8.0B.⁷ So the rise in customer down payments was more than twice the top of the range. In 2026 they’ve done it again on a far greater scale. I am not suggesting anything improper. The plan is what it is, and the deposits are real money. But it pays management heavily for a number that rises when customers pre-pay, and that reverses later, after the three-year window. Nothing in the plan measures return on capital, or cash after the deposits unwind. If I wrote this plan, I would measure cash excluding the change in contract liabilities.

In fairness, the 2025 plan did dock pay for safety. The company had four worker fatalities, and the board cut the modifier by ten points, though the bonus still reached the 200% cap on an individual score of 130%.¹⁰

How much of their own money is in

A good deal. Strazik held 152,282 shares after his last reported transaction, about $131M at this price, and a guideline of six times salary would need about $9M C (calculated).¹¹¹⁰ I read 97 Forms 4 from April 2024 to now. He has reported no open-market sale. Every disposal was shares withheld for tax on vesting or option exercise. No insider made an open-market purchase in that time. A few other officers sold small amounts: the largest, by Maví Zingoni, was about $7M.¹¹ In April 2026, at about $1,115, Strazik exercised 44,496 options and kept the net shares after tax, with the stock not far from its high.¹¹ Keeping them says more than a speech would.

What they did with capital

On a partial record, I think this team allocates capital well. Prolec looks well bought. The capacity plan has been careful. The 2025 buybacks, at about $400, were excellent. The question is what happened to the discipline near the top. Of the $9.9B, $6.6B has gone out in 2026, when the stock was highest. That is the usual pattern for corporate buybacks and it’s no credit to anyone. A company that owes its customers $45.7B of goods and holds $16.8B of cash is using a form of interest-free customer finance to retire stock at 35 times next year’s earnings.

Trench is the larger question. It may well be right: control of a critical part, some 80 framework agreements, and a 36% margin.⁴²⁴ But it is a high price. Trench was independent only from April 2024, and revenue is estimated at $1.4B for 2026.²⁴⁴ The $50M to $100M of cost savings is small against $9.0B, so the deal needs growth or pricing to pay.⁴ Over a full cycle these people haven’t yet been tested on capital. The old GE, whose record they inherited, bought Alstom’s power business and wrote off $22.0B of goodwill. Strazik didn’t make that decision, and he helped clean up after it. What I would want to see is a large buyback in a downturn, when the stock is cheap and customers’ cash is draining. That is when this kind of promise is hard to keep.

VFinancials

GE Vernova reports audited numbers only from 2021, because that’s how far the Form 10 carve-out goes. The older record is in GE’s own filings, under different segments, and I come to it below. The first thing to know is that every headline number in this table is distorted in some way.

Revenue is growing at last. It was flat to down from 2021 to 2022, up 12% in 2023, 5% in 2024 and 9% in 2025, and up 22% so far in 2026, or 13% organically.¹⁸⁷² Margins are the better story. Adjusted EBITDA went from a loss in 2022 to 8.4% in 2025 and 15.4% in the third quarter.⁷² Gross margin in that quarter was 24.6%.³ For the first time since the spin-off, the core of this business earns a good return on sales.

Free cash flow is mostly other people’s money

This is the most important paragraph on this page. Free cash flow was $15.3B in the first nine months of 2026, more than three times adjusted EBITDA.² That can’t come from operations. The cash flow statement shows the source: contract liabilities rose $19.5B. Working capital as a whole added $17.0B.³ Take working capital out, and free cash flow for the nine months was about −$1.7B C (calculated). Cash tax of $2.7B against a book charge of $1.1B, and a $350M retiree payment, make that look worse than the run rate.³² Even so, the plain fact is that customers funded all of this year’s cash, the buyback and the Prolec purchase too.

Down payments are a sign of strength. Customers don’t pay in advance for what they can get elsewhere. They’re also a loan with no interest, repaid in turbines. As those turbines ship, revenue is recognized and no new cash comes in for that part of the price. So the cash boom comes first, and the profits come later. Management has said as much: first-half cash would run well ahead of the second as reservations convert to orders.¹⁹ The 2026 guide, with free cash flow of $16B to $17B, implies about $0.7B to $1.7B in the fourth quarter.²

What an owner can really take out

So here’s my number. I take the middle of 2026 guidance, about $6.6B of adjusted EBITDA, which is after stock pay, since the company doesn’t add it back. I take off about $1.2B of tax at 24% on operating profit, and about $0.3B of the restructuring and separation costs that adjusted EBITDA leaves out. Then I take off the capital needed just to stand still. Depreciation of about $0.75B is my proxy for that, since the company reports only total spending. That leaves about $4.3B for 2026 E (estimate).

That’s about 53 times owner cash for this year. It’s a poor measure of what’s coming, since the profits are still on their way. My base has owner cash before working capital of $8B in 2028 and $14B in 2030. But it shows how far the headline numbers sit from what is yet earned: reported free cash flow is nearly four times what an owner could take out and leave the business standing.

The balance sheet

At 30 September cash was $16.8B and long-term debt $2.8B, the $2.6B of senior notes issued in February and leases.³ Fitch moved its rating to A− in August.² Equity is $12.5B, and that’s after $9.6B of goodwill and $4.3B of intangibles.³ At $228B the market pays about 18 times book, but book means little here, because the assets that matter are fleets and factories long written down.

The liability that matters is contract liabilities of $45.7B, of which Power is $33.2B.³ These aren’t debts in the usual sense, since they’re settled with goods. Asked on the call whether customers were asking to exit slot reservations, Strazik said no, and that cash terms had not changed.⁴ But what a customer gets back on leaving is in neither the filings nor the call. I can’t tell you how much of the deposit cash would stay with the company if a customer walked. At GE in 2002 such exits gave rise to $0.9B of net termination fees, and they showed up as profit, on the way down.⁵ Other items: warranties of $1.7B; $1.0B of indemnities; and an open IRS audit of GE’s 2016 to 2020 returns that could fall partly on GE Vernova in an amount it can’t yet estimate.³

The long memory

The company’s record starts in 2021. The turbine cycle’s doesn’t, and GE’s filings tell it plainly. In the 2004 annual report GE described demand that “peaked in 2002, a period often referred to as the ‘U.S. power bubble.’”⁵ Look at what followed.

Units fell 62% in two years. Profit fell 55%. Revenue fell only 27%, because services kept growing, and that’s the good news in this table. Notice the timing, too. Orders collapsed in 2002, the year revenue and profit set records. A reader who watched earnings would have seen the best year ever, with the order book already falling away.

The second fall was 42% in one year. GE had just bought Alstom’s power and grid businesses, in November 2015, and within three years it was writing off the goodwill.²⁸⁶ Then, in 2020, it was still planning for 25 to 30 GW.²⁰ Six years later, one maker took orders for 36.7 GW in nine months.³ That’s the lesson of these tables, and it runs both ways. People inside this industry, with the best information anyone has, have been badly wrong about the next five years, at both turns.

VIValuation

First, how this goes badly wrong

Before any numbers, I asked what would most likely ruin this investment, and worked back from there. Three things can do it.

• The order book turns out to be a queue, and queues empty fast. Customers who fear being left out reserve early. Some reserve more than they’ll need. The 10-K says plainly that reservations may not result in orders, and that orders may be deferred, reduced or cancelled.⁷ In 2002 GE’s orders fell 42% in a year, and customers paid to get out.²⁵ Third-quarter additions to gigawatts under contract, at 6 GW net, were already the smallest of the year, though output was the highest.²²⁹
• Power demand from AI falls short, just as all three makers’ new capacity arrives. The 2030 capacity at GE Vernova, Siemens Energy and Mitsubishi is being added for a market far above anything seen. Data centers are about 20% of gas contracts.⁴ Yet they set the marginal price, and the customers’ own financing is the swing factor. Asked on the call if financing was a problem, Strazik said no.⁴ The stock fell 10.6% on 22 October, days before results, in what looked to me like a scare about AI spending, though I couldn’t confirm the cause.¹
• The cash goes out the door before the deliveries do. The company holds $16.8B and owes $45.7B of goods. It has approved another $10.1B of buybacks and committed to $9.0B for Trench.²³ If deposits run off in a downturn, the buyback and the deal would have used the customers’ money at the top, and a company with an A− rating would be borrowing to build machines it has already been paid for. This is how an excellent quarter can become a poor decade for an owner.

None of these needs a surprise. Each has happened in this business, and two are in the company’s risk factors. They set my bear case, and my reason for keeping the 2030s below the 2020s. An industry peak with no later decline is the one thing its history has never shown.

What the price implies

At $857.87 the market values GE Vernova at about $228B on 266M diluted shares, after the 1.4M bought in October.²¹ That’s 35 times Nasdaq’s consensus for 2027, 26 times 2028 and 20 times 2029, though only four analysts reach that year.¹² The mean target is $1,173, from 18 analysts, so the Street sits well above this page.³⁰ The stock is 27% below the $1,174.86 close of 30 June.¹

A single growth rate means little when one number, gas output after 2031, carries so much. So I take my base cash through 2031 and ask what the business must make after that. Those five years, after paying for Trench, are worth $32B today. The price pays for $180B beyond that. It needs $20B of owner cash a year from 2032, growing 3%, for good C (calculated). My base has $14B to $15B. The most this company reported in free cash flow in any year before 2026 was $3.7B.⁷

Put another way, on my base cash flows the price equals value at a discount rate of about 8.2%. That is the plainest way to read this page. If you accept about 8% from a long-cycle equipment maker, near the peak of its order book, the price is fair. I don’t. And more volume alone doesn’t rescue it. Shipments would need to settle near 62 GW, about twice today’s industry-leading plan, because extra gas at my margin doesn’t add as much as people expect once tax and capital are paid for.

What I can’t judge

Two inputs decide this page, and I can’t know either. The first is how many gigawatts a year the world buys from GE Vernova after 2031. Nobody has that information. The 2030 slots are sold, the 2032 ones are being sold, and customers’ plans that far out are hopes with deposits attached. The second is the normal margin for Power, because no one has run this mix before: 2030 equipment pricing about double what ships today, on top of a far larger service fleet. So I show the range.

I also can’t tell how much of the deposit cash stays if a customer leaves. And I can’t judge whether the demand forecasts are right, or if data centers will use as much power as their builders expect. I’d rather say so than dress it up.

The discount rate

I use 10.0%, the same as on my NVIDIA page, which is a beta of about 1.2. The capital asset pricing model starts with the 4.94% ten-year Treasury in Q4 2026 and Damodaran’s 4.20% implied equity risk premium.³¹³² GE Vernova has only 31 months of trading. On month-end returns its beta is 0.97, which gives 9.0%. On daily returns it’s 1.90, which gives 12.9% C (calculated).¹³³ Neither means much on so short a record, in a stock that has traded as an AI name. A business with a long backlog and a service annuity earns something lower than a chip maker’s rate. One with this history of order collapses shouldn’t get much lower. I take 10.0%. It’s the third most important input on the page, and the table shows it.

Three cases

I build the model by segment. For Power I multiply gigawatts shipped by equipment revenue per kilowatt, and add services. That revenue per kilowatt starts at about $530 in 2026, Power equipment revenue over gas gigawatts, which also carries nuclear and hydro equipment. It climbs through the backlog: Power’s equipment backlog of $54.4B against 64 GW is about $850 a kilowatt, and the rise in that backlog this year, plus shipments, works out near $990 C (calculated).³ Electrification grows at a rate I set each year, Trench comes in from 2028 on the company’s forecast at 36%, and Wind follows its outlook.²⁴³⁴ I take off corporate costs, tax at 24%, capital spending of 3.5% to 5.5% of revenue and $0.3B a year of restructuring. Stock pay is already a cost in adjusted EBITDA.

Customer money runs through working capital. Receivables, inventory and contract assets, less payables and contract liabilities, were −$22.5B at 30 September, nearly half of a year’s sales C (calculated).³ I let that fall to −23% of revenue by 2035 in the base as output catches up with orders, and −14% in the bear. The $9.0B for Trench goes out at the end of 2027. The company’s by-2028 outlook, $56B of revenue, 20% margin, Power and Electrification each at 22%, was set in January, and management calls its targets floors; my base is above it at $64B and 22.6%, since 2026 is already well ahead.³⁴⁴

Note how close I am to consensus through 2029. This page doesn’t doubt the next three years. It says that buying at 35 times 2027 earnings pays for those years in advance, plus a good part of the decade after. My terminal value is about 9.8 times 2036 adjusted EBITDA, which is a sober multiple for a company with so much of its profit in services by then. Someone paying 15 times would add about $180 a share. That’s a real argument, and I show it so a reader can use it, though I count it as hope about a multiple, not as cash.

Other uses of the same dollar

Price only means something next to what else the money could do. The ten-year Treasury pays 4.94%, without risk of default. My base returns 5.9% a year, so about one point over bonds, for a company with about a seventh of its value in the next five years and the rest in the fog E (estimate).³¹ The best comparable on this site is ASML: also the sold-out maker of a machine with a large service fleet, and few rivals, a far stronger position. It returns 7.4% a year on my base there and 5.0% on my bear, and its price sits 25% above base value. TSMC returns 9.4% on my base and 5.3% on my bear, and NVIDIA 9.7% and 5.5%. Those pages use their own prices and rates. All three sell the same demand story with better records through bad years. Against GE Vernova’s −3.0% bear return, they look safer. Siemens Energy, at less than half the multiple of current profit, is the nearest direct alternative, and I have no page on it yet.

The margin of safety

• Price$857.87Close, Q4 2026 [1]
• Base value$638My estimate E (estimate)
• Margin of safety below$412set from my bear case C (calculated)

I set the margin of safety from the bear case, because the cycle drives this business and the service book sets the floor. Below $412, 35% under my base value, my $309 bear value is at least three quarters of the price paid. My base would return about 17% a year and my bear 5.9% E (estimate). The stock last closed below $412 in May 2025, about 18 months ago, before the order surge.¹ So this isn’t a fantasy price, but the facts have improved a lot since. The price sits 108% above it, and there is no margin of safety at $857.87. The cushion covers what the model doesn’t hold: the IRS audit, a warranty failure on a new class of turbine, and my own error. A worse repeat of 2002, in which output falls to 8 GW and deposits run off to 10% of sales, gives $203 E (estimate). That stress case isn’t covered.

Two different questions

How good is the business, and how good is the price? They have different answers. This is the best GE Vernova, or any version of GE’s power business, has been in at least twenty years: sold out, pre-paid, repricing its whole book, with a large annuity under it and a capable team. The service fleet alone would be a fine thing to own at a fair multiple. The price is another matter. At 34% above my base it pays in full for the boom, and for much of what comes after. So this looks like a very good business, at a price that leaves the buyer little to be wrong about. Those are two separate conclusions, and the first doesn’t lead to the second.

Where this could be in 5 and 10 years

For the base to stand at 5 years, GE Vernova has to ship 30 GW of gas turbines a year in 2030 and 2031 at close to $1,000 a kilowatt, and reach $88B of revenue and $15B of owner cash in 2031. For it to stand at 10, output has to settle at 21 GW from 2033 and Power's margin at 29%. My base value goes from $638 today to $735 at the end of 2031 and $869 at the end of 2036. The owner cash paid on the way is $436 a share. All of the year-10 value is cash after 2036, the part I know least E (estimate).

The bear and the bull differ on gas output and Power's margin. In the bear, the order book proves to be a queue, output falls to 12 GW by 2036 and Power's margin returns to 20%. That is $425 at year 10 and a loss of 3.3% a year from $857.87. In the bull there is no bust. Output reaches 36 GW, Power earns 36%, and revenue passes $141B. Capital needed is the same kind in all three: capital spending of 3.5% to 5.5% of revenue, and $9.0B for Trench in 2027. The discount rate is 10% in all three. At 8.5%, my base value today is $816 E (estimate).

The most likely way the base is wrong is gas output after 2031. I have it settling at 21 GW, nearly twice 2024. This company's turbine market has collapsed twice, and I can't see the market of 2033.

These are estimates of value, not forecasts of the share price, and the market price can sit above or below value for years.

VIIThe case for the price

What most people miss

Three things cut in the company’s favour and don’t make the headlines. The growth is paid for by customers. The step to 30 GW in 2030 comes from machines in existing factories, with supply secured, “all funded by customer down payments.”¹⁹ Capital spending is about 5% of revenue. So on the capital it has put in, this is close to free growth for now. Next, the factories have a second life. Strazik says the 325 or so HA units commissioned or on contract reach their first outages in the middle of the next decade, and that capacity will then be needed for parts.¹⁹ In past busts the factories had nothing to do. In the next one the service wave may arrive to fill them. And third, Power made 21.4% in what is its weakest services quarter, with its first full quarter of better-priced shipments.⁴ The margin on the book being built this year doesn’t show until January.⁴

The bull case

In plain words: electricity has become the limit on growth, few firms can supply it, and this is the start of a decades-long build. Strazik calls it the early stage of an electricity investment super cycle, and says his 2030 guide, when it comes, will be a floor.⁴ Rivals back the story. Siemens Energy’s chief says, as Utility Dive reports it, that the turbine market could reach 120 GW a year, and Mitsubishi Power, as Bloomberg reports it, is talking about delivery around 2031.⁸³⁵ The Street’s mean target is $1,173, and the high is $1,450, as Nasdaq reports it.³⁰

The holdings show who else is in. Coatue held 2,558,164 shares at 30 June, down from 3,370,174 at the end of 2025. Tiger Global had 797,511, down from 972,994. Fundsmith, Terry Smith’s firm, opened a new position of 409,835 shares in the quarter.³⁶³⁷³⁸ Those filings are four months old and don’t say why. Fundsmith’s buying is the most telling, since that firm is known for owning high-return companies and leaving them alone. I could not open its letter.

Against reality

Today’s numbers support the near part. Orders, backlog, margin and cash all beat this year’s plan, and the 2026 free cash flow guide has gone from $4.5B to $16B or more in less than a year.¹⁵² If they ship 30 GW in 2030 at about $1,000 a kilowatt, reach 30% margins in Power and 25% in Electrification, and keep services compounding, then $55 of earnings in 2030 is possible, and the price can be defended on that year. That’s my base. So delivery through 2031 isn’t really the argument.

The argument is what happens next. The price sits about halfway between my base and bull. To get there, the 2030s must look much more like the late 2020s than the early 2030s have ever looked after a turbine boom. It helps to multiply it out. Say orders stay above shipments through 2031, 65%. Gas output then settles above 30 GW and doesn’t fall back, 35%. Price per kilowatt keeps above $1,100 with three makers at full stretch, 50%. Power reaches a margin of 35%, 60%. Electrification reaches 28% to 30%, 55%. Those are my guesses, and they aren’t independent, since strong demand helps most of them. Multiplied, they give 4% E (estimate). Allowing for the overlap, perhaps 20%, which is the weight I give the bull. Each step looks reasonable. All five together do not.

What a 10x would take. Ten times $857.87 is $8,579, about $2.3T of market value. At 25 times earnings that needs about $91B of net income. At an 18% net margin, well above anything in my base, it needs about $510B of revenue, eleven times the 2026 guide, and more than the $300B Strazik puts on the whole electrification market in 2030.⁴ By 2036 the share price would have to compound at 26% a year. GE Vernova has already risen more than six times since its first close at $131.25, but that began at about $36B of value.¹¹⁸ From about today’s size, the best run I know of is NVIDIA’s: ten times from about $310B in about three and a half years, selling chips at software-like margins.³⁹ A maker of turbines and transformers would have to do it with factories, long lead times and margins in the 20s. It is close to impossible, and I would not build a case on it.

Where I come down

The down payments and the service fleet moved me: without them my settled margin would be several points lower. The headline backlog did not, since backlogs have fallen fast here before. Evidence that orders can stay above 25 GW a year after 2031 would raise my values most, and framework deals for 2031 to 2035 with real money behind them would be the first sign.

VIIIRisks, and what would change my view

The three ways this goes badly wrong are at the start of Valuation. These are the rest.

• Bottlenecks outside the factory. A turbine doesn’t make power until the contractors, pipelines and permits are ready. Strazik says the company won’t book orders without a firm path to schedule, and that this will decide the pace of conversions.¹⁹ Slower projects mean slower deposits.
• Quality. Output has risen by about a third in a short time, with hundreds of new tools, and the newer HA machines have the least field history.¹⁷⁷ The warranty liability is $1.7B.³ The old Vineyard Wind blade removal shows what a field failure costs, and a problem in the HA fleet would touch the most valuable part of the company.⁷
• Substitutes. Batteries, solar, engines, fuel cells and small turbines are all being pushed by the same shortage. Each dollar of scarcity is a dollar of incentive for someone else.
• Policy. Tariffs cost under $100M in 2026 after recoveries, an estimate the company says may prove quite different.³ Wind depends on permits and on rules that have already changed. Emissions rules can shift the economics of gas in any market.
• Trench. It closes in late 2027 if regulators agree, and GE Vernova pays a fee if they don’t.³ It’s also a big bet on a fixed euro price near a high point for the sector.
• Legacy. The IRS audit of GE’s 2016 to 2020 returns may leave GE Vernova with a bill it can’t estimate and hasn’t accrued, and $1.0B of indemnities remain.³
• Offshore. Dogger Bank has one more year to go.⁴ Offshore wind has repeatedly cost more than planned.
• Safety. Four workers died in 2025.¹⁰ The third quarter had none.² This is a heavy industrial company at full stretch, and it’s not just a pay modifier.
• The stock itself. It fell 38% in less than two months in early 2025, and 27% from June to now.¹ Even if I’m right on value, the path can be violent both ways.