GE Vernova (GEV) — The Customers Paid First

GE Vernova's third quarter booked $29.3 billion of orders, up 96% organically, and raised the 2026 free-cash-flow guide by $4.5 billion. The 10-Q shows where the cash came from: customer deposits added $19.5 billion in nine months, more than all of the free cash flow. That money is real, but it is paid against turbines and grid equipment still to be built. Counted as timing rather than earnings, it leaves GE Vernova worth $534.79 a share, from $672 in July, against a price that needs the sold-out market for gas turbines to last until about 2048.

TON618 Capital Research · As of 27 October 2026 · Download PDF · Download model

In July we valued GE Vernova as the best business in the AI power buildout, at a price that already assumed the boom lasts a decade. The third quarter made the business case stronger: some gas-equipment slots are now sold for 2032 delivery, backlog rose $20 billion in three months to $197 billion, and margins rose in all three segments. It also made the cash flow harder to read. The quarter's $5.4 billion of free cash flow, and the raised $16–17 billion for the year, rest on down payments from buyers holding a place in the production line. This note asks what that cash is worth to a shareholder, and what $895.08 requires once it is counted as timing rather than earnings.

Terms used throughout, defined once. Orders are new firm contracts; backlog is contracted revenue not yet earned (the remaining performance obligation). A slot reservation is a paid agreement that holds a gas-turbine production slot before a firm order; the company reports gigawatts (GW) of backlog and reservations together as "under contract". Customer deposits are contract liabilities: cash received ahead of the revenue it pays for. Adjusted EBITDA is earnings before interest, tax, depreciation and amortization, as the company adjusts it (a non-GAAP measure). Free cash flow (FCF) is cash from operations less capital spending; earnings FCF is our estimate of the part that comes from profit rather than deposits; free cash flow to equity (FCFE), the cash a shareholder could be paid, is the basis of the fair value. Growth stages are forecast periods, each lasting as long as its driver. ROIC is after-tax operating income over invested capital, trailing twelve months, against the weighted average cost of capital (WACC). Terminal growth is the rate a business grows at forever after the forecast, set from long-run GDP. Fair value is the FCFE value weighted across bear / base / bull at 25 / 45 / 30; the EV/EBITDA value is a cross-check (§4).

Revenue was $12.64 billion, up 27% reported and 18% organically. Net income was $1.29 billion, or $4.81 a diluted share against $1.64 a year earlier. Adjusted EBITDA was $1.94 billion, a 15.4% margin, against $0.81 billion and 8.1% a year earlier and 11.3% in the June quarter. The company credits higher price, volume and productivity.

• Power. Orders $21.8 billion, revenue $6.07 billion (+25%), segment EBITDA $1.30 billion at 21.4%, up from 13.4%. In gas, the company signed 11 GW of new contracts (8 GW of slot reservations and 3 GW of orders), converted 13 GW of reservations into orders and shipped 5 GW. Equipment backlog rose from 53 to 64 GW and reservations fell from 63 to 58 GW. More of the book is now firm orders than reservations. Power backlog reached $130.9 billion.
• Electrification. Orders $5.37 billion, down 9% organically. Management attributed the fall to "a significant order for $1.6 billion of synchronous condensers in Saudi Arabia" a year earlier, with book-to-bill about 1.4 and North American equipment orders doubling. Revenue was $3.86 billion, up 51% reported. Of that, $945 million came from Prolec GE, the transformer business acquired in February, and organic growth was 16%. Segment EBITDA was $739 million at 19.1%; on the organic basis, without Prolec and currency, 17.9%. Equipment backlog was $42.2 billion, up 60% on the year, $5 billion of it from Prolec.
• Wind. Orders $2.14 billion, up 16% organically on onshore repowering in North America. Revenue $2.74 billion, segment EBITDA $72 million (2.6%). The full-year Wind loss guide widened to $400–500 million from about $400 million, partly because scheduled maintenance was pulled forward from 2027. Management said there is "one more year to go in the offshore execution in the North Sea."

The guide. 2026 revenue $46.5–47.5 billion (from $45.5–46.5 billion), adjusted EBITDA margin 13.5–14.5% (from 12–14%) and free cash flow $16–17 billion (from $11.5–12.5 billion). Power organic growth is now 19–21% at a 19–21% margin. Electrification revenue is $14.5–15.0 billion, including about $3.1 billion from Prolec, at 18–20%, "trending toward the higher end". The revenue and margin raises are modest; the free-cash-flow raise is not. Nine months of free cash flow already total $15.3 billion. The guide therefore implies $0.7B–$1.7B for the fourth quarter, against $5.4 billion in the third, and management guided only to "positive free cash flow". No 2027 figure was given.

The 10-Q's cash-flow statement answers the question the guide raises. In the nine months, cash from operations was $16.7 billion, and the increase in contract liabilities alone contributed $19.5 billion. The 10-Q attributes it to "down payments on orders and slot reservation agreements at Power, and down payments at Electrification". Everything else in free cash flow came to −$4.2B: operating earnings, less $2.7 billion of cash taxes, $1.4 billion for pension and retiree plans net (including $0.9 billion of voluntary payments), a $2.5 billion net working-capital build outside deposits (receivables $1.7 billion, inventory $2.1 billion and contract assets $0.3 billion, less $1.5 billion of added payables), and $1.4 billion of capital spending. On the call the chief financial officer described the third quarter the same way: "Working capital was a $5.3 billion cash benefit, driven primarily by down payments on increased orders and slot reservations at Power." The third quarter's $5.4 billion of free cash flow included $5.8 billion of deposits.

What the business earns on its own. Deposits are not earnings. When the turbine ships, the revenue is booked, but the cash was collected years before, so the delivery year carries the cost with no matching inflow. For 2026 we estimate free cash flow from earnings at about $3.4B, on a normalized basis: adjusted EBITDA at the guide midpoint ($6.58B), less a 23% cash tax on operating profit and about $2.0 billion of capital spending, before pension payments. The nine months ran below that pace: free cash flow before any working-capital change was −$1.7B, after $2.7 billion of cash taxes and $1.4 billion for pension and retiree plans. The rest of the $16.5 billion midpoint, about $13.1B, is a residual: deposits and other working capital, net of taxes and pension payments above the normalized level. On the guided figure the enterprise value is 13.7 times free cash flow; on earnings free cash flow it is 67 times.

Deposits held by the company stood at $45.7 billion on 30 September, 104% of trailing revenue. At the end of 2024, before the gas order cycle, the ratio was 50%; at the end of 2025, 68%. At the 2024 ratio the balance would be about $23.5B lower. That is more than the company's $14.0B of net cash: in effect, the net cash is customers' money, held against work still owed.

How the model treats it. Deposits follow orders, not revenue, so the model ties them to the gas slot book. Above their pre-cycle level of half of revenue, they rise when far-dated slots are sold and fall as the book ships at about 24 GW a year (the mechanics are in §8). In the base case 2027 signings run at the third- and fourth-quarter pace, which the chief executive said 2027's quarters were "more likely to feel like": about 48 GW on our arithmetic. The book peaks near 156 GW, and deposits peak with it in 2027, then run down while revenue still grows, because the book ships faster than it refills. The bull signs at that pace for a second year; a third would carry the book past the bull's own stage, which ends in 2032. The bear signs no further far-dated slots. The company's 30 GW target for 2030 is left out: faster shipments would run the book down sooner, and would also lift revenue, which the growth paths set separately. Counted this way, deposits still add value, the time value of cash collected early: +$50.65 a share at their step in the attribution table, about $49 in the final model.

The strongest reading of the company's position is that deposits this size are a new norm set by scarcity, not a cycle; the chief executive said there is "no change in our cash terms" required to secure equipment. That reading is ours, not the company's words. No case carries a higher norm, because the only pre-cycle observation is 2024's, and the link between deposits and the book is fitted to two year-ends, not tested. Held at their 2027 share of revenue for good, deposits would give $704, still well below the price. The end of 2028 tests it: deposits more than $5 billion above what the slot book implies would move the model toward that case; more than $5 billion below would move the base case to the bear's book ($528).

How long the slots run. Gas equipment under contract was 62 GW at 30 September 2025, 83 GW at the end of 2025 and 122 GW at 30 September 2026. The company now ships at what the chief executive called "this 5 gigawatt quarterly run rate", about 20 GW a year. At that rate 122 GW is about six years of output on our arithmetic, consistent with slots sold for 2032 delivery. Prices are still rising: full-year 2026 gas orders are expected at about 50% more per kilowatt than 2025's. The one soft reading is the fourth quarter. Management expects gas orders "to have a price per kW in 4Q that is lower than 3Q", on mix (fewer aeroderivatives, more units for 50-hertz grids). It guided about 13 GW of signings in the quarter, close to the third quarter's 11 GW. About 20% of gas customers are data centers.

The 2028 framework is stale, and is being replaced. In December 2025 the company framed 2028 as at least $52 billion of revenue at a 20% adjusted EBITDA margin, about $10.4 billion of EBITDA. July's EV/EBITDA value was anchored there. 2026 is now guided at about $47 billion, so $52 billion would need only about 5% a year of growth after a year of 23%. The framework was not restated, reaffirmed or updated on the call or in the slides. The chief executive said the company will "create a 2030 financial guide" at a Capital Markets Day, to be dated on the January call. Until then the house uses each case's own 2028 path: about $11.4B of EBITDA in the base case on $60.0B of revenue (19.0% margin), $12.6B in the bull and $8.7B in the bear. That re-anchoring raises the EV/EBITDA value; its weight is a separate question (§4).

Trench Group. GE Vernova agreed on 25 October to buy Trench, a maker of high-voltage bushings, instrument transformers and coils, from Triton Partners for about €8.0 billion ($9.0 billion at $1.12). Trench expects $1.9 billion of revenue in 2028 at about a 36% adjusted EBITDA margin, with free cash flow of $0.2 / $0.3 / $0.4 billion in 2026–28. The price is 13 times Trench's 2028 EBITDA before synergies, funded "using an equal mix of debt and cash on hand", about $4.5 billion of each, with close expected in late 2027. The model takes the $9.0 billion out of cash in 2027 (discounted with the year's other flows) and adds Trench's own cash flows from 2028. At our cost of equity the deal costs $5.30 a share, close to value-neutral (−$0.55 at its step in the attribution table, before the method changes). Each $1 billion paid above or below what Trench is worth moves fair value by about $3.53 a share. On the base path, net cash at the end of 2027 is about $19.7B after Trench, before any further buybacks.

The buyback. The company bought $1.6 billion of stock in the third quarter and $1.3 billion more from 1 to 23 October. Since December 2024 it has bought 15.6 million shares at an average of about $635. The board doubled the authorization to $20 billion, leaving $10.1 billion. Repurchases are a transfer between shareholders who sell and those who stay. At our fair value, each $1 billion bought at $895.08 moves about $0.40B of value to the sellers. Spending the remaining authorization at that price would lower fair value per remaining share by about $16. The model counts the shares and cash as they stood after the October purchases and assumes no further buybacks.

How long each part grows faster than normal. Each stage lasts as long as its driver, and every length is argued from something dated in the filings or on the call.

Why these lengths. Above-average growth rarely lasts. Across US companies from 1951 to 1998, sales growth stayed above the median five years running for 6.3% of firms and ten years running for 0.9%, against 3.1% and 0.1% if growth were random (Chan, Karceski and Lakonishok, 2003). Turbine cycles have followed the same pattern: after the US gas-turbine order boom of 1999–2001, GE forecast in July 2002 that US orders and shipments would fall by 80% or more over two years. Hence a straight-line fade to terminal by 2035 wherever the sold-out stage ends. The bull case's sixth year rests on slots already sold for 2032, and on 2027–28 signings that would carry the book to about 180 GW (tripwire 7). A longer stage is the single input most in the price's favor, and it is priced below. The usual counter-argument is that today's buyers pay deposits years ahead, where a large share of the 1999–2001 boom was ordered by merchant (independent) power developers; management did not address the comparison on the call. We cannot test it inside the model; it is what the bull case's sixth year and Exhibit 4 price.

The services annuity. Gas and steam services growing from about $12 billion in 2025 to at least $22 billion by 2035 is about 6% a year (the company's December 2025 framing, which an analyst cited on the call and the chief executive called likely "conservative"). The base path grows the whole company at least that fast through 2032 and at 4% from 2035, so the company's services framing is consistent with the base path at the company level; the model does not split services from equipment, and the chief executive's "conservative" is credited only through the terminal-margin sensitivity. What the base case does not credit is services growing faster than the economy after 2035; that would sit in the terminal value, and a terminal margin two points higher is priced below.

Weighted 25 / 45 / 30, fair value is $534.79. Terminal value is 67% of the base case (65% to 83% across the three). That is why the terminal assumptions are priced below. The bear's 83% reflects a business whose deposits are returned and whose growth stops; its explicit years carry little cash.

What the price requires. We solve the base-case model for the $895.08 close, one input at a time. Held at its 2031 rate (7% growth, a 21% margin) the sold-out stage would have to last until about 2048, not 2031. Alternatively, growth through the stage would have to run 1.8 times the base case's, or the cost of equity would have to be 7.6% (from 9.60%), or terminal growth 6.6%, above any long-run GDP projection. At a cost of equity of 8.6% the price needs 1.5 times base stage-1 growth; at 10.6%, 2.1 times. The bull case, which carries the sold-out stage one year further and pays 23% margins, is worth $764.82; the price is 17% above it.

Why the number moved from $672. Each row of the table below re-runs one engine with one more factor changed, in order, so the rows add to the move; July's configuration reproduces its published $672 first ($672.28 unrounded). The quarter raised the value. Re-basing to the 30 September balance sheet and the raised guide, rebuilding the cash flows from drivers with a deposit line and re-anchoring 2028 EBITDA take July's model to $779.70. Two rows correct errors in the July model: it added the 30 June cash, which already held the first half's $9.9 billion of free cash flow, and also counted the full-year 2026 guide; and it discounted calendar 2026 as if received in July 2027. Seven rows change how the model is calculated, taking it to $660.82 on July's weights; the last row is the weighting judgment above.

Return on capital. Return on invested capital, gross of cash, was 17.8% over the twelve months to September against a 9.5% cost of capital, up from 8.1% a year earlier: strength, in the house's mechanical label (above the hurdle and widening). The business first cleared its cost of capital in the twelve months to June 2026. The move is operating income, $2.8bn over the trailing year against $1.4bn, on pricing in gas and grid equipment. It is earned on a small capital base: equity plus debt averaged $12.5bn over the year, less than the $13.9bn of goodwill and intangibles on the balance sheet at 30 September, because customer deposits fund the working capital. Deducting cash would leave invested capital near zero, so the return is shown gross of cash.

Sensitivities (weighted fair value). Cost of equity one point lower $648, one point higher $456; at the peer-median estimate (10.4%) $470. Terminal growth half a point lower $502, higher $573. Terminal margins two points higher $574, lower $492. Base case on the bull's margin path $561, on the bear's $467. Base deposits on the bull's slot book $542, on the bear's $528; deposits held at their 2027 share of revenue $704; no deposit line at all $486. Without Trench $540. The EV/EBITDA value at 25% $598, at July's 50% $661, and at 50% on Trench's 13 times $490. No single input in its plausible range reaches the price. Apart from the held-ratio deposit case, which is a structural alternative rather than a range ($704; the price is 27% above it), the nearest is a cost of equity one point lower, which leaves it 38% above.

The July note armed four tripwires. None fired. Two are dated 2028 and can only be read as waypoints.

Each tripwire carries its basis, its tolerance and what the model does with the answer, so the next update applies rules written now. The fourth-quarter release is expected in January 2027; the 2030 guide comes at a Capital Markets Day to be dated then.

1. Fourth-quarter deposits. Basis: the FY2026 change in contract liabilities less the nine-month $19.5 billion, and fourth-quarter free cash flow against the $0.7B–$1.7B the guide implies. A deposit inflow above $3 billion in the quarter moves the base case to the bull's slot book, a second year of signings at the 2027 pace ($542). A fall of more than $2 billion, deposits earned out faster than they are replaced, moves it to the bear's, no further far-dated signings ($528). In between, the path holds.
2. Gigawatts under contract at year-end. Basis: the fourth-quarter gas backlog plus slot reservations, against at least 130 GW guided. At or above 135 GW the base case holds the sold-out stage one more year ($547). Below 125 GW, which would mean new reservations no longer replace conversions, it moves to the bear's slot book ($528).
3. The 2030 guide (Capital Markets Day). Basis: the guided 2030 revenue and adjusted EBITDA margin, which management says will be a floor, against the base case's 2030 revenue of $71B and 21.0% margin. A margin floor at or above 22% moves the base margin path halfway to the bull's ($548). A revenue floor above the bull's $79B moves base growth halfway to the bull's ($556). A margin floor at or below 20% moves the base margin halfway to the bear's ($501). A revenue floor below $63B, as far under the base case as the bull's is over it, moves base growth halfway to the bear's ($476). No numbers by the end of 2027: the paths hold and the note says so.
4. Electrification orders and data centers. Basis: Electrification organic orders and the segment data-center orders the company gives on the call. Management expects the fourth quarter "to be our largest quarter orders in the year", including the $1.5 billion Barmer HVDC order. Fourth-quarter segment orders below the third quarter's $5.4 billion, or data-center orders below $1.0 billion in two consecutive quarters, cut base growth by one point in 2028–30 ($528) and carry the July data-center trigger forward as a reading. Orders above $7.5 billion, or data-center orders above $2 billion in two consecutive quarters, raise it by one point ($542).
5. Gas pricing. Basis: management's price per kilowatt on new gas orders against the prior year. A full-year 2027 statement that new orders priced below 2026's, with the slot book still above 120 GW, moves the base margin path halfway to the bear's ($501), because that path rests on the backlog repricing upward as it converts. Pricing level with 2026's, about 50% above 2025's, holds the path; pricing above 2026's moves it halfway to the bull's ($548).
6. Trench at close. Basis: the closing 8-K (expected late 2027) and the first full year's segment disclosure. A price above $9.5 billion, or Trench 2028 EBITDA below $0.6 billion, is recorded as overpayment at about $3.53 a share per $1 billion. A close that slips past mid-2028 is recorded and the cash outflow moved.
7. The stage the EV/EBITDA multiple assumes (restores its weight). Basis: gas equipment under contract against each case's slot book, and the 2030 guide. The base case's book is about 133 GW at the end of 2028 (slots into about 2034); the bull's is about 180 GW (into about 2036). A book at or above the bull's at the end of 2028, or 2030 floors that put the base case on the bull's path (revenue above $79B with a margin of at least 22%), moves the evidence toward the sold-out stage the multiple assumes; the EV/EBITDA value returns at July's 50% weight ($661 at today's inputs). A book about halfway between, 156 GW or more, restores half the weight, 25% ($598). Neither by the end of 2028: the cross-check stays unweighted and the next update says so.
8. The price. Fair value $534.79. A close at or below it resolves the gap. At $401, a 25% margin of safety, the house would look for an entry if the operating tests above are intact.

Fair value $534.79 a share (bear $192.72 / base $571.48 / bull $764.82 at 25 / 45 / 30), from $672 on 22 July. At the $895.08 close of 27 October the expected return to fair value is −40%; at the print-day close of $878.03, −39%. The quarter was better than July assumed: the order book, the slot calendar and the margins all moved up, and before any change of method the rebuilt model gives $779.70. What the quarter did not do is earn the free cash flow the guide headlines. Most of it is customers' money, collected because the equipment is scarce and owed back in turbines. Valued on the cash the business produces, with each stage of growth argued, the price sits above even our bull case on cash flows ($764.82; $926 if its slot book arrives and the multiple's weight returns), which already assumes scarcity into the next decade at margins the company has not yet guided. The January call puts the first number on 2027; the Capital Markets Day puts the first number on 2030.