This is BMInsider PRO content
You're reading a Deep Dive — our most in-depth research. Unlock full access to all analyses, Smart Money breakdowns & more.
Unlock BMInsider PRO →One-time payment · instant access · no account needed
In the first half of 2026 Safran shipped 1,030 LEAP engines, 41 percent more than a year earlier. By its own chief financial officer’s account, it lost money on every single installed one of them. In the same six months, the operating margin of its propulsion business rose to 24.5 percent. Resolving that contradiction does more than explain an industry. It exposes the single question that decides whether the most reliable profit machine in modern manufacturing is worth what the market is currently paying for it.
The loss is the entry ticket
A commercial jet engine is not a product. It is a contractual relationship that opens with a delivery and closes twenty-five to thirty years later. In between, the engine is fully torn down, overhauled and reassembled several times. Each of those events — a shop visit, in the trade’s language — costs the operator a mid-single-digit number of millions of dollars. And in nearly every case, a large share of that money flows back to the company that built the engine in the first place.
That is why the manufacturer can afford to hand over the engine below cost. Safran has said as much repeatedly about the LEAP: the program of installed plus spare engines together has only been profitable for roughly three years, and the company still takes a loss on each installed engine — the ones delivered hanging under the wing of a factory-fresh Airbus A320neo or Boeing 737 MAX. The profit sits in the spare engines airlines buy as reserves, and above all in what comes afterwards.
Economically this is an investment in a position, not in revenue. Whoever has the engine on the wing holds, for the life of that airframe, the type certificate, the approved repair procedures and the parts catalogue. The operator cannot change engine without changing aircraft. There are very few customer relationships in industry with that kind of lock.
How much of the money is actually in the parts
GE Aerospace discloses the arithmetic cleanly enough to measure. For fiscal 2025 the company reported GAAP revenue of $45.9 billion, up 18 percent, and operating profit of $9.1 billion at a 21.4 percent margin. Free cash flow was $7.7 billion, orders rose 32 percent to $66.2 billion, and the backlog stands at roughly $190 billion.
The composition matters more than the totals. Commercial Engines & Services accounted for about 73 percent of company revenue in 2025, and within that segment services made up roughly 75 percent. Do the multiplication and well over half of every dollar the company collects comes not from selling engines but from maintaining them. The growth rates say the same thing louder: services revenue grew 26 percent in 2025, internal shop visit revenue 30 percent, and spare parts revenue more than 25 percent. In the same year CFM International — the fifty-fifty joint venture between GE Aerospace and Safran — delivered more than 1,800 LEAP engines, a record.
The margin is the punchline. On roughly $32 billion of segment revenue, Commercial Engines & Services earned $8.9 billion of operating profit, up 26 percent, a 26.6 percent segment margin. That is a software margin earned by a heavy-metal business, and it rests on a parts catalogue nobody else is fully licensed to serve.
The number the whole thesis hangs on
This is where most investors make the decisive mistake. They treat “aftermarket” as one idea. Maintenance demand is booming, so buy whatever stands near maintenance demand. The margins actually reported in 2026 show how little that category is worth.
| Business | Period | Operating margin | What is actually sold |
|---|---|---|---|
| TransDigm (group) | Q3 FY 2026 | 52.8% EBITDA margin | proprietary parts, largely sole-source |
| GE Aerospace, Commercial Engines & Services | FY 2025 | 26.6% | own parts plus own shops |
| Safran, Propulsion | H1 2026 | 24.5% | own parts, LEAP and CFM56 catalogue |
| HEICO (group) | trailing twelve months | 24.3% | PMA replacement parts, DER repairs |
| Rolls-Royce, Civil Aerospace | FY 2025 | 20.5% | availability per flying hour |
| Pratt & Whitney (RTX) | Q2 2026 | 8.3% | own parts, carrying a recall |
| MTU Aero Engines, commercial MRO | H1 2026 | 8.0% | labour hours and shop capacity |
Top row to bottom row is a factor of six and a half. Every one of these companies makes its money because aircraft must be maintained. The difference is not demand, not the cycle and not management quality. The difference is whether the company owns the part number or merely the workbench the part gets installed on.
Why the part number decides everything
MTU Aero Engines demonstrates the point better than anyone because it carries both sides in one set of accounts. In the first half of 2026 its adjusted commercial maintenance revenue rose 21 percent to €3.4 billion — 29 percent measured in dollars, stripping out the currency effect. Splendid growth. Adjusted EBIT for the same business rose only 13 percent, to €271 million, and the margin fell from 8.6 percent to 8.0 percent. Roughly 46 percent of that business was maintenance on Pratt & Whitney’s geared turbofan.
There it is: MTU overhauls, at industrial scale, an engine whose parts catalogue belongs to someone else. The shop earns on labour and on a handling markup over pass-through material. The catalogue owner earns on the part number. In that structure more revenue does not automatically mean more margin; in the first instance it means more material flowing through. Which is how one and the same boom lifts one company’s margin to 26.6 percent and pushes another’s from 8.6 down to 8.0.
The inverse holds too, and American investors have the purest examples of it. TransDigm sells mostly proprietary components for which there is frequently no second approved source, and reported a 52.8 percent EBITDA margin in the third quarter of fiscal 2026 with $870 million of free cash flow in a single quarter, against full-year revenue guidance of about $10.51 billion. HEICO attacks the same economics from the other direction: it certifies replacement parts under the FAA’s Parts Manufacturer Approval route and takes catalogue margin away from the originals one part number at a time. Third-quarter 2026 revenue rose 23 percent to $1.41 billion, with the Flight Support Group up 18 percent to $947.8 million.
The question for a portfolio is therefore never “does this company benefit from the maintenance cycle.” It is: where on the part number does it sit?
The gap in the schedule
Now the part today’s margins do not show. What GE Aerospace and Safran earn right now, they largely earn on an engine generation sold in the 1990s and 2000s: the CFM56, the most-produced commercial jet engine in history. That fleet is old enough that every unit has been through the shop several times, and those visits are expensive precisely because older engines consume more parts.
The data here is unusually well documented. CFM56 shop visits are running at 2,300 to 2,400 per year and are expected by industry analysts to hold near that level through 2027 and 2028; the peak originally pencilled in for 2025 has slipped, because the successor engine ramped more slowly than planned and airlines are flying older aircraft longer. The LEAP, by contrast, generated only around 1,000 shop visits in 2025 — though it is accelerating fast, with LEAP shop visits up more than 50 percent in the first quarter of 2026.
The reason for the gap is physics and the calendar, nothing more. A narrowbody engine typically reaches its first full overhaul in its seventh year of service; on modern widebody engines the trigger is closer to 20,000 engine flight hours. The 1,800-plus LEAP engines CFM delivered in 2025 alone therefore will not generate their first large bill until the early 2030s.
Which produces an uncomfortable possibility that almost no valuation model shows: the old fleet that pays today shrinks before the new fleet that pays tomorrow comes due. That transition gap is the central timing risk in the entire model. It has repeatedly been pushed out — every delivery delay at Airbus and Boeing extends the life of the CFM56 fleet and with it the good years. But being postponed is not the same as being cancelled.
What an overhaul really costs, and who pays for it
| Shop visit | Typical cost | Note |
|---|---|---|
| Narrowbody performance restoration | $0.5m–$1.5m | limited workscope, mature types |
| LEAP-1A performance restoration | $2.0m–$4.5m and above | new generation, costlier materials |
| LEAP high-pressure turbine durability kit | an extra $0.3m–$0.6m | often embodied in the same visit |
| Widebody engine overhaul | $3m–$5m | before life-limited parts |
Who carries that bill depends on the contract, and this is the second underappreciated fork in the road. An airline flying on time and material pays for each overhaul as it happens, and the manufacturer earns on every part fitted. An airline on a long-term service agreement pays a fixed rate per flying hour, and the manufacturer carries the risk that the repair costs more than assumed.
Both models make the manufacturer rich. But they invert the incentives. Under a parts-sales model, an engine that lasts longer is lost revenue. Under a per-flying-hour contract, the identical engine is margin gained, because the hourly rate stays and the cost falls. Anyone valuing a stock in this sector should know which incentive system they are looking at.
The contract is the asset — and the contract is an estimate
This is where it gets awkward for anyone who reads financial statements carefully. GE Aerospace discloses how these long-term service agreements are booked: terms generally run ten to twenty-five years, and revenue is recognised on a percentage-of-completion basis, using costs incurred against estimated total costs over the life of the contract.
The company names the required assumptions itself, and unusually plainly: first, how the customer will utilise the assets covered; second, the expected timing and extent of future overhaul services; third, the future cost of materials, labour and other resources; fourth, forward-looking information about market conditions.
Sit with that for a moment. A meaningful share of the profit reported today by one of the most valuable industrial companies on earth rests on a judgement about how often an airline will fly in the 2040s and what a turbine blade will cost by then. This is not accounting sleight of hand — it is the prescribed treatment for contracts of this type, and the alternative would distort the economics more, not less. But it does mean the quality of these earnings depends on assumptions no outsider can audit and which can be revised in either direction. Rolls-Royce, for its part, explicitly attributes part of its much-improved civil margin to better commercial terms and higher service-agreement margins across its in-production widebody engines.
Pratt & Whitney: when the annuity becomes a liability
The clearest proof that an installed base runs both ways has been parked outdoors since 2023, on aprons from Toulouse to Doha. Contaminated metal powder was identified in safety-critical components of Pratt & Whitney’s geared turbofan family, including high-pressure turbine discs, where it can seed cracks. The consequence was accelerated inspection across the fleet.
The numbers have become brutal. What was initially communicated as roughly a $3 billion hit is now put by industry estimates at $6 billion to $7 billion gross. At the end of 2025 some 770 GTF-powered jets sat in storage; the count grounded by the engine issue fell 15 percent in the first quarter of 2026. Across 2026 a further 600 to 700 engines remain in rework.
Here the downside of the availability contract becomes visible. A manufacturer that has promised readiness in exchange for a per-flying-hour payment still owes that readiness when the reason for its absence is its own material — and pays compensation on top to the airlines whose aircraft are stationary. The installed base, ordinarily an annuity, becomes a liability with precisely the same duration.
The operational recovery is now real. Pratt & Whitney reported second-quarter 2026 adjusted sales of $8.9 billion on 17 percent organic growth, with commercial aftermarket up 25 percent and military up 23 percent, partly offset by an 8 percent decline in commercial original equipment. MRO output rose 40 percent and aircraft-on-ground were down 25 percent year to date. Adjusted operating profit climbed 22 percent to $740 million — an 8.3 percent margin. That is roughly a third of what the structurally comparable business earns at GE Aerospace. The successor GTF Advantage has now been certified by both EASA and the FAA and is due to enter service later in 2026.
Rolls-Royce: same fleet, different contract
Rolls-Royce runs the control experiment. It sells its large engines almost exclusively on availability contracts, which gives it a direct financial interest in its engines needing the shop as rarely as possible. That is exactly what the durability programme is for: the target is now more than a doubling of time on wing across in-production Trent engines by the end of 2027, with more than half of that already delivered, and the life-extension programme for the Trent XWB-84 completing in 2026.
The margin records the result. Civil Aerospace delivered a 20.5 percent underlying operating margin in 2025 against 16.6 percent the year before, driven by stronger large-engine aftermarket performance, improved contractual terms and more profitable spare engines. Large engine flying hours grew 8 percent. Group underlying operating profit rose from £2.5 billion to £3.5 billion at a 17.3 percent margin. Guidance for 2026 is £4.0 billion to £4.2 billion with flying hours at 115 to 120 percent of 2019 levels, and mid-term targets have been raised to £4.9–5.2 billion of operating profit and £5.0–5.3 billion of free cash flow.
Note the direction of travel: a manufacturer that causes fewer shop visits earns more here. For a pure parts seller, the identical engineering improvement would be a revenue warning.
What the market is paying for it
| Company | Price | Market cap | P/E (trailing) | EV/EBITDA |
|---|---|---|---|---|
| HEICO | $325.48 | $45.5bn | 54.3 | 32.4 |
| Rolls-Royce Holdings | about £14.89 | £123.6bn | 41.2 | 22.4 |
| GE Aerospace | $337.12 | $349.8bn | 39.7 | 31.4 |
| Safran | €333.50 | €138.3bn | 35.8 | 21.1 |
| TransDigm | $1,162.05 | $64.2bn | 35.3 | 18.6 |
| MTU Aero Engines | €347.60 | €18.7bn | 20.1 | 11.3 |
Prices and multiples as of 4–6 September 2026. The table answers the obvious question by itself: no, the market does not confuse these business models. MTU trades at 11.3 times enterprise value to EBITDA, GE Aerospace at 31.4 times. The premium for owning the part number is nearly threefold — roughly the same size as the margin gap.
That shifts the actual investment question. It is no longer whether the business model is good; that is settled and priced. It is whether an EBITDA multiple north of thirty is the right price for an earnings stream whose present base is an ageing engine generation, whose future base does not start paying until the 2030s, and whose reported profit rests on twenty-five-year estimates. At TransDigm there is $30.7 billion of net debt on top — supportable at a 46.7 percent operating margin, but it is a leveraged model and will be treated as one in every rate discussion.
MTU’s position is the instructive one. Eleven times EBITDA for a company whose maintenance revenue is growing 29 percent in dollar terms is not a mispricing; it is the honest price of an eight-percent-margin business. Buying MTU buys volume. Buying GE or Safran buys a position in the catalogue. Both can work. The investor should simply know which one they are doing.
Three scenarios
| Scenario | What happens | Valuation consequence |
|---|---|---|
| Bull: seamless handover | Delivery delays at Airbus and Boeing keep the CFM56 fleet in service longer while LEAP shop visits already grow above 50 percent; the two waves overlap instead of relaying. | Today’s multiple is defensible because the earnings stream never breaks. The winners are catalogue owners, not shops. |
| Base: an air pocket in the middle | CFM56 overhauls roll off after 2027–28 as expected, and the LEAP fleet does not hit its first large wave until the early 2030s. | One to three years of slower spare parts growth. At above 30 times EV/EBITDA that is enough for a visible de-rating even with the model intact. |
| Bear: technical fault meets contract risk | A materials or durability problem in a newer type lands on availability contracts, as at Pratt & Whitney in 2023, while PMA parts and used serviceable material press on catalogue pricing. | The installed base inverts: estimated future revenue becomes estimated future cost. Six to seven billion dollars gross is the order of magnitude. |
One detail from the base case deserves attention because it shows how tight the market has become: operators needing a spare engine at short notice are currently paying premiums of 8 to 12 percent over market lease rates to be supplied within two weeks, while those paying market rates wait eight to twelve weeks. Premiums like that are the thermometer of a capacity shortage — and they disappear once shop capacity catches up.
How a US-based investor can actually own this
American investors have an unusual advantage in this sector: the two purest expressions of the thesis are domestic listings. GE Aerospace and RTX cover the engine makers; TransDigm covers the proprietary-parts model in its most concentrated form; HEICO covers the attacker taking catalogue share through PMA approvals. All four are ordinary US common stocks, eligible for retirement accounts, with no foreign withholding and no currency conversion.
The European names are reachable but less clean. Safran, Rolls-Royce and MTU all trade as unsponsored or sponsored ADRs on the US over-the-counter market, which brings three frictions worth pricing in: depositary fees deducted from dividends, wider spreads and thinner volume than the home listings, and dividends declared in euros or sterling, so the dollar amount moves with the currency. France withholds tax on Safran dividends, recoverable in part under the treaty and creditable via the foreign tax credit; the United Kingdom levies no dividend withholding on Rolls-Royce, which makes it the least friction-heavy of the three. Note also that a foreign tax credit is of no use inside an IRA or 401(k), so treaty-withheld income is simply lost there — a reason to hold the foreign names in a taxable account if you hold them at all.
On the tax treatment of the position itself: gains on shares held longer than a year are taxed at long-term capital gains rates, gains inside a year at ordinary income rates, and qualified dividends from US corporations at the preferential rate. Those thresholds argue for holding a thesis with a ten-year clock — which this one has — in a form that does not force annual turnover. For anyone preferring a fund wrapper, the listed aerospace and defense ETFs are the common route; check the holdings before buying, because most of them carry a heavy defense weighting and therefore track a completely different cycle from the one described here, driven by government budgets rather than by flight hours.
Three questions before you buy
Anyone assessing a stock in this sector will get further with three questions than with any industry outlook.
First: does the company own the part number or the workbench? The margin difference between those two positions, in the same quarters of the same boom, is a factor of three to six. That is a structural fact, not a cyclical one, and it does not change.
Second: which engine generation is paying today’s profits, and when does the next one start paying? Between CFM56 and LEAP there is an interval that current metrics do not capture. Paying more than thirty times EBITDA means implicitly buying the assumption that the interval does not exist.
Third: which contract form is the company sitting on, and who carries the repair cost? An availability contract turns durability into profit and a material defect into a catastrophe. A parts contract turns durability into lost revenue and a material defect into the customer’s problem. Both are labelled “aftermarket” in the annual report, and they are not the same business.
The answers to all three are in the filings, usually stated with surprising candour. They are simply never in the headline you read before you buy.