The backlog stopped growing. The margin started working.

01Two things happened in the first half, and they point in opposite directions

The operating story is excellent. First-half revenue of $5,255.8m was 10.2% ahead of last year, operating profit rose 28.6% to $835.7m, and diluted earnings per share went from $0.97 to $1.53. Adjusted EBITDA margin reached 19.9% against 18.1%, and in the second quarter alone the Subsea segment posted a 23.2% adjusted EBITDA margin. Free cash flow of $764.8m funded $724.6m of dividends and buybacks — 94.7% of it returned.

The demand story is not. Subsea inbound orders fell 17.4% to $4,410.8m. Book-to-bill was 0.94x for the half, having been 1.29x a year earlier. Subsea backlog on 30 June 2026 stood at $15,833.2m against $15,810.0m twelve months before — growth of 0.1%. Total backlog, including Surface Technologies, actually shrank.

Neither observation is a surprise on its own. Together they define the question. The company is converting a backlog built in 2023 and 2024 at margins nobody expected, while replacing that backlog at roughly the rate it consumes it. One of those is a level and the other is a rate of change, and the market appears to be extrapolating both.

02The order book has flattened and the forecast is generous about it

My model carries FY2026 Subsea inbound of $10,271m. The first half delivered $4,411m. Annualise that and the model is 16.4% above the run rate. That is deliberate — subsea awards are lumpy, the second half is seasonally stronger, and management has guided to roughly $10.0bn — but it should be stated plainly rather than buried: the single most optimistic line in this model is the order forecast, and it is already above what the company has actually booked.

Beyond FY2026 the model has book-to-bill converging on 1.0x by FY2030. That is not a bearish assumption. It says the deepwater cycle stays at today’s level of activity for another five years, which is a real and defensible view of Brazil pre-salt, Guyana and the Norwegian shelf. It is simply not a growth assumption, and the backlog line in Exhibit 3 flattens accordingly.

03The backlog number is noisier than the way it gets used

Backlog drives forecast revenue through a conversion rate: revenue in a year divided by opening backlog. The obvious way to set that rate is to average the last four years. My April model did exactly that and arrived at 65.4%.

That number does not survive contact with the roll-forward. If backlog behaved mechanically, closing backlog would equal opening backlog plus inbound orders less revenue. It does not. Reported backlog also moves for foreign exchange translation, scope changes and cancellations — movements that never pass through the revenue line. The residual was −$559m in FY2023, −$1,230m in FY2024 and +$457m in FY2025.

A conversion rate computed from a denominator that swings by more than a billion dollars for reasons unrelated to operations is not a clean operating statistic. In this version the forecast rate is reset to 60.0%, anchored to the H1-2026 run rate of roughly 59% rather than to a four-year average of a contaminated series.

04Converting the backlog faster is worth nothing

This is the result I did not expect, and it is the reason the conversion rate debate that dominates coverage of this sector is largely beside the point.

Raise the conversion rate from 60.0% to 65.4% and revenue rises in every forecast year. The implied share price does not move. It stays at $39.99 in the low case, to the cent.

The same logic constrains how much comfort the backlog can offer. It tells you the revenue is coming. It tells you almost nothing about the margin at which it arrives, and the margin is where the entire valuation gap lives.

05The forecast, and where it is deliberately too conservative

Revenue reaches $11.9bn by FY2031 and EBITDA $2.33bn. EBITDA margin steps up to 20.2% in FY2026 and then sits there. That plateau is an assumption, not a forecast: Subsea gross margin is held at 25.0% and Surface at 22.2% for six consecutive years, after four consecutive years of Subsea margin expansion.

It is the single largest source of conservatism in the model and I am flagging it rather than defending it. The second quarter of 2026 produced a 23.2% Subsea adjusted EBITDA margin. If that is the new structural level rather than a mix effect, the flat-margin assumption is wrong and the valuation below is too low. Exhibit 8 prices exactly how wrong: moving Subsea gross margin from 24.0% to 25.5% is worth $3.40 a share.

Against reported results and company guidance the rest of the model sits mid-range or below. Free cash flow is 14.1% under the first-half run rate and below the bottom of guidance. Earnings per share is 4.3% below the run rate. The tax rate is set above the 25.2% actually paid in the first half. Where the model departs from guidance it departs downward — with one exception, the order forecast discussed above.

One further point about what is actually being discounted. Named, awarded contracts cover 37% of FY2026 Subsea revenue, 13% of FY2027 and 1% of FY2028. Everything after that is the backlog conversion model. This is normal for a business with a two-to-four year cycle from award to revenue, and it is the correct way to build the forecast — but it means the visibility that supports the multiple is contractual for the current year and statistical thereafter.

06What my own judgement is worth, priced call by call

A discounted cash flow is mostly an argument about four or five numbers that cannot be observed. Rather than present a single target and a sensitivity table, this model isolates each judgement call and prices it. Start from the low case, where all four calls sit at the conservative end, and move one at a time.

The largest single lever is beta, worth $4.46 a share between 0.90 and 0.764. It is also the least observable input in the model, which is an uncomfortable but honest place to end up: the biggest swing factor in my valuation is the number I can defend least. The central case uses 0.85.

Beneath the levers, the arithmetic of the discount rate is unglamorous. A 4.3% risk-free rate, a 5.5% equity risk premium and a beta of 0.85 give a cost of equity of 8.97%. With debt at 1.5% of capital the weighted average cost of capital is 8.89%. Terminal growth of 2.25% is a nominal GDP proxy. On those inputs the present value of six years of unlevered free cash flow is $5,780m, the terminal value is $13,015m, and enterprise value is $18,796m — with 69% of it beyond FY2031.

07What you would have to believe to own it at $78.31

Run it backwards. Hold the operating forecast still and ask what discount rate and terminal growth rate reproduce $78.31. Neither lever gets there alone. Cut the discount rate all the way to 7.39% while holding terminal growth at the central 2.25%, and the answer is $59.42. Raise terminal growth to 3.75% while holding the discount rate at 8.89%, and the answer is $55.86. Both at once, at the extreme corner of the grid, produce $78.29 — two cents short of the close.

A 7.39% weighted average cost of capital on these inputs implies an equity beta of about 0.57 for a company whose revenue depends on operator sanctioning decisions. Terminal growth of roughly 3.3% is a full percentage point above the nominal GDP proxy this model uses, in perpetuity. Each is arguable in isolation. Both at once, plus margins at or above the second quarter exit rate, is the position embedded in the price. Of the forty-nine cells in Exhibit 9, none reaches today’s price.

Two cross-checks say the same thing in different units. On enterprise value the market pays 14.9x FY2026 estimated EBITDA against 8.7x on the model’s own enterprise value. And the nine-cell scenario matrix — every valuation case crossed with every operating scenario — spans $35.95 to $58.67. The market price lies outside all nine.

08Where this is most likely to be wrong

The honest reading of the gap is not that the market is irrational. It is that the market is underwriting a margin level and a cost of capital that this model declines to underwrite. Four places where that decision could prove mistaken:

• The margin plateau. Subsea gross margin is flat for six years after four years of expansion, and the second quarter printed a 23.2% segment EBITDA margin. If iEPCI has structurally changed the economics rather than flattered a favourable project mix, the forecast is too low from FY2027 onward and every downstream number moves with it.
• Beta. A net cash balance sheet, 1.8 years of revenue visibility and long-cycle contracts are a real argument for a lower equity beta than the oilfield services peer norm. The gap between 0.90 and 0.764 is already $4.46 a share; the gap to the 0.58 the market appears to use is considerably more.
• The oil deck, and the war behind it. The regional inbound model runs Brent at $87 in 2026 and $69 thereafter, following the EIA Short-Term Energy Outlook of 11 August 2026. That $69 is not a neutral baseline. The same outlook has Brent falling into the $60s specifically as the geopolitical effects of the conflict in Iran ease, with Hormuz transit normalising and Middle East production returning to near pre-conflict levels in early 2027. This model adopts that view without having argued for it. Brent traded near $97 in early September; the deferred strip — the price at which 2027 and 2028 barrels actually transact — sat closer to $77 and declining, which is the right benchmark for an out-year deck and much nearer the number used here. Regional sensitivities range from 0.05 in Brazil to 0.25 in Africa, so even a permanent $95 deck moves inbound by 2% to 9%, worth roughly $1 to $4 a share against the $2.06 the inbound calibration factor carries. It is a judgement call like the other four, and it is worth about a tenth of the gap.
• The valuation date. The model is anchored at 31 December 2025 with FY2025 net cash and share count, for internal consistency of the discounting. Rolling it to mid-2026 would use a 3.6% lower share count, which raises per-share value modestly. It was not applied.

09What would change the number

Four observable things, each with a price attached:

• Second-half book-to-bill above 1.15x, taking FY2026 Subsea inbound past $10.5bn. That would make the first-half order stall a timing artefact and validate the inbound calibration factor at 1.000, worth $2.06 a share on its own.
• Subsea EBITDA margin sustained above 23% through FY2027 without a mix explanation. That breaks the flat-margin assumption. The 24.0% to 25.5% gross margin step is worth $3.40; a structurally higher level is worth considerably more than that.
• Realised volatility supporting a beta materially below 0.80. The largest lever in the model and the one most likely to be re-rated by the market before it is re-rated by the analyst.
• Brent holding above $85 through 2027, with the out-year curve following spot rather than reverting. The deck here is $87 in 2026 and $69 thereafter, from the 11 August EIA outlook — a forecast conditioned on the conflict in Iran easing. On regional sensitivities of 0.05 to 0.25, a permanent $95 deck adds 2% to 9% to inbound, worth roughly $1 to $4 a share against the $2.06 the calibration factor carries. It moves the answer. It does not close the gap.

None of this is an argument that TechnipFMC is a poor business. It is the best-executing company in its peer group, it has the only net cash balance sheet among them, and the deepwater cycle it serves is structurally supported through the end of the decade. The question this note answers is narrower: what price the company’s own numbers support once each judgement call is made explicit and priced.

On my numbers that price is $46.22, and no combination of the four calls I am willing to make reaches $78.31. I would rather publish a valuation the model can actually produce than reverse-engineer one that agrees with the tape.