Cost Benchmarking

Benchmarking FPSO Costs: Why $/Ton Comparisons Break Without Matching Contract Models

Updated September 2026·14 min read·Public data only

Every FPSO cost conversation eventually lands on the same shorthand: dollars per ton. It's an easy number to reach for, and a reasonable one when you're comparing two ships built to the same contract, by the same yard, in the same year. The trouble starts when people use it to compare FPSOs in general — because most of what gets called a "cost" for an FPSO is not the same kind of number at all.

I built a dataset of fourteen Brazilian FPSOs from public sources — trade press, the winning contractors' own announcements, Petrobras communications — to see how far a $/ton comparison could honestly go. This is an update to an earlier, smaller version of this piece: five more Búzios-field platforms are now in the dataset, and for the first time there's enough clean data to show a real spread instead of a single number. Here's what the full picture looks like, and why it still doesn't support the kind of one-line comparison most benchmark reports imply.

The dataset

† Total value over the full firm charter term — see below before comparing this to anything else in the table.
‡ Only a combined figure exists (110,000t for P-80 and P-82 together) — not split by unit in any public source, so left blank rather than inferred.
§ P-84 and P-85 were awarded in a single joint contract (~US$8.15bn combined, per one source) with no public per-unit breakdown.

Three points survive the checklist — and they still don't agree

A defensible $/ton figure needs four things to line up at once: a real (not estimated) headline value, a weight figure, a matching contract model, and — for charters — a contract normalized to a comparable time window. Of fourteen platforms, exactly three clear all four bars: P-78, P-80, and P-83. All three are EPC contracts, all three were built by the same contractor family (Keppel, now Seatrium, post-2023 merger), all three sit in the same field and basin.

That ~15% spread, inside the tightest possible comparison this dataset allows, is itself the finding worth sitting with. If three near-identical contracts from the same yard can differ by that much, a $/ton figure pulled from a platform with a different contractor, a different contract model, or an unconfirmed weight should be treated as a rough order of magnitude at best — not a number to build a bid position on.

Why the other eleven don't qualify

Weight data is the most common gap. Only four of the fourteen platforms report any weight figure, and they aren't all the same kind of weight. P-78, P-80, and P-83 report topside weight — the modules sitting on the hull. Almirante Tamandaré's 460,000 tonnes is displacement — the full loaded weight of the entire structure, hull included, ballast and cargo included. Dividing a contract value by displacement and calling it the same metric as dividing by topside weight isn't a rounding error — it's comparing a module to the ship it sits on.

Combined figures masquerade as per-unit figures. P-82's topside weight only exists in the public record as part of a joint 110,000-tonne figure shared with P-80. P-84 and P-85 were awarded — and are reported on — as a single joint contract, with no public split between the two units. In both cases, it would be easy to divide a shared number by two and present it as a clean per-unit data point. That number would be an estimate wearing the clothes of a fact, and it's exactly the kind of thing that survives into a slide deck unlabeled.

Charter values are a different kind of number entirely. Anna Nery's $5.5bn and Maria Quitéria's $5.2bn look, at a glance, like they belong in the same column as P-78's $2.3bn. They don't. Those figures are the value of a 25-year and 22.5-year firm charter — decades of day-rate revenue for operating the unit, bundled into one headline number. P-78's $2.3bn is a construction price. Comparing them directly is comparing the sale price of a house to the total rent someone will pay over a 25-year lease.

Even sister hulls don't move together. SEAP-I and SEAP-II are as close to identical twins as this dataset gets outside the Búzios EPC cluster — the same SBM Offshore Fast4Ward hull design, signed the same day, for neighboring fields in the same project. SEAP-II is publicly reported with higher gas-handling capacity than SEAP-I, yet its headline value is roughly $200 million lower ($4.1bn versus $4.3bn). The Búzios EPC cluster shows the same pattern in miniature: real variation persists even under the tightest possible matching.

What's next: the pipeline behind these numbers

The fourteen platforms above are the built and contracted record. What comes after them matters just as much for anyone reading this as a forward-looking market, not just a historical one.

Petrobras' 2026–2030 Business Plan puts total exploration-and-production capex at roughly US$109 billion over the period, with about US$91 billion already tied to projects in active implementation. Within that, the plan calls for 8 new production systems between 2026 and 2030 — 7 of which were already under contract at the time the plan was published — plus a further 10 projects targeted for 2030 and beyond. That's a multi-year pipeline still being built out, not a market that's finished contracting.

Two structural shifts are worth watching alongside the raw count. First, the Mero field cluster is now complete — its four FPSOs (Guanabara, Sepetiba, Marechal Duque de Caxias, and Alexandre de Gusmão, the last installed in 2025) are all in operation, closing out what was one of the largest concentrated FPSO build programs of the past decade. Búzios, by contrast, is still active: P-83 through P-85 are under construction as of this writing, on top of the nine units already producing there.

Second, and more relevant to everything above: Petrobras has been publicly signaling a shift toward the BOT model for future proprietary units, explicitly aimed at controlling costs after a run of expensive long-term charters. If that shift holds, the contract-model mismatch this piece keeps returning to — EPC construction price, BOT global price, and charter total value all describing fundamentally different economics — is going to keep getting more relevant, not less, as the mix of contract types in the Brazilian FPSO market changes under it.

What this is actually useful for

None of this means public benchmark data is worthless — it's the opposite of worthless, it's just narrower than the headline numbers suggest, and it comes with a margin of error that most reports never state. It's good for sanity-checking order of magnitude, for understanding who builds what kind of unit, and for spotting when a number being quoted to you doesn't match its stated scope. The ~15% spread among the three cleanest points in this dataset is a reasonable working margin of error to assume for any less clean comparison — which is most of them.

What public data can't do is substitute for building a comparison field by field, with each contract's value tied explicitly to its own scope, duration, and weight basis — which is the same discipline that has to hold inside an actual competitive bid, not just in a public write-up like this one. Out of fourteen well-documented Brazilian FPSOs, exactly three data points survive that discipline, and even they disagree by 15%. That ratio, and that margin, are the actual findings here — not any single dollar-per-tonne number.