Published 9 October 2026 · ECO EFX Solutions GmbH. Analysis as of 9 October 2026. Market prices are those of 8–9 October 2026 and are not updated afterwards.

Both companies announced the agreement on 7 October 2026 (CMA CGM — 7 October 2026, DSV — 7 October 2026). It extends a partnership the two already had.

The target. DSV will use ACT+ to support a targeted reduction of 12,000 tonnes of CO₂ over two years. The reductions are to come from second-generation biofuel based on used cooking oil methyl ester, known as UCOME. Like any fuel-switch figure, the reduction will be calculated from the fuel — from the quantity of biofuel used and its emissions over its life cycle — rather than measured at the funnel. That is how every fuel-based CO₂ figure works, including ours.

How ACT+ works. According to CMA CGM’s product page, customers choose a reduction level of 10%, 25%, 50% or 83%, counted well-to-wake — over the fuel’s production, transport and combustion — and can offset the remainder through a portfolio of projects. The quantities of low-carbon fuel are tracked and allocated to customers’ shipments under a mass-balance approach, so the benefit is not necessarily tied to the particular ship carrying a particular container. Each purchase comes with a declaration verified by an independent third party; the full methodology is available on request (cma-cgm.com — ACT+, accessed 9 October 2026). Prices are not published.

A precedent with public numbers. In December 2025, DHL Global Forwarding and CMA CGM agreed to use 8,990 tonnes of UCOME under the same offer, for an estimated reduction of 25,000 tonnes of CO₂e well-to-wake (DHL — 22 December 2025). It is the only public example we found that links a quantity of fuel to a reduction under ACT+, and like the DSV target it is calculated from the fuel.

The fleet behind it. CMA CGM’s announcement also sets out its fleet programme: the 23,000-TEU CMA CGM Jacques Saadé, delivered in 2020 as the first LNG-powered ship of that size; CMA CGM NOTRE DAME, delivered in 2026 as the first of ten 24,212-TEU LNG-powered ships; and around 200 dual-fuel LNG- and methanol-capable vessels expected by 2031. Those ships are not part of the DSV agreement, and we do not connect the two. They are context: this is what the long-term route looks like when a carrier commits to it.

Low-carbon fuel and new engines are the right long-term answer. They are also expensive, and the cost falls in three places: the fuel, the conversion of existing ships, and new ships.

The fuel

On 8 October 2026, VLSFO cost USD 728 per tonne in Rotterdam and USD 908 in Singapore; marine gasoil cost USD 1,416.50 and USD 1,291.50. A day later, ENGINE priced pure biodiesel at USD 1,185 per tonne in Rotterdam and USD 1,445 in Singapore (shipandbunker.com — Rotterdam, accessed 9 October 2026, shipandbunker.com — Singapore, accessed 9 October 2026, ENGINE — 9 October 2026).

Converting an existing ship

DNV puts the cost of converting a ship to dual-fuel operation at USD 5–15 million, depending on the fuel and including tanks and fuel systems (DNV — 16 May 2023). Its rule of thumb is that a conversion should cost no more than about a quarter of the ship’s newbuild value, so it typically sees ships worth around USD 50 million or more new as suitable — though some conversions, such as to methanol, can cost less. For methanol, the Mærsk Mc-Kinney Møller Center for Zero Carbon Shipping puts a conversion at 10–16% of a standard newbuild’s cost (zerocarbonshipping.com — 28 September 2022). Add the time the ship spends in the yard instead of at sea.

Building new

Hapag-Lloyd ordered eight 4,500-TEU container ships with dual-fuel methanol engines in December 2025, with an investment volume of more than USD 500 million and deliveries expected in 2028 and 2029 (Hapag-Lloyd — 12 December 2025). By simple division, that is more than USD 60 million a ship, for vessels that arrive two to four years after the contract was signed.

Fossil fuel is expensive because of the conflict around the Strait of Hormuz. The US Energy Information Administration expects Brent crude to average around USD 105 per barrel in the fourth quarter of 2026 and USD 96 for the year as a whole, then around USD 84 in 2027 as Middle East exports recover through convoys and bypass routes (eia.gov — STEO, October 2026). The 2025 average was USD 69. That is the EIA’s forecast, not ours, and it has moved this year: in August, the same outlook expected around USD 87 for 2026.

Biofuel prices move less with the conflict — less, not independently. After the conflict broke out, the premium of Rotterdam’s B30 blend over pure VLSFO contracted from USD 250 to as low as USD 128 per tonne in mid-March, according to ENGINE, before recovering to USD 269 by 10 July (ENGINE — 10 July 2026). In the week to 9 October, conventional VLSFO rose by USD 66 per tonne in Rotterdam and the blend by USD 2, and the premium fell to USD 191.

The same logic cuts both ways, and we apply it to ourselves. On this year’s pattern, the biofuel premium would widen again as fossil prices ease. But every tonne of fuel a ship does not burn would also be worth less, which lengthens the payback of any efficiency measure, ours included. That is why the scenario below is calculated at two prices: today’s, and the level before the conflict.

New ships and conversions change a fleet one ship at a time, and most of today’s fleet will not take part. In 2023, DNV regarded fewer than 10% of the existing global merchant fleet as even theoretical candidates for a dual-fuel conversion (DNV — 16 May 2023). Many existing ships can burn biofuel blends such as B30 without conversion, subject to the fuel specification, the engine maker’s guidance and compatibility checks, and at the premium shown above. A B30 blend is still nominally 70% conventional fuel by volume.

Those ships burn fuel every day, at today’s prices, and on voyages to and from EU ports the larger ones pay for their emissions under the EU ETS. For them, the question is not which fuel to switch to in 2030. It is what can be done with the engines they already have, this year, without a yard stay and without a capital budget that is not there.

Fuel replaced and fuel not burnt add up; they do not compete. A ship that needs less fuel needs less of whichever fuel it burns, so lower consumption makes every tonne of a costly low-carbon fuel go further. That is why we see efficiency on the existing fleet as a complement to fuel switching and new engines, not as an alternative to them.

What it is. NanoEFX is a water-based treatment applied to the existing air filter or a suitable intake filter medium. It is not a fuel additive, not an oil additive, not an ECU remap and not a retrofit — nothing is introduced into the fuel or oil systems, and no hardware is changed. On a ship, the crew applies it to the intake filter medium during routine maintenance, without a yard stay. Some large two-stroke engines do not use a paper air filter; whether it applies there depends on whether another treatable intake medium is present.

What our evidence covers. Our published fuel results come from road vehicles and land-based engines, collected on our evidence hub. None of them is from a ship. Some individual trials were weak, and some were mixed or inconclusive. Results depend on engine type, duty cycle, filter condition and correct application. On board, we have one engine-inspection study: a Komatsu marine main engine opened after 1,875 operating hours, with no treatment-related damage and no corrosion — a compatibility study, not a fuel result.

The auxiliary engines. A large ship is rarely one engine. Its generators are four-stroke engines with conventional intake filtration, closer to the engines our evidence comes from. That makes them the natural place to start, not a place where we can promise a result. Of our 17 published results, one comes from generators, and it is the one to know about. The weakest is a Nigerian generator fleet, IHS Towers, running continuously at near-zero load: a small positive median of +1.7%, too small and variable to establish a clear treatment effect for that duty cycle. Auxiliary generators on board normally run at considerably higher loads than that near-idle duty cycle, so its result cannot be carried over in either direction.

Biofuel. We have no documented field validation for engines running on biofuel or LNG either. Whether a treatment on the intake side behaves the same on a ship burning a B30 blend is a question for a trial, not for us to assume.

The figures below are scenario calculations, not results. They use the same model vessel and the same cost assumption as our May analysis of fuel-cost volatility: an 8,000-TEU container ship burning 17,664 tonnes of fuel a year, and a first-year cost of treatment of approximately USD 120,000, a modelled total covering product, application and first-year logistics. Our main case assumes far less than we have seen on land: 2% less fuel, because nothing has been measured on a large marine engine. For comparison we also show 5%: less than half the median of our land-based results, and below the 5.5% we assumed for the same ship in May. Neither is an expectation for ships. Both are assumptions, there to show what the numbers do.

The saving is large because the fuel bill is large, not because the effect is. Two per cent of an annual fuel bill of USD 12.86 million is USD 257,188. That is the whole argument, and it works in both directions: if the effect on a given ship is smaller than assumed, the payback stretches; if there is none, it never comes. That is why the next step is a measurement, not a purchase.

Two different jobs. A conversion gives a ship access to a different fuel for the rest of its life. A treatment of the intake filter aims at burning less of the fuel the ship already uses, by an amount that on a ship has yet to be measured. We do not rank the two. What matters for a ship that will never be converted is simpler: the treatment needs no conversion and no yard stay, and it can be tested within a normal maintenance window.

The same scenario in tonnes

The CO₂ side, calculated from the fuel. At 3.114 tonnes of CO₂ per tonne of fuel — the EU MRV standard factor for heavy fuel oil, which we use as a model assumption for VLSFO — 2% less fuel means 353 tonnes of fuel a year not burnt on this ship and about 1,100 tonnes less CO₂; at 5%, 883 tonnes of fuel and about 2,750 tonnes of CO₂. On a large ship the lever in tonnes is large because the fuel volume is large, not because the effect is. These are tank-to-wake figures for CO₂ alone. Fuel-switch programmes usually report well-to-wake CO₂e, a different basis, so the two cannot be set side by side. On voyages under the EU ETS, every tonne not burnt also avoids the allowances for its emissions; we have left that out here.

We have deliberately not multiplied any of this by a fleet. A projection built on an unvalidated application does not become more reliable by being multiplied. To run your own vessel’s numbers, use the ROI calculator.

We are looking for a shipowner or operator willing to run the first controlled trial with us. It is the honest next step, and the one that would let us write about ships with a measured figure instead of a scenario.

This is our own assessment. Replacing fossil fuel, and building ships that can burn something else, is the long-term answer for shipping. It reaches a fleet one ship at a time, and most of today’s ships will run largely on conventional fuel for years without being converted. For those ships, a measure that needs no conversion and little capital is worth testing — if the result holds on the ship’s own load profile.

It does not replace fuel switching, new engines or the regulation that drives both. It runs alongside them, and it is available now.

For every material figure above, we state either its source or the assumptions behind it. Live price pages show the current price, not the value on the date we used, so the assessment date is given here. The figures about the agreement are the companies’ own; we report them and have not verified them.

The agreement and ACT+

Prices and forecasts

Conversion and newbuilding

How the scenario is calculated

These are scenario calculations, not field results. Every input is stated so you can substitute your own and see the answer change.