Most CDR efficiency claims are framed as a ratio. They should be framed as a balance. A ratio of 1.1 — 100 tonnes emitted across the lifecycle to capture 110 tonnes — is a vanity metric. The number that matters is the net balance: gross capture minus embodied plus operational emissions.
That is what we mean when we say a pathway is “85% net.” For every 100 tonnes of CO2 pulled out of the air or fixed in stable form, 85 tonnes survive after you subtract everything the project emitted to make those 100 tonnes happen. The 15 tonnes you are losing went into making steel, running compressors, trucking rock dust, or heating a sorbent. That is what credible buyers should pay for: the 85, not the 100. The credible registries already refuse to issue credits for anything else.
What net removal efficiency actually measures
A proper lifecycle assessment counts every carbon-relevant flow across the project’s life:
- Embodied emissions in the equipment: steel and concrete in a direct air capture (DAC) plant, the grinder fleet for enhanced rock weathering, the pyrolysis kiln for biochar, the ship for ocean alkalinity delivery.
- Operational energy: electricity for fans, pumps, and compressors; heat for the sorbent regeneration step in DAC; diesel for crushing and trucking rock.
- Feedstock supply chain: where the biomass came from for biochar and bioenergy with carbon capture and storage (BECCS), including any indirect land-use change.
- Storage and transport: pipelines, injection wells, the long trip from a DAC outlet to a permanent geological reservoir.
- End-of-life: decommissioning, sorbent disposal.
Subtract those tonnes from the gross capture and divide back into gross capture. That is your net efficiency. Anything that ignores parts of the chain is closer to a marketing brochure than an LCA. The single biggest variable across pathways is the electricity mix the project actually runs on, not the hardware.
What the literature shows, pathway by pathway
The numbers below are typical ranges across peer-reviewed lifecycle assessments and the methodologies that operating registries use today. Real projects move inside the range based mostly on their electricity mix and feedstock provenance.
Direct air capture. On dedicated low-carbon power with nearby storage — the canonical example is Climeworks plus CarbFix in Iceland, on geothermal — the literature converges on roughly 85-95% net efficiency: of every 100 tonnes captured, 85-95 stay banked once the LCA is done. On hub-tier US grids the same hardware drops to about 70-85%. On coal-heavy grids the same machine can be net positive: more carbon emitted to drive the fans and the heat than it pulls from the air. Madhu and colleagues’ 2021 LCA in Nature Energy made this case crisply, and Terlouw and colleagues’ 2021 critical review in Energy & Environmental Science catalogued the same pattern across European grid scenarios. The point is not that DAC is bad; it is that you cannot evaluate DAC without specifying the energy source.
Enhanced rock weathering (ERW). Basalt spread on cropland — the dominant deployment shape today — typically lands between 60% and 85% net. The variance comes mostly from grinding energy and trucking radius: a project quarrying 50 km from the field is in a different regime than one trucking from 500 km. Beerling and colleagues’ 2020 Nature paper laid out the economics under different transport assumptions; UNDO and Lithos publish per-deployment LCAs that sit inside this band, with Lithos’ methodology document on Frontier’s Stripe page showing the boundary choices explicitly.
Biochar. Pyrolysis of biomass to fixed carbon typically recovers 30-40% of the input carbon as durable char, with operational emissions adding around 5-10%. Net efficiency lands at roughly 60-85% when the feedstock is a waste residue that would otherwise have decomposed. The Puro.earth biochar methodology requires the LCA to start from the residue’s counterfactual, which is what keeps biochar honest versus pulling forest material on purpose.
BECCS. The widest spread in the field. With genuine waste-stream feedstock (sawmill residues, agricultural residues that would otherwise rot or burn) net efficiency can reach 80-85%. With dedicated energy crops on land that was previously forest or peatland, indirect land-use change can push the net negative — some lifecycle scenarios in the IPCC AR6 WG3 chapter 12 on CDR put it at -20%, meaning the project emits more than it captures. BECCS deserves project-by-project scrutiny more than any other pathway.
Mineralization. CarbFix-style basalt injection delivers 80-95% net, with most of the operational footprint already on the gas-supply side. Gunnarsson and colleagues’ 2018 paper on CarbFix2 showed mineralization completing within two years underground — the durability boundary is short and verifiable. Industrial mineralization on slag tailings (Arca, Heirloom, Neustark, CarbonCure) sits in a similar 70-90% band depending on whether the tailings are produced explicitly for sequestration or already exist as a waste stream.
Ocean alkalinity enhancement (OAE). Still maturing. Land-side preparation (olivine grinding) plus ship distribution adds roughly 10-30% to lifecycle emissions; net efficiency in the published modelling sits at 50-80%. The MRV layer here is not yet defensible enough to price confidently — that was the substance of Hoffmann and colleagues’ marine CDR verification paper from earlier this month.
Permanent biomass burial. Charm Industrial, Vaulted Deep, and Graphyte report supplier-side numbers that cluster around 70-85% net, with the durability boundary set at the methanogenesis risk: anaerobic shallow burial of cellulose can degrade to methane, which unwinds part of the climate benefit. Bumping the durability claim up the geological ladder — deep injection of bio-oil instead of shallow wrap-and-bury — protects the number.
The MRV standards do this work already
Every credible registry — Puro.earth, Isometric, the IC-VCM Core Carbon Principles, the European CRCF when it lands — requires full upstream and operational lifecycle accounting before issuing credits. A “ton” sold on these standards is the net ton, with the boundaries disclosed. A 100-tonne-in, 110-tonne-out project would not clear their thresholds, and the credit would not exist.
Where the system still leaks: not every supplier is on a credible standard yet, and an LCA is only as honest as its boundary choices. Variances between registries on what counts as “embodied” or “indirect land-use change” mean two suppliers can publish very different net efficiencies for the same hardware on the same grid.
What buyers should ask for
Three questions cut through almost all the noise:
- What is the lifecycle electricity mix assumed in your LCA, and does it match the grid you actually run on?
- What is the boundary on embodied emissions — does it include manufacturing, transport, and end-of-life?
- Which standard issues the credit, and have you published the methodology document?
If a supplier cannot answer those three quickly, the project is operating outside the credible MRV layer and the net number on the brochure is whatever marketing chose. If they can, the answer is specific and falsifiable, and the conversation moves from “what is the ratio” to “is the assumed grid mix realistic over the contract period.” That second conversation is where the actual carbon balance lives.