Every PFAS Technology Removes PFAS... But The Decision Is What You Do With The Leftovers

Most PFAS treatment evaluations I’ve seen start with the wrong question. The utility asks which technology removes PFAS best, gets three vendors to present, and picks the one with the strongest removal data. Removal is the part of this problem that’s already solved. Granular activated carbon, single-use anion exchange resin, and high-pressure membranes can all bring a contaminated source below four parts per trillion when they’re sized correctly for the specific application.

What separates them shows up later, in a twenty-year operating cost that nobody modeled and a waste stream that nobody wanted to own. I’ve watched a utility select on capital cost and unit removal efficiency, then spend the next two budget cycles discovering that its spent media had become the most difficult line item in the plant to defend publicly.

This post assumes you already know what the PFAS rule requires and where the deadlines are. Read that first if you haven’t already!

The removal question is largely settled

All three PFAS removal techniques are proven to work. The differences that matter cluster in three places.

Short-chain performance. GAC adsorbs longer-chain compounds like PFOA and PFOS well and handles short-chain compounds poorly. Short-chain PFAS break through a carbon bed far earlier than the long-chain compounds do, which means the compound mix in your source water sets your change-out frequency more than your influent concentration does. Anion exchange resin holds a broader range, including the short-chain variants, which is why utilities chasing very low targets or facing a messy compound profile tend to land there. In pilot work at a matched five-minute contact time, resin has run six to eight times the bed volumes GAC managed before breakthrough, and across the literature resin media use rates come in a factor of two to ten below carbon. Nanofiltration and reverse osmosis are indifferent to chain length; they reject by size and charge.

Footprint and hydraulics. GAC needs empty bed contact time (EBCT), and contact time needs vessels. Carbon in PFAS duty is generally designed around ten minutes or more of contact, commonly eight to thirteen; anion exchange typically runs at two to five. That difference decides retrofits. A utility working inside an existing train with limited real estate often finds the carbon option it preferred on paper doesn’t physically fit, or fits only by cutting contact time to a point where bed life collapses. Membranes bring their own demand in pressure, energy, and pretreatment.

Pretreatment sensitivity. This is where site-specific reality overrides the general comparison, and it’s covered below.

If your evaluation gets stuck arguing removal percentages between two of these, the evaluation is stuck on the wrong axis. Ask each vendor to size for your water and state a bed life; the credible ones will decline to do it without pilot data, which is itself useful information.

Where the PFAS goes after you catch it

GAC, ion exchange, and membranes share a property that rarely appears on a comparison slide. None of them destroy anything. Each one moves PFAS out of the finished water and concentrates it somewhere else, and that somewhere else becomes yours.

The endpoint options are limited, and each carries exposure the utility should price before signing.

Reactivation. Spent carbon can be shipped to a thermal reactivation facility and returned to service, at roughly $0.90 to $1.30 per pound of media in published fee ranges. Commercial reactivation capacity exists for GAC and not for other sorption media, it’s mainly practical for large-volume users, and utilities that assumed a slot would be available on their schedule have been surprised. EPA’s 2026 guidance notes that certain reactivation units may operate under conditions favorable to destroying the PFAS they drive off, with uncertainties remaining.

Regeneration. Single-use resin is the common configuration, but on-site regenerable ion exchange is a real third path, using a brine and solvent regeneration in place. Roughly nine full-scale PFAS regenerable systems are operating globally, so the operating record is thin. The trade is a much smaller solid waste stream in exchange for a concentrated liquid regenerant that still needs an endpoint, which suits a utility that would rather manage one small high-strength stream than a recurring media logistics problem.

Incineration. High-temperature incineration runs about $1.00 to $1.50 per pound of spent media, and the regulatory picture around it moved in 2026 after several years of doubt. The Department of Defense lifted its own incineration prohibition in February, allowing PFAS materials to go to facilities holding both RCRA and Clean Air Act permits, running secondary chamber temperatures above 1,100 degrees Celsius, and testing emissions to EPA’s protocol. EPA’s 2026 interim guidance reports promising results from thermal treatment under those kinds of conditions. The open question narrowed rather than closed, and it now sits mainly on lower-temperature units such as municipal waste combustors, where EPA says more information is needed on whether products of incomplete combustion form. The practical read is that permitted high-temperature destruction is a more defensible path than it looked two years ago, and that the specific facility and its operating temperature are now part of what you’re evaluating.

Landfill. Landfilling spent media is the cheapest endpoint and the one whose standing weakened most recently. EPA’s 2026 guidance states that new information shows landfilling could release more PFAS to the environment than the agency believed in 2024, and it steers higher-concentration material toward hazardous waste landfills with the strictest engineering controls. The loop behind that is familiar to anyone who has run a wastewater plant: about sixty percent of Subtitle D landfills send their leachate to a treatment plant, which discharges to surface water. Studies of that pathway generally find leachate doesn’t move PFAS levels much at the receiving plant unless it makes up a large share of flow, so the exposure here is less about mass loading than about acceptance. Wastewater utilities have been tightening leachate acceptance, adding conditions, and turning loads away, which means the cheapest endpoint is also the one whose availability is least certain over a twenty-year horizon.

Membrane concentrate. RO and nanofiltration produce a continuous concentrate stream containing everything they rejected. A coastal utility with an outfall has options. An inland utility doesn’t, and “concentrate management” on a vendor’s slide can mean an evaporation and crystallization system that costs more than the membranes. This is a common reason a membrane option that looked competitive at the feasibility stage stops being competitive at thirty percent design.

The pattern across every one of them is the same. The technology decision looks like a treatment decision and behaves like a waste management decision, and the waste side is governed by rules that are less settled than the drinking water rule that started the project.

The co-contaminants that decide your bed life

The number that drives twenty-year cost is media change-out frequency, and change-out frequency is driven less by PFAS than by everything else in the water.

Total organic carbon is the main one. TOC competes with PFAS for adsorption sites on carbon, and a source with elevated organics can cut GAC bed life sharply against a vendor model built on cleaner water. It cuts the other way too, and this surprises people: anion exchange resin is more susceptible than GAC to fouling and scaling where organic carbon, iron, and competing anions run high, so resin’s bed-volume advantage narrows in exactly the waters that need the most help. Nitrate and sulfate compete directly for exchange sites. Iron and manganese foul media and membranes both. A source with any of these may need pretreatment that wasn’t in the original capital estimate, and the pretreatment decision can flip which technology wins.

Published bed-life ranges are wide for exactly this reason. Carbon in PFAS duty commonly runs somewhere between six months and two years between change-outs and single-use resin often runs longer, but those spans are broad enough that the midpoint isn’t a planning number. The peer-reviewed work is blunt about why: media use rates are dictated by the treatment target and the influent water quality, which is another way of saying they’re dictated by your site. Only pilot or rapid small-scale column testing on your own source water produces a figure you can put in a budget, and running that testing early is the cheapest risk reduction available in the whole project.

Two practical consequences follow. Put the full water quality picture in front of every vendor before asking for a proposal, TOC and competing ions included. And treat any bed life quoted without site data as a sales estimate.

What to require before you pilot a destruction technology

Destruction technologies break the carbon-fluorine bond instead of relocating the problem: electrochemical oxidation, supercritical water oxidation, plasma, and several others. The category is real and the direction is right, because it’s the only branch that ends the residuals question.

It’s also early. Utility-scale operating history is thin, most units in the field treat concentrated side streams and few run full plant flow, and the vendor population spans a wide range of credibility. That mix is the normal condition of an emerging category and it calls for the same discipline any technology being vetted for specification deserves.

Before committing staff time to a destruction pilot, ask for four things:

• Fluoride mass balance. Destruction produces fluoride. A vendor who can account for fluorine in equals fluorine out has evidence of destruction; a vendor reporting only that PFAS concentrations dropped has evidence of removal, which is what you already had.
• Byproduct analysis. Partial breakdown can generate short-chain PFAS that the influent didn’t contain. Ask what the effluent looks like across the full compound panel, and ask whether anything appeared.
• Energy per unit treated, at your matrix. Bench results on clean spiked water don’t transfer to a real concentrate stream with competing organics.
• A referenceable installation running on something like your waste. Treating ion exchange regenerant is a different job from treating landfill leachate or RO concentrate.

The cost argument is what keeps the category interesting. Published figures put supercritical water oxidation near $0.44 per kilogram of spent media against roughly $2.00 to $3.30 for reactivation or high-temperature incineration, and a destroyed compound generates no downstream acceptance risk. Those numbers come from a small number of installations, so treat them as a direction of travel and not as a quote.

The realistic near-term role for destruction in most systems is as a back end on a concentrate or regenerant stream, sitting behind a proven removal step. Sequencing it that way also gives you a defensible answer when a board member asks why you didn’t wait for the newer technology.

How to put residuals into the procurement

The fix for most of this is procedural, and it belongs in the bid documents.

Specify performance and let the technology compete. A spec that states the finished water target, the design flow, the source water quality including co-contaminants, and the required demonstration of compliance will surface options a technology-named spec forecloses. The procurement mechanics behind that differ by utility type, which is covered in the material on how municipal and investor-owned buyers run these processes.

Then make residuals an evaluated element instead of an assumption:

• Require each bidder to name the disposal endpoint, the receiving facility, and what happens if that facility stops accepting the material.
• Ask for twenty-year lifecycle cost including media replacement, transport, and disposal at current rates, and ask for the same cost recalculated at a disposal price two or three times higher. The spread between those two numbers is the risk you’re actually buying.
• Put bed life on the vendor. Guaranteed-performance and media-as-a-service arrangements exist, and moving change-out frequency risk to the party who modeled it changes the incentive to model it accurately.
• Confirm who holds title to the spent media and for how long. A utility can find itself the generator of record for material sitting in a third party’s yard.

Here is the comparison in one place:

Technology	Removal strength	Site sensitivity	What you own afterward	Where it goes wrong
Granular activated carbon	Strong on long-chain, weak on short-chain	Needs 10+ min contact time and the vessels to hold it; high TOC cuts bed life	Spent carbon, reactivable at ~$0.90-1.30/lb	Retrofit doesn’t fit, or organics halve the modeled bed life
Anion exchange, single-use	Broad, including short-chain; good at very low targets	Runs at 2-5 min contact; fouls faster than GAC on high organics and iron	Spent resin, incineration or landfill	Higher media cost, and no thermal reactivation route
Anion exchange, regenerable	Same as single-use, with far less media consumed	Same fouling sensitivity, plus regeneration handling	A concentrated liquid regenerant	Thin operating record, roughly nine full-scale systems globally
Nanofiltration / reverse osmosis	Effective across the panel by size and charge	Fouling, energy load, and pretreatment demand	A continuous liquid concentrate stream	Inland concentrate management costs more than the membranes
Destruction technologies	Ends the residual instead of moving it	Performance varies sharply by waste matrix	Fluoride and breakdown products	Thin utility-scale record, and bench data that doesn’t transfer

Every technology in that table is a defensible choice for some utility. The selection turns on which set of problems your system is equipped to carry for twenty years, and that’s a question about your source water, your footprint, your disposal geography, and your operating staff. A removal curve won’t answer it.

FAQ

Which PFAS treatment technology is cheapest?

Over a twenty-year horizon, none of them reliably. GAC usually shows the lowest capital cost and can look cheapest at bid, but its operating cost is driven by media change-out frequency, which elevated total organic carbon in the source water can increase sharply against the vendor’s model. Ion exchange typically carries higher media cost with longer runs between change-outs and no reactivation option, so its disposal cost per cycle is higher. Membranes carry the highest energy and pretreatment load, and for an inland utility the cost of managing concentrate can exceed the treatment system itself. The comparison only becomes real when each option is priced with site-specific bed life from pilot data and with disposal costed at more than today’s rate. For scale, reactivation runs about $0.90 to $1.30 per pound of spent media and high-temperature incineration about $1.00 to $1.50, so a change-out schedule that turns out twice as frequent as modeled moves real money.

What happens to the PFAS after GAC or ion exchange removes it?

It leaves the water and concentrates in the media, and the utility becomes responsible for that media. Spent carbon can be thermally reactivated and returned to service at roughly $0.90 to $1.30 per pound, subject to capacity at a reactivation facility, and commercial reactivation capacity exists for carbon and not for other sorption media. Single-use resin has no equivalent thermal reactivation route and goes to incineration, around $1.00 to $1.50 per pound, or to landfill. On-site regenerable ion exchange is a separate option that consumes far less media and produces a concentrated liquid regenerant instead, with about nine full-scale systems running globally. High-temperature incineration became a more defensible path in 2026, after the Department of Defense lifted its prohibition for permitted units above 1,100 degrees Celsius and EPA reported promising thermal treatment results; the remaining doubt centers on lower-temperature units. Landfilling is cheapest and carries the most uncertainty, and EPA’s 2026 guidance says it may release more PFAS than the agency previously thought. Membrane systems don’t produce spent media; they produce a continuous liquid concentrate that needs its own disposal path.

Do destruction technologies replace GAC or ion exchange?

Not in most systems today. Utility-scale operating history is limited, and most deployed units treat concentrated side streams; full plant flow is rare. The practical near-term configuration is a proven removal step at the front, with a destruction process treating the resulting regenerant or concentrate at the back, which addresses the residuals problem without betting compliance on an early-stage technology. Before piloting one, ask for a fluoride mass balance, a full-panel byproduct analysis, energy consumption measured on a waste matrix similar to yours, and a reference installation treating a comparable stream.

How do we compare vendor proposals when they all claim compliance?

Move the evaluation off removal efficiency, since all three proven technologies can meet four parts per trillion when sized correctly. Give every bidder the same complete source water profile including TOC, nitrate, sulfate, iron, and manganese, and require a stated bed life with the basis for it. Require each proposal to name a disposal endpoint and the contingency if that endpoint closes. Ask for twenty-year lifecycle cost at current disposal rates and again at two to three times current rates. The proposals will separate quickly, and they’ll separate on the factors that determine what the plant costs to run in year twelve.

---

HydroKnowledge advises water utilities on technology selection and regulatory transitions, and advises water technology companies on positioning and go-to-market. Get in touch to talk through a PFAS treatment evaluation or a PFAS market strategy.