The Biggest Untapped Efficiency Reserve in Direct Air Capture Is Not a Better Material. It Is Smarter Timing.

Direct air capture has a public relations problem and a physics problem, and they feed each other. The PR problem is cost: the technology that pulls carbon dioxide straight out of the atmosphere remains stubbornly expensive, with disclosed deal prices averaging around $516 per tonne according to CDR.fyi data compiled in October 2026. The physics problem is quieter, but a study published this week suggests it might be more solvable. The air never stays the same, and the machines that harvest carbon from it barely notice.

Temperature swings, humidity surges, shifting electricity prices, and a power grid that is dirtier at some hours than others all change how much CO2 a direct air capture machine can grab and how much energy it burns doing so. Most machines, in models and in practice, run on fixed schedules anyway, regenerating their sorbents at the same temperature at the same cadence regardless of conditions. New research from the National Laboratory of the Rockies in Golden, Colorado, published in Frontiers in Environmental Engineering, asks what happens when the machine adapts instead. The answer, in the scenarios examined, is striking: roughly double the net climate benefit.

Why fixed schedules are a silent tax

To understand the finding, it helps to know how a leading class of capture machines works. In temperature-vacuum swing adsorption, a solid sorbent material grabs CO2 molecules out of passing air, then releases them when heated and placed under vacuum, in a six-stage cycle of vacuum extraction, heating, desorption, cooling, pressurization, and adsorption. Every cycle leaves residue: incomplete regeneration means the sorbent never fully clears, and the leftover loading carries over into the next cycle.

Because the sorbent's state at the start of each cycle depends on everything that happened before, performance cannot be predicted from any single steady-state snapshot. A machine that regenerates during a humid heatwave fights water that has co-adsorbed onto the sorbent, demanding extra thermal energy to boil it off. The same machine, regenerating a few hours later on cheap, clean electricity in dry air, would glide through the same cycle at a fraction of the energy.

Historically, engineers could not explore these tradeoffs properly because the physics is brutally expensive to simulate. Detailed models of adsorption systems were so computationally heavy that running a full year of hour-by-hour operation was impractical. The team, led by Keju An, solved that with a physics-informed reduced-order model: a stripped-down but physically faithful mathematical description that tracks the six coupled states that matter (fast and slow CO2 pools, free and bound water, gas-phase CO2, and bed temperature) forward through time.

120,000 simulations, one tiny neural net

The reduced-order model was benchmarked against published experimental data from Sendi and colleagues. Across twelve benchmark cases, it reproduced productivity with a mean absolute error of 10.2 percent and electricity demand with an error of 11.4 percent, with deviations growing largest under extreme conditions. Direction and relative sensitivity were faithful, which is what matters for large-scale screening.

To make the framework fast enough for serious optimization, the researchers generated roughly 120,000 cycle-resolved simulations, sampling across environmental conditions, operating strategies, system designs, and sorbent states. That dataset trained a compact neural network, a single hidden layer of 39 neurons, that reproduces the physics model's state transitions at a tiny fraction of the computational cost. The surrogate was then embedded inside a model predictive control framework, effectively giving the machine a weather-and-grid-aware operating schedule.

The biggest untapped efficiency reserve in direct air capture may not be new materials or bigger machines, but smarter timing.

Sensitivity analysis confirmed where the leverage is. Regeneration temperature dominated everything: raising it from 70 to 130 degrees Celsius cut annual productivity by roughly 10 tonnes of CO2 per cubic meter of bed while adding 0.7 to 0.8 megawatt-hours of electricity demand per tonne captured, a punishing double penalty from hotter, slower cycles. Humidity was the second most influential factor for energy, with relative humidity climbing from 0.1 to 0.9 adding roughly 0.5 megawatt-hours per tonne, a direct consequence of water co-adsorption.

Texas and Wyoming, hour by hour

The real test ran the adaptive controller against hourly 2025 weather, electricity price, and grid carbon intensity data for two contrasting American locations: humid Atoka, Texas, and dry, cooler Madison, Wyoming. A baseline machine running fixed settings completed about 1,094 cycles per year. The adaptive controller pushed that to roughly 1,449 cycles in Atoka and 1,379 in Madison, shifting energy-hungry regeneration steps into hours when electricity was cheap and clean and exploiting favorable humidity windows for capture.

In Atoka, electricity demand fell 18.8 percent while annual CO2 productivity rose 25.5 percent. In Madison, energy dropped 8.9 percent and productivity climbed 22.4 percent. The headline number is the carbon-adjusted net productivity, capture minus the emissions embedded in the electricity consumed, which nearly doubled at both sites, rising 95.2 percent in Texas and 94.4 percent in Wyoming. That is the number that matters for the climate: a tonne captured on a dirty grid is not the same as one captured on clean power.

One wrinkle deserves honesty. The economics did not always align with the climate math. Electricity cost fell slightly in Atoka but rose 3.7 percent in Madison, showing that carbon-aware control sometimes accepts higher operating costs to harvest cleaner power. For plant operators paid by the tonne captured rather than by the tonne net-removed, that tradeoff would need repricing.

What this does not prove

The authors are careful about scope, and so should everyone reading this. Their model is a control-oriented screening tool, not a plant-specific predictor. Because the surrogate learns from the physics model, the physics model's uncertainties propagate into the optimization results, meaning the reported gains are relative, model-based estimates rather than guarantees. No one has yet run a real capture plant on this controller for a year.

Still, the direction of travel is hard to argue with. The industry's roadmap for reaching affordable carbon removal has leaned heavily on novel sorbents and economies of scale: bigger plants, better materials, cheaper construction. This study points at a complementary lever that every existing plant can pull in software. That matters for the economics of a field where every marginal gain counts. Shaving nearly a fifth off energy demand while capturing a quarter more carbon is the kind of compounding improvement that, stacked over a decade, bends the cost curve toward the levels buyers need.

There is also a policy angle. Grid-aware carbon removal rewards the electricity system too: machines that schedule their heaviest loads for clean, cheap hours act as flexible demand, the same way smart charging does for electric vehicles.

We have spent a century engineering machines to run the same way no matter what the world does around them. The cheapest carbon removal breakthrough of the decade might not be a new chemical at all. It might just be paying attention.