The System Case Against Hydrogen for Grid Storage

Support CleanTechnica’s work through a Substack subscription or on Stripe.

Every time hydrogen is critiqued as an energy carrier for the power sector, the same question reappears. If not hydrogen, where does long duration storage come from? I received it related to my recent critique of Germany’s attempt to force the EU to double green hydrogen and synthetic fuel quotes for transportation to justify their already built and already stranded hydrogen infrastructure. It is a fair question, if naive. It is also the hundredth iteration of it, and the repetition suggests that the frame is wrong. The premise that hydrogen is required for long duration storage only holds if the system boundary is drawn too narrowly and the problem is defined too simply.

Before debating molecules, we have to define what problem we are solving. Storage is not one thing. Grid operators manage seconds to minutes of response for frequency and voltage stability. They manage intra day shifting to move solar from midday into the evening peak. They manage multi day variability driven by weather patterns. Finally, in a few geographies, they plan for rare multi week low wind and low solar periods that are often described in Germany as dunkelflaute. The first two categories are daily operational realities. The third is manageable variability. The fourth is a strategic reserve problem. Much of the hydrogen debate jumps straight to the fourth category and treats it as if it defines the entire system.

Seconds to minutes of response are already dominated by batteries and power electronics. Lithium-based battery systems respond in milliseconds. A 100 MW battery providing frequency response can earn revenue in multiple markets while cycling only small fractions of its capacity. These services do not require hours of storage. They require speed. Hydrogen is not even a candidate here because the round trip through electrolysis and turbines cannot respond in milliseconds and would destroy equipment if cycled at that frequency.

Intra day shifting is increasingly the domain of lithium batteries as well. In the United States and Europe, four hour systems are common and eight hour systems are appearing regularly in procurement. Fully built, grid connected utility scale systems outside China are now often in the $100 to $150/kWh range, with BloombergNEF reporting global averages near $117/kWh and large Chinese tenders signaling equipment and lifecycle service stacks closer to $65/kWh at scale. At $125/kWh, an eight hour 100 MW system is 800 MWh and roughly $100 million of capital. If it cycles daily, that is 292 GWh of annual throughput. Over 20 years, that is about 5.8 TWh delivered. Even before layering in financing, augmentation, and O&M, that implies capital of roughly $0.017/kWh spread over lifetime throughput. High utilization matters. An asset that runs daily has a fundamentally different economic profile than one that runs a few hours per year.

Pumped hydro has been providing 8 to 24 hours of storage for decades. China alone has about 365 GW of pumped hydro in operation, under construction, or approved for construction by 2030, and whatever China does at massive scale the rest of the world really should pay attention to. With typical durations of 12 to 24 hours, that represents on the order of 10 to 15 TWh of storage capacity. Round trip efficiency is commonly 75% to 85%. Asset life is measured in 50 to 100 years. Capital costs vary widely, but once built, marginal operating costs are low and utilization can be daily. These are not pilot projects. They are infrastructure at continental scale.

Flow batteries are moving from demonstration to deployment. By separating power components from energy storage tanks, they decouple power and duration. If a 50 MW system requires more hours, additional tank volume can be added without duplicating inverters and transformers. That makes 10 to 24 hour systems more cost effective per added hour than lithium systems in some cases. Round trip efficiency is often 65% to 75%, still far higher than hydrogen pathways that can fall below 35% when accounting for electrolysis, compression, storage losses, and reconversion. While they’ve often been encapsulated as complete systems in containers historically due to the low requirement for longer duration storage, OEMs are starting to unbundle them, decoupling power and energy.

As an aside, it’s also far better than the SEC filing documented 36% round trip efficiency of Form Energy’s apparently inflexible iron air batteries, which my electrochemistry acquaintances tell me is lab efficiency that is unlikely to be replicated in the real world at scale. Form of course remains silent on the problem of inevitable hydrogen formation inside the closed cells in the presence of oxygen and electricity, a major explosion risk drawback of the iron air chemistry that’s well understood outside of venture capital due diligence and marketing material from the firm. There’s likely a reason why Form Energy is pivoting—for the fourth time by my count—to attempting to be a green iron manufacturer instead of an energy storage firm.

Multi day variability is more complex, but it is still not automatically a hydrogen problem. Large grids smooth variability across regions. High voltage direct current interconnectors can move gigawatts across hundreds or thousands of kilometers with losses around 3% per 1,000 km. Northern Europe already shares wind and hydro across borders. Norway’s reservoirs effectively provide multi day and even seasonal flexibility for Denmark, Germany, and the Netherlands. Demand response shifts industrial loads by hours or days. Overbuilding wind and solar by 10% to 20% reduces the depth of shortfalls. Current constraints of transmission are simply that, current constraints that will be overcome by expanding transmission, dynamic line rating and reconductoring. None of these require building a parallel hydrogen grid.

The hydrogen argument tends to crystallize around rare multi week events in northwestern Europe. True multi week low wind and low solar events occur, but they are infrequent. A system that is 80% to 90% renewable can experience several days of low output every year and a deeper event once every decade or two. That is a strategic reserve problem. Strategic reserves are not daily balancing tools. They are insurance. Insurance assets do not need to operate frequently. They need to be available when required.

Insurance does not require inventing a new fuel or building new energy storage facilities. Europe already maintains large strategic gas reserves. Germany alone has underground gas storage capacity on the order of 20 to 25 billion cubic meters. One cubic meter of methane contains about 10 kWh of energy. Twenty billion cubic meters represent roughly 200 TWh of primary energy. Even accounting for 55% efficiency in modern gas turbines, that is over 100 TWh of dispatchable electricity potential. A severe two week event requiring 50 GW of backup generation would consume about 16.8 TWh of electricity. That is well within existing storage capacity.

The overlooked resource in this discussion is anthropogenic biomethane. Agriculture, landfills, and waste streams emit methane equivalent to hundreds of millions of tons of CO2 annually. Capturing as much of that methane as possible at point sources such as landfills and dairy barns is climate mitigation regardless of grid design. Anaerobic digesters convert waste into biogas and nutrient rich solids, avoiding untrapped methane creation from the same biomass. After upgrading to pipeline quality, biomethane can be injected into existing gas networks and stored in existing caverns. Europe’s technical biomethane potential is often estimated at 30 to 40 billion cubic meters per year. At 10 kWh per cubic meter, that is 300 to 400 TWh of primary energy annually. Not all of that is economical or sustainable, but even a fraction covers rare electricity shortfalls as well as methane for industrial feedstocks.

Maintaining existing gas turbines and filling reserves with biomethane avoids building a hydrogen stack that includes electrolyzers, compressors, pipelines, salt cavern storage, and hydrogen turbines. The $500 to $1,000 per kW figures often cited are closer to stacks or best case equipment pricing, not the full installed facility cost once you include power conversion, water treatment, cooling, buildings, safety systems, controls, electrical tie-in, and EPC. The International Energy Agency’s Global Hydrogen Review 2025 puts the cost of making and installing an electrolyser outside China in 2024 at about $2,000 to $2,600 per kW, and real project FIDs are landing in that range.

Reuters’ reporting on the Air Liquide and TotalEnergies FID for two Dutch projects points to over €1 billion of investment for 200 MW and 250 MW class facilities, which is consistent with roughly €2,000 to €3,000 per kW before you add the rest of the hydrogen system. Using $2,000 to $2,600 per kW as the economic baseline, a 10 GW electrolysis buildout is $20 to $26 billion for electrolysis plants alone, before compression, salt cavern development or modification, pipelines, and hydrogen capable generation. If that system operates 2% of the year, its capacity factor is 175 hours annually. The levelized cost of energy from such low utilization assets is high even before accounting for round trip losses that commonly leave only 30% to 40% of the original electricity available after electrolysis, storage, and reconversion.

If you add, conservatively, another $5–$10 billion for storage, pipelines, and generation upgrades, you are quickly north of $2,500–$3,000/MWh on a capital-only basis for delivered electricity in a 2% utilization scenario. Is insurance really worth 100x the cost of electricity?

The biomethane pathway is not a single use storage asset. It is a stacked valorization chain across sectors that already exist. Methane from agriculture, landfills, and wastewater carries a high global warming potential, roughly 28 to 34 times that of CO2 over 100 years. Capturing it avoids emissions that would otherwise face tightening regulation under schemes such as the EU Emissions Trading System. Avoided methane leakage reduces compliance risk for operators and avoids future carbon costs. Anaerobic digestion produces digestate rich in nitrogen, phosphorus, and potassium, which substitutes for synthetic fertilizers whose production is energy intensive and carbon priced in Europe, something farmers will pay for.

Upgraded biomethane injected into the gas grid displaces fossil natural gas and reduces ETS exposure for industrial users. When used in direct reduced iron production, it lowers emissions intensity and reduces the number of ETS allowances steel producers must surrender. The biogenic CO2 stream separated during direct reduction of iron can be routed to greenhouse agriculture, where growers would otherwise require fossil CO2. Each step creates measurable economic value through avoided carbon costs, avoided fossil fuel purchases, and avoided synthetic fertilizer production, all things which real economic actors will pay for, before any residual volumes are allocated to strategic electricity reserves.

Hydrogen built primarily for grid backup does not stack value at all, rather destroys it at each step. Electrolyzers consume electricity that could otherwise serve direct loads or reduce fossil generation. The hydrogen has no inherent carbon avoidance value unless it displaces a fossil input elsewhere, and when used only for rare strategic electricity generation it sits idle for most of the year. The capital invested in electrolysis, compression, storage, and reconversion assets earns little revenue outside emergency periods. In ETS terms, hydrogen produced from renewable electricity may help an industrial user avoid carbon costs if it replaces fossil hydrogen, but hydrogen produced solely to reconvert to electricity during rare grid events does not generate layered carbon savings across agriculture, fertilizer markets, industrial decarbonization, and CO2 reuse. It is largely a single purpose insurance asset. Biomethane, by contrast, passes through multiple economic value nodes, each of which either avoids carbon pricing exposure under schemes like the EU ETS or displaces carbon intensive inputs, before finally serving as low utilization strategic generation fuel.

A large part of the long duration storage argument is driven by the assumption that winter heat requires molecules. It does not. Aquifer thermal energy storage systems store summer heat in underground water bearing layers and recover it in winter. Borehole thermal energy storage stores heat in rock volumes. Seasonal pit storage systems in Denmark shift heat across months at costs per kWh of storage that are lower than electrochemical systems. If district heating networks use large heat pumps and seasonal storage, winter electricity peaks fall. Lower peaks reduce the scale of backup generation required.

Finally, there is demand response. For extremes of dunkelflaute conditions every 10 to 50 years, the assumption that all industries will stay open during low energy availability periods will fall by the wayside. We already manage many industries with demand response contracts for daily peaks, and extending that to staff vacations every one to five decades is very manageable. Knowing that it might occur will require power heavy industries to stockpile a bit more, and climactic and weather modeling will enable them to know how much. That turns into a cost which factors into negotiations with grid operators over contract terms. There’s nothing unusual about. Refineries on the Gulf Coast of the USA shut down when hurricanes threaten all the time.

In the Netherlands, scenario modeling I was involved in last year with transmission systems operator TenneT using the open source Energy Transition Model, which includes hourly weather data for multiple representative years including high dunkelflaute years, shows this effect clearly. Expanding aquifer thermal storage, maintaining combined heat and power units for strategic use, maximizing biomethane for industrial feedstocks such as direct reduced iron, and routing biogenic CO2 to greenhouse agriculture can create a near balanced mass flow of about 5 million tons of CO2 per year in one region. In those scenarios, hydrogen for energy shrinks to a small residual category tied to industrial processes rather than grid balancing, and one mostly provided by existing by product hydrogen such as that from chloralkali plants.

Eliminating hydrogen as a bulk energy carrier reduces total system energy demand because hydrogen pathways are inefficient. If electrolysis is 70% efficient and turbine reconversion or fuel cells are 55% efficient, round trip efficiency is about 38%. That means 2.6 units of electricity in for one unit out. Avoiding that pathway reduces required generation capacity. In the Netherlands exercise, avoiding hydrogen allowed TenneT to eliminate the requirement for the expensive nuclear plant and the most expensive offshore wind in the scenario we were adjusting for pragmatism.

The capital allocation question is central. An asset that operates 300 cycles per year spreads its capital cost over hundreds of revenue events. An asset that operates 100 hours per year spreads capital over a handful of events. If a hydrogen system costing $10 billion delivers 10 TWh of electricity in a severe year and near zero in mild years, its effective cost per kWh over its life is high. Maintaining existing gas turbines, which are already depreciated, and filling them with biomethane leverages sunk capital rather than duplicating it.

Some argue that producing hydrogen directly at wind turbines through integrated electrolysis avoids transmission investment. Electricity must still be moved from dispersed turbines to centralized storage or end users unless hydrogen pipelines are built to the same scale as today’s electricity networks. Compressing hydrogen to 200 to 700 bar requires additional energy. Cavern storage requires drying and conditioning. Reconversion requires turbines or fuel cells. This substitutes one infrastructure stack for another rather than eliminating infrastructure. I assessed the DNV study on this and found it sadly lacking in realistic assumptions.

Another point is that northern latitudes lack consistent solar and require molecular storage, while sunny regions can rely on batteries. Many parts of the world do not experience prolonged low wind and low solar periods. Equatorial regions have limited seasonal variation in solar. Regions with strong hydro resources already have seasonal storage embedded in reservoirs. Even within Europe, southern countries have different weather correlations than northwestern Europe. Designing a global hydrogen energy backbone around the most challenging weather pattern in one region does not make economic sense for the rest of the world. Europe’s decarbonization challenge should not be exported to Africa, slow their decarbonization of their economies.

Hydrogen does have a role. Ammonia production requires hydrogen. Biofuel processing and upgrading pathways can require hydrogen inputs. Some industrial processes need hydrogen as a feedstock. These uses justify electrolyzers where alternatives do not exist, but not necessarily in Europe. Treating hydrogen as a bulk energy storage medium for the power sector extends it beyond those roles and assigns it to a task where it is neither the most efficient nor the most capital efficient option.

The recurring question about long duration storage assumes that one technology must fill a single gap. In practice, reliability emerges from layering solutions. Batteries handle seconds to hours. Pumped hydro and flow batteries handle hours to a day. Interconnection, demand flexibility, and modest overbuild handle multi day variability. Seasonal thermal storage reduces winter peaks. Existing gas assets running on declining volumes of biomethane handle rare strategic events. When the system boundary is widened and the math is done, hydrogen as a grid scale energy storage backbone is not necessary. It becomes one option among many, and far from the most economically compelling one.