Learning curves used to forecast the cost of hydrogen appear more optimistic than they should be when they link cost reductions solely to doubling the installed capacity of electrolyzers. Electrolyzers are manufactured in factories, but hydrogen plants are built as integrated industrial projects, and a large share of learning and cost lies outside the stack manufacturing plant.
“Experience curves” rely on the assumption that each doubling of cumulative capacity represents another similar round of manufacturing and installing similar units. When this occurs, the historical rate of cost reduction can be used to extrapolate what may happen after future doublings. But this assumption is weaker in the electrolyzer sector, because capacity growth may result partly from manufacturing larger stacks, expanding shared equipment, and building larger facilities, rather than from repeating the same number of units.
Raw Figures Conceal More Than One Learning Mechanism
The material draws on a European study of electrolyzer projects published in 2025, which compiled capacity and capital-cost data for projects dating back to 2005. The unadjusted data showed a 23.3% cost reduction with each doubling of cumulative installed capacity for all projects, 32.1% for proton-exchange membrane (PEM) electrolysis projects, and 22.9% for alkaline electrolysis.
These rates appear close to strong historical learning curves for solar systems and batteries, but the picture changes when costs are normalized according to estimated project scale economies. The rates then fall to 13.3% for all projects, 17.6% for PEM technology, and 7.3% for alkaline technology, while the remaining relationship for alkaline technology is no longer statistically significant. This does not mean that costs did not decline; rather, the raw curve attributed multiple effects to “learning” at the same time, including scale economies and the transition from pilot projects to industrial facilities.
Installed Capacity Is Not Always a Good Measure of Manufacturing Repetition
The analysis provides a numerical example that illustrates the problem. If cumulative electrolysis capacity rises from 5 to 50 gigawatts, this represents 3.32 doublings of capacity. If the average stack size remains at 1 megawatt, the number of completed stacks rises from 5,000 to 50,000, meaning that the number of units has doubled at approximately the same rate.
But if the average stack size increases during the same period from 1 to 5 megawatts, 50 gigawatts of capacity would require only about 10,000 stacks. Installed capacity would still have increased tenfold, while the number of stacks would have doubled only once. Thus, gigawatt data suggest more manufacturing experience than the actual repetition at the level of the final equipment.
The number of stacks is not a complete measure either; larger stacks contain repeating cells, membranes, plates, and electrode areas, each with its own learning path. Improvements in materials and increases in current density can also raise output without a proportional increase in the quantity of materials. But the central point is that rapid expansion in equipment size changes what each “doubling” in the data means.
The Stack Is a Limited Part of Plant Cost
According to the International Energy Agency’s 2025 electrolyzer cost breakdown, the stack accounts for only about 15% to 20% of total installed capital cost. The “balance of plant” accounts for about 25% to 30% and includes power electronics, piping, compressors, and gas treatment. Engineering, procurement, construction, and contingencies may account for more than half of the cost.
Therefore, even a 20% reduction in the cost of a component that represents 20% of capital expenditure would reduce the cost of the integrated project by only about 4%, before accounting for potential improvements in the other components. Compressors, transformers, civil works, electrical connections, and construction do not automatically inherit the learning rate of the stack manufacturing plant; each has different scale economies and productivity curves.
Why Does This Matter?
The material does not conclude that electrolyzer projects will remain at the cost of first-generation facilities. Stack designs and materials can improve, engineering, procurement, and construction companies may standardize plant layouts, and repeated construction can reduce the errors associated with initial projects. But combining all these mechanisms into a single learning rate and then mechanically extending it to 2035 or 2050 may overestimate future reductions.
Capital-equipment cost is also not the final price of hydrogen. Electricity is a central operating cost, and cheap electricity from wind and solar may be intermittent, while capital equipment requires high utilization rates. After production come the costs of compression, storage, and distribution, each with its own infrastructure economics.
The editorial reading from certi.news is that a defensible forecast requires separating improvements in manufacturing, electrochemical performance, and stack size from plant-scale economies and the repetition of engineering and construction, and then separating all of these from electricity, utilization, financing, and logistics. Hydrogen will likely become cheaper than the cost of its first projects according to the information cited, but doubling global electrolyzer capacity alone is not enough to demonstrate a steep and sustained cost-reduction curve. The practical question is the cost of the complete system and which uses justify paying for it.