Alkaline electrolyzers accounted for 84% of projects under construction worldwide in 2026, with PEM filling the high-purity and dynamic-response niche and AEM trailing as the third option still to prove membrane life at scale [S2][S5].
Western alkaline systems ran at about 54.6 kWh/kg versus roughly 55.1 kWh/kg for PEM at the system level in 2025, the narrowest gap the two technologies have ever shown in a published survey [S2]. Combined with a stack unit cost that remains below PEM, that efficiency parity is the single most important procurement fact for buyers in 2026.
Where alkaline and PEM actually win in 2026
Alkaline is the workhorse for large steady-power industrial sites, running on a nickel catalyst supply chain free of precious-metal risk, with stack life around 80,000 hours and field references that go back 30 to 40 years at MW to multi-hundred-MW scale [S5]. It is the technology China scaled to build roughly half of global green hydrogen capacity, and BNEF 2026 puts its 2026 share of projects under construction at 84% [S2].
PEM holds the high-purity, fast-cycling segment because it produces hydrogen at roughly 40 bar directly from the stack, which sidesteps a first-stage mechanical compressor, and its dynamic response handles variable renewable input without the slow ramp that penalizes alkaline [S3][S4]. For projects where the electricity profile is intermittent, or where hydrogen feeds directly into a fuel cell or a mobility application, PEM's operating envelope still beats alkaline on the duty cycle even when stack capex is higher.
Cost structure: stack vs BOP vs system
The headline rule is simple: alkaline stack capex per kW is lower than PEM, but PEM's balance-of-plant (BOP) cost grows more slowly as system size increases, so the crossover point sits near 10 MW according to the Fraunhofer ISE analysis cited by Plug [S3]. Above that size, PEM's avoided first-stage compression and its 4 to 40 times higher output pressure relative to a 1-10 bar alkaline system start to offset the higher stack cost, and PEM service costs run roughly one-third of alkaline service costs on a multi-year basis [S3].
U.S. DOE technical targets for low-temperature PEM stacks in 2026 set stack capital cost at 100 $/kW and uninstalled system cost at 250 $/kW against a 2022 baseline of 450 $/kW stack and 1,000 $/kW system, with a stretch ultimate target of 50 $/kW stack and 150 $/kW system [S1]. Whether the market reaches those targets depends on platinum-group-metal loading, which DOE tracks at 0.5 mg/cm² (0.1 g/kW) for 2026 against 3.0 mg/cm² in 2022, with iridium in the anode doing most of the heavy lifting on the PGM line [S1].
Efficiency, degradation, and lifetime numbers
Both technologies can land in a 55 to 70% electrical efficiency band depending on stack age and operating point, and PEM typically holds a small efficiency edge at part load where alkaline's older diaphragms and lower current densities show their age [S7]. The 2026 BNEF system-level numbers narrow that edge to roughly 0.5 kWh/kg, with alkaline actually ahead on the system efficiency line in Western units [S2].
DOE's 2026 PEM targets include a stack electrical efficiency of 48 kWh/kg (69% LHV), a system energy efficiency of 51 kWh/kg (65% LHV), a stack average degradation rate of 2.3 mV/kh (0.13%/1,000 h), and a stack operating life of 80,000 hours [S1]. Both status and target numbers include a 2.5% efficiency allowance for H₂ crossover as an upper safety limit, and end-of-life is defined as 10% voltage loss from beginning-of-life at the same current density [S1].
Supply-side constraints: iridium, PFAS, and the materials ceiling
Global iridium production sits at roughly 7-8 tonnes per year, a hard ceiling that caps how much PEM capacity the world can physically build in any given year, and that bottleneck is the single biggest reason AEM was designed without iridium in the first place [S5]. Platinum on the cathode side is less of a constraint but still adds to the PGM load, and DOE's 2026 PGM-loading target of 0.5 mg/cm² combined is the metric to watch for whether PEM stack capex can actually fall to the 100 $/kW line [S1].
The other structural risk on PEM is regulatory: Nafion-type PEM membranes are PFAS, and the EU PFAS restriction is moving through ECHA evaluation in 2026 and a Commission decision in 2027, with take effect expected in 2028-2029 [S5]. Buyers specifying PEM for European projects after 2026 should treat the membrane as a regulatory variable, not a fixed input, and this is one of the supply-side pressure points tracked in the PTFE, PVDF, FKM price 2026: where the cost is moving under PFAS rules coverage as well. Alkaline, with its KOH/NaOH liquid electrolyte and asbestos-free or polymer diaphragms, sits outside the PFAS restriction pathway entirely.
Decision matrix: pick by duty cycle, not by headline
For a project that runs at steady full load 8,000 hours a year, has cheap grid or dedicated renewable power, and sits in a region where footprint is not the binding constraint, alkaline is the correct call: lowest stack capex, no precious-metal exposure, and the longest reference list. For a project that needs fast ramp, partial-load efficiency, high hydrogen purity, or direct delivery at 30-40 bar into a pipeline or refueling station, PEM wins on the operating envelope even at a higher stack $/kW [S2][S3][S5].
AEM is the right answer only for buyers willing to accept pilot-scale risk on membrane durability and to wait on volume manufacturing economics through the late 2020s, and even then only when the application needs PEM-like response without the iridium or PFAS exposure [S5]. Procurement should weight four questions in order: what is the electricity duty cycle, what pressure does the downstream require, what is the regulator's view on PFAS in the target market, and what is the project's tolerance for the 2-3x footprint of an alkaline plant [S5].
What the field data does not yet show
BNEF 2026 and the World Bank's February 2026 technical-economic study both warn that there are no industry standards for how degradation, efficiency, minimum load, or restart times are calculated across vendors, and most published figures are still lab-derived rather than verified by long-term field operations [S2]. PEM has scaled past 50 MW in operation, but long-term field data at that scale is still limited, so any capex comparison that ignores the degradation-rate assumption is overstating certainty on both sides [S2].
Buyers in 2026 should ask each OEM to state the same stack current density, the same end-of-life voltage loss definition, and the same dynamic-duty-cycle profile before comparing numbers side by side, and treat OEM scope-of-supply differences (BoP inclusion, transformer, water purification) as a first-order cost variable rather than a footnote [S2]. For project owners who already have a hydrogen offtake agreement but face the delivery slippage seen across 2024-2026, the Electrolyzer Delivery Delays and Project Cancellations Reshape the 2026 Hydrogen Build coverage lays out which delivery promises are still holding and which slipped, and that data should feed into the alkaline-vs-PEM call when a 2027 commissioning date is at stake.
Trackable signals for the next quarter: the EU Commission's PFAS restriction decision expected in 2027, any update to the DOE 2026 PEM technical-target scorecard (PGM loading and stack capex), and the first set of >50 MW PEM field-degradation reports from Western plants commissioned in 2024-2025.
For the relevant spec sheets and selection criteria, see power supply, dc power supply, and switching power supply.