An LCOH sensitivity analysis tells you which assumptions could break a hydrogen project before you pay for a feasibility study. In the worked 100 MW example below, operating hours and power price move the cost of hydrogen most, and moving six inputs to the bad end of their ranges at once takes the cost from $5.72/kg to $8.61/kg.
Short answer: set a base case, move one input at a time across a range you can defend from public data, rank the swings, then find each input's break point: how far it can slip before the project misses its price. If a plausible slip in one input breaks the project, that input is what the feasibility study must pin down first.
This is general pre-feasibility guidance, not engineering, investment or financial advice. Confirm all figures with vendor quotations, power offers and qualified advisers.
The investor's question: do the numbers hold up?
A pitch deck usually shows one levelized cost of hydrogen (LCOH). An investor or analyst needs to know something different: how wide the honest range is, and which single assumption carries most of the risk. A $5/kg project that becomes $8/kg if the power contract slips is a different investment from one that stays under $6/kg in every reasonable case.
Our hub guide on what LCOH is and how it is calculated explains the cost lines. This article tests them. The method is simple enough to run before any money goes into a feasibility study, and it tells you which questions that study should answer first.
Documented inputs and how far they can move
A stress test is only as good as its ranges. Use ranges you can trace to a source, not round numbers chosen to look reassuring. Public reference data gives a defensible starting point for most inputs:
- Installed electrolyzer cost. The IEA Global Hydrogen Review 2026 assumptions annex puts the 2025 installed cost of water electrolysis at $1,920–2,500/kW as a global average and $550–1,100/kW in China. Its figure includes the electrolyzer system, electrical equipment, gas treatment, balance of plant and EPC, but not the power generation assets.
- Fixed operating cost and stack life. The same annex assumes fixed operating cost of 3% of capital cost a year and a stack life of 50,000 operating hours, noting stacks can reach up to 95,000 hours.
- Electricity use and stack life range. The U.S. Department of Energy's PEM electrolysis targets give a 2022 system status of 55 kWh/kg with a 51 kWh/kg target for 2026, and a stack life of 40,000 hours today against an 80,000-hour target.
- Electricity use over plant life. DOE Hydrogen Program Record 24005 (May 2024) uses an average of 57.5 kWh/kg over life, against 55.2 kWh/kg at beginning of life, because stacks degrade.
- Cost of capital. The same DOE record uses financial assumptions chosen to give a weighted average cost of capital of about 10% nominal, and notes that assumptions such as WACC and analysis period "can highly influence the LCOH."
- Renewable power price and hours. The IEA annex lists 2025 generation costs for good resource conditions in North America of $37/MWh for solar PV at 2,480 full-load hours and $36/MWh for onshore wind at 3,640 hours.
DOE's own results show why a single number is not enough. Record 24005 reports roughly $5–7/kg in 2022 dollars, without subsidies, for an installed cost averaging $2,000/kW, renewable power around $0.03/kWh and capacity factors of about 50–75%. Its solar PV case at about 35% capacity factor reaches about $9/kg.
Worked example: an LCOH sensitivity analysis for a 100 MW plant
All inputs below are illustrative assumptions, not quotes, project data or H2Hub outputs. The base case is deliberately the same as the 5,000-hour case in our guide to electrolyzer utilization and the cost of hydrogen, so you can compare the two.
Base case (stated assumptions):
- 100 MW electrolyzer, $2,200/kW installed (inside the IEA global range)
- 5,000 full-load hours a year, for example a hybrid wind and solar supply with some oversizing
- Electricity at $40/MWh, paid only on energy used
- 55 kWh/kg system electricity use (DOE 2022 status)
- Fixed operating cost 3% of capital cost a year (IEA annex)
- Stack replacement $300/kW every 50,000 operating hours, charged per hour run (stack life from the IEA annex)
- 8% discount rate over 20 years, giving a capital recovery factor of 0.10185
- Water, oxygen sales, compression, storage, grid fees, taxes and incentives excluded
Base result: about 9,091 tonnes of hydrogen a year at $5.72/kg, made up of $2.46 capital recovery, $0.73 fixed operating cost, $2.20 electricity and $0.33 stack replacement.
Step 1: move one input at a time
Each input is moved to the low and high end of a range, with everything else held at the base case.
| Input | Range tested | Basis for the range | LCOH at favourable end | LCOH at unfavourable end | Swing |
|---|---|---|---|---|---|
| Full-load hours | 6,000 to 4,000 | ±1,000 hours around base | $5.19 | $6.52 | $1.33 |
| Power price | $30 to $50/MWh | ±$10/MWh around base | $5.17 | $6.27 | $1.10 |
| Installed cost | $1,920 to $2,500/kW | IEA 2025 global average range | $5.31 | $6.16 | $0.84 |
| Discount rate | 6% to 10% | DOE record uses about 10% WACC | $5.37 | $6.10 | $0.73 |
| Electricity use | 51 to 57.5 kWh/kg | DOE 2026 target to DOE lifetime average | $5.30 | $5.98 | $0.68 |
| Economic life | 25 to 15 years | ±5 years around base | $5.52 | $6.08 | $0.56 |
| Stack life | 80,000 to 40,000 hours | DOE target to DOE 2022 status | $5.60 | $5.80 | $0.21 |
Sorted by swing, this table is the data behind a tornado chart. In this example, operating hours and power price are the two inputs to pin down first, and stack life barely matters at screening stage. That ranking is specific to these ranges: a wider power-price range or a narrower hours range would reorder it, which is why the ranges must be defensible.
Step 2: find each input's break point
A swing shows sensitivity. A break point shows risk. Suppose the project needs hydrogen at no more than $6.72/kg, $1.00 above the base case, to meet a buyer's price. How far can each input slip on its own before the project misses?
| Input | Base | Break point at $6.72/kg | Slip allowed |
|---|---|---|---|
| Full-load hours | 5,000 | about 3,807 | about 1,190 fewer hours (24%) |
| Power price | $40/MWh | about $58.2/MWh | about $18/MWh (46%) |
| Installed cost | $2,200/kW | about $2,889/kW | about $690/kW (31%) |
| Discount rate | 8% | about 13.1% | about 5 percentage points |
| Electricity use | 55 kWh/kg | about 64.6 kWh/kg | about 9.6 kWh/kg (17%) |
Read the two tables together. Power price has a large swing but also a lot of headroom before it breaks the project. Full-load hours combine a large swing with the smallest proportional headroom, so a supply that delivers a quarter fewer hours than planned is enough to miss the price. That is the input to challenge hardest in any power offer. Our guide to the renewables-to-electrolyzer ratio shows how oversizing changes those hours.
Step 3: test the combined downside
Inputs rarely fail alone. Moving six inputs to their unfavourable end together (4,000 hours, $50/MWh, $2,500/kW, 10% discount rate, 57.5 kWh/kg and 40,000-hour stacks, with life held at 20 years) gives:
| Cost line ($/kg) | Base case | Combined downside |
|---|---|---|
| Capital recovery | $2.46 | $4.22 |
| Fixed operating cost | $0.73 | $1.08 |
| Electricity | $2.20 | $2.88 |
| Stack replacement | $0.33 | $0.43 |
| Total | $5.72 | $8.61 |
Moving only the three most market-driven inputs (hours, power price and discount rate) gives $7.54/kg. The same six inputs at their favourable end give $3.63/kg. A credible screening result is therefore a range, roughly $3.6–8.6/kg here, with the most likely outcome near the base case and the downside driven mostly by fixed costs spread over fewer hours. Our electrolyzer cost per kW guide explains why installed cost should be tested on a full scope, not an equipment price.
What this simple stress test leaves out
This is a one-at-a-time and corner-case test on an annual model. It does not:
- Capture correlations. Cheap power often comes with fewer hours. Test linked pairs together rather than treating them as independent.
- Assign probabilities. A Monte Carlo analysis can do that later, but only once the ranges are well founded. At screening stage, ranges matter more than distributions.
- Model hourly dispatch. Minimum load, part-load efficiency, storage and delivery profile all need an hourly model.
- Include revenue or policy. Tax credits, subsidies, oxygen sales and the rules on how power is sourced can change the answer in either direction. Check them for your jurisdiction on the date you decide.
A five-step stress-test checklist
- Write down the base case and its sources. Every input should have a source, a date and a range.
- Set ranges you can defend. Use published ranges, real quotes and the terms of actual power offers, not ±10% on everything.
- Rank the swings. The top two or three inputs are the feasibility study's first questions.
- Find the break points against the price the project needs. Small headroom on a big-swing input is the warning sign.
- Run the combined downside. If the project only works when everything goes right, say so before anyone commits money.
Try it in H2Hub
You can start a project in H2Hub for free, without an account or a card, for the site you are screening. Keep this example's base case beside you: 100 MW, $2,200/kW installed, 5,000 full-load hours, $40/MWh power and 55 kWh/kg, and see where your project's inputs differ.
Your first full project and its PDF report are free; downloading the PDF needs a free sign-in. Re-running with new inputs, such as the downside power price or fewer hours, or starting a second project requires a subscription, so it is worth choosing your strongest base case first. The LCOH Lab is a subscriber tool for working through cost inputs in more depth; signed-out visitors see an introduction.
A note on pre-feasibility
The numbers above are calculations from stated assumptions. They are not forecasts, market prices or H2Hub-generated results, and they do not replace a feasibility study, vendor quotes or financial advice.
Frequently asked questions
What is an LCOH sensitivity analysis?
It measures how much the levelized cost of hydrogen changes when each input changes. The usual form moves one input at a time across a range and ranks the results, often shown as a tornado chart. It identifies which assumptions deserve the most scrutiny.
Which input affects the cost of green hydrogen most?
It depends on the project and the ranges tested. In our example, full-load hours and electricity price had the largest swings. DOE's Record 24005 also says LCOH depends strongly on capacity factors and electricity prices.
How does the discount rate change LCOH?
A higher discount rate raises the annual cost of recovering capital. In our example, moving from 6% to 10% raised the cost from $5.37/kg to $6.10/kg. Capital-heavy, low-hour projects are the most exposed.
Should I use Monte Carlo simulation at screening stage?
Usually not first. Monte Carlo adds probabilities, but its results are only as good as the ranges and distributions behind them. Start with one-at-a-time tests and break points, then add probabilities once the key ranges are backed by quotes.
Primary sources and further reading
- International Energy Agency — Global Hydrogen Review 2026: Assumptions Annex. Electrolysis installed cost and scope, fixed operating cost, stack life; renewable generation costs and full-load hours.
- U.S. Department of Energy — Technical Targets for Proton Exchange Membrane Electrolysis, page updated 29 May 2026. System electricity use and stack life, status and targets.
- U.S. Department of Energy — Hydrogen Program Record 24005, clean hydrogen production cost with PEM electrolyzers, 20 May 2024. Scenario costs, lifetime electricity use and financial assumptions.
Editorial verification: 2026-11-10. Worked examples are calculations from stated assumptions, not forecasts, market prices or H2Hub-generated results.
