Blue Hydrogen: The Cheapest Clean H2 — So Why Is It Losing?

YouTube Spotify Apple

Key Takeaways

• It is a split verdict. As a new energy vector — the thing that was supposed to displace fuels — blue hydrogen is overhyped, and four dead projects in eight months are the market saying so out loud. As a retrofit — existing molecule, existing storage, existing customer — it is a real bet.
• Making hydrogen was never the hard part. The carbon is. It doesn’t disappear; it only changes form and address. There are exactly three things you can do with it: bury it (blue), sell it (turquoise), or skip hydrogen and burn the gas for electricity — because a data center will definitely buy that.
• The fleet captures ~60% by design; the standards want 90–95%. The cheap, concentrated process stream gets captured; the flue gas is left alone. And behind that sits a wall no capture rate can touch: upstream methane. At the US-average 1.5% leakage, even 100% capture stays above 2.5 kg CO₂e per kg H₂.
• One subsidy carries the math. Grey runs $1.50–2.50/kg, blue $2.00–3.50. 45Q pays $85/tonne (EOR now at parity), which converts to ~$0.77/kg at ~9 kg captured — about three-quarters of the blue-grey gap. Canada adds a 50% capture-equipment ITC; Alberta’s nominal C$95 carbon price trades nearer C$20.
• What actually outbid blue hydrogen wasn’t green or turquoise. It was the AI data center — the Teesside site is literally becoming one, and Monolith is reportedly planning one on its own plant.

The Graveyard: Eight Months, Four Projects

Start with the obituaries, because they are the evidence. In November 2025 ExxonMobil paused Baytown — roughly a billion cubic feet of gas a day, one of the largest clean-hydrogen plants ever attempted. The CEO’s reason was not technical: no signed offtake customers. In December, bp withdrew H2Teesside, a 1.2 GW project whose application had been pending since March 2024. The site’s new plan: an AI data center. Air Products stopped spending on its Louisiana blue-ammonia complex and listed the capture and ammonia pieces for sale, while its Alberta project — announced in 2021 at $1.3 billion for 2024 delivery — is now a $3.3 billion, end-2027 asset management itself calls underperforming, with the CEO admitting part of the miss was self-inflicted. And Suncor + ATCO, 300,000 tonnes a year announced in 2021 with FID planned for 2024, still has no FID; Alberta’s major-projects inventory now says earliest 2029.

Four companies, four balance sheets, four different technologies of excuse — one shared cause of death. Nobody would sign the contract.

What failed was not the plant. It was the business model: a premium product with no premium buyer.

It’s Not a Technology Problem

Steam methane reforming is a century-old process that makes about 95% of the world’s hydrogen. Shell’s Quest facility in Alberta has been injecting CO₂ since 2015 — roughly nine million tonnes stored so far. The equipment works, the geology works, and on paper blue is the cheapest clean hydrogen there is: $2.00–3.50 per kilogram against green PEM at roughly $5. IEA’s 2026 outlook puts the theoretical floor in cheap-gas regions just above $1/kg — with the important footnote that no operating project reaches it at the 95% capture rate the marketing implies.

So the failure is not in the reactor. It is in the spreadsheet — and specifically in the two places the spreadsheet is honest about: how much carbon you actually catch, and what the gas leaked before it ever arrived.

The Capture-Rate Gap, and the Methane Wall Behind It

The running SMR-plus-CCS fleet captures about 60% of its CO₂ — by design, not malfunction. An SMR makes CO₂ in two places: a concentrated process stream that is cheap to capture, and a dilute flue gas that is not. Most projects capture the first and vent the second. Standards and subsidy rules want 90–95%. The technology that can get there — autothermal reforming — targets 95% and is projected to grow from roughly zero today to about 37% of new capacity by 2030.

Then the wall. Methane warms more than 80 times as much as CO₂ over twenty years, and it leaks upstream of any capture box you build. At the US-average 1.5% leakage rate, even a plant capturing 100% of its stack CO₂ still lands above 2.5 kg CO₂e per kg of hydrogen. At 0.4% leakage with 90% capture you are near 2 kg. Which means “blue” is not a property of the plant. It is a property of the gas source — and most gas sources cannot sell you a clean one.

Does It Pencil?

Grey hydrogen costs $1.50–2.50/kg; blue costs $2.00–3.50. Call the premium a dollar, for a molecule the customer cannot tell apart. The entire case for blue therefore runs through policy, and policy mostly runs through one number: 45Q at $85 per tonne of CO₂ stored geologically — and since the July 2025 budget bill, enhanced oil recovery and utilization pay the same $85. Capture about 9 kg of CO₂ per kilogram of hydrogen and the credit is worth ~$0.77/kg: three-quarters of the blue-grey gap, gone.

Canada stacks a refundable 50% investment tax credit on capture equipment on top. But Alberta’s TIER market shows how soft the floor can be: a nominal industrial carbon price of C$95/tonne, against compliance credits actually trading near C$20. The spreadsheet says 95; the market pays 20. Blue pencils where the subsidy is bankable and the storage is already proven — and almost nowhere else.

Where Blue Actually Works: Four Conditions

There is exactly one big blue FID to study, so study it: Shell’s Polaris, about 650,000 tonnes of CO₂ a year at Scotford. Read the four conditions straight off it. Existing molecule — the refinery already uses the hydrogen. Existing storage — Quest, in the ground since 2015. Existing customer — Shell’s own plant. Existing permits. Blue as a retrofit of an asset you already own, decarbonizing a molecule you already sell to yourself.

Miss any one of the four and you are asking a stranger to pay a premium for a commodity, with storage you haven’t permitted, on policy that can be repriced. That is the graveyard list.

Turquoise: The Honest Answer with a Small Door

Methane pyrolysis is the one route that takes the carbon problem seriously: split CH₄ into hydrogen and solid carbon and there is no process CO₂ at all — nothing to capture, compress, inject or monitor for a century. It sips 7–12 kWh per kilogram, about a fifth of electrolysis, and uses no water. Monolith has run commercially in Nebraska since 2020; BASF and ExxonMobil signed a joint demonstration at Baytown — 2,000 tonnes of hydrogen, 6,000 tonnes of solid carbon — weeks after the blue pause.

The catch is pure arithmetic. The chemistry is fixed: every kilogram of hydrogen brings 3 kilograms of solid carbon. Sell it into the carbon-black market — about 18 million tonnes a year globally — and the route caps out near 6 million tonnes of hydrogen: ~6% of world demand. A peer-reviewed analysis in April 2026 flags precisely this: carbon supply outrunning real carbon demand is a fundamental oversupply risk that breaks the techno-economic assumptions. Turquoise solves the molecule and hits a market-size wall. Worth watching — and it inherits the upstream methane wall in full.

The Third Answer: Don’t Make Hydrogen

Now step back and look at what the market actually did with these sites. Teesside becomes an AI data center. Monolith is reportedly planning one on its own Nebraska plant. Everything a blue-hydrogen project assembles — the gas supply, the grid connection, the water, the permits — is exactly what a data center wants to buy, today, at scale, in cash.

That is the third answer for the carbon: don’t make the hydrogen at all. Burn the gas, sell the power. When the same inputs have two bidders and one signs today while the other asks you to wait for policy, the market has already voted. What outbid blue hydrogen wasn’t green or turquoise. It was the AI data center.

China: Same Physics, Different Problem

China’s hydrogen economy runs on coal: about 20.7 million tonnes of coal-to-hydrogen a year at a carbon intensity of ~19–20 kg CO₂ per kg — roughly 2.5× the gas route. The costs are public: coal ~RMB 12, gas ~RMB 15, electrolysis RMB 20–30 per kilogram, and adding CCS to coal-to-hydrogen costs about +RMB 7/kg at RMB 350–400 per tonne captured.

But the binding constraint is not price. National CCUS capture capacity is about 9.4 million tonnes a year; coal-to-hydrogen alone emits about 390 million. That is 2.4% — the tool and the problem differ by two orders of magnitude. For China, the blue question is not “is it cheap enough.” It is “is there anywhere to put it.”

The Verdict

Maturity 7/10 — century-old reforming, a decade of running storage; the highest score in this series. Economics 6/10 — the only clean hydrogen that comes within a dollar of the incumbent, and one subsidy covers most of it. Scale-up 3/10 — storage permits, leakage scrutiny, four dead projects in eight months. Moat 4/10 — geology and existing assets are real moats; the molecule is a commodity. Timing 4/10 — 45Q is at its peak with EOR at parity, and the buyer market just walked next door.

So, two stamps. As a new energy vector — the fuel that was going to displace other fuels — blue hydrogen is overhyped: the customer cannot tell the molecule apart, the fleet captures 60% of the carbon it advertises, and the market said no four times in eight months. As a retrofit — existing molecule, existing storage, existing customer, existing permits — it is a real bet: the cheapest cleanup of hydrogen we already make, with one subsidy carrying most of the gap. Turquoise is the honest physics with a 6% door: worth watching. And the force actually repricing all of it isn’t a better hydrogen at all. It’s the data center next door.

Blue hydrogen doesn’t lose to green. It loses to the contract nobody signs — and to the data center that signs today.