Loa Carbon · Hydrogen Supply · Research Brief
A decision-grade look at the Terraform/Rivan class of low-CAPEX electrolyzer — what's actually buyable at multi-MW scale, whether the low-efficiency-for-intermittent bet holds, and what it costs you downstream of the stack.
Bottom line
Your instinct is directionally right — but for a subtler reason than "low efficiency is better." At low capacity factor with cheap power, efficiency is a second-order lever: the extra electricity of a worse electrolyzer costs only ~$0.20–0.40/kg. That means the whole bet collapses onto three things the "cheap-and-crude" framing hides: real installed cost, guaranteed turndown under intermittency, and cycled-duty lifetime.
The catch: the genuinely cheap-and-low-efficiency units (Terraform, Rivan) aren't for sale — they're captive to fuel producers. The buyable cheap units are Chinese alkaline at ~$185–219/kW as a stack in China — but ~$800–1,200/kW installed, and they can't safely follow deep solar swings. The world's largest such plant (Sinopec Kuqa, 260 MW) has run at under ⅓ capacity for exactly this reason.
Efficiency barely matters here. That's the whole point — and the trap.
Levelized cost of hydrogen has two dominant terms. Capital-per-kg scales as 1 / capacity factor — so when the box runs part-time, capital dominates. The electricity term is capacity-factor-independent and scales with efficiency × power price. This is not opinion; it's the cost identity, and NREL's 2025 peer-reviewed grid-responsive TEA models exactly your question and concludes: as capacity factor and power price fall, the optimal move is higher current density — i.e. lower efficiency — because it lowers $/kW. HIGH
Key insight The bet isn't about efficiency
The headline "10× cheaper stack" ($150 vs $1,500/kW) is real at the stack level. But the efficiency penalty you'd accept for it is tiny in dollars, and the stack is only ~half the installed system. So the decision reduces to: is the cheap unit genuinely cheap once installed, can it actually load-follow your power, and will it survive the cycling? Optimize for verified installed $/kg-delivered-steadily over 20 years — not efficiency, and not stack price.
A narrow, well-defined corner — and Terraform's own numbers prove how narrow.
Terraform Industries published the cleanest worked version of your thesis ("How to Produce Green Hydrogen for $1/kg," 2023). Read the middle row carefully — at today's prices the low-efficiency machine is actually slightly more expensive. It only wins once power hits ~$10/MWh and CAPEX halves. The thesis is a bet on future ultra-cheap power, not a present-day win. HIGH
| System | Efficiency | CAPEX | Utilization | CAPEX $/kg | Power $/kg | Total $/kg |
|---|---|---|---|---|---|---|
| Legacy high-eff | 50 kWh/kg | $1,000/kW | 50% | $1.00 | $1.00 @$20/MWh | $2.00 |
| Terraform low-eff (today) | 80 kWh/kg | $100/kW | 25% | $0.64 | $1.60 @$20/MWh | $2.24 |
| Terraform target (~2028) | 80 kWh/kg | $50/kW | 25% | $0.32 | $0.80 @$10/MWh | $1.12 |
The break-even rule of thumb: the efficiency penalty (Δ kWh/kg × power price) must be smaller than the CAPEX-per-kg you save. From the cited worked cases, that flips in favor of cheap-and-inefficient roughly when power < ~$15–20/MWh and capacity factor stays low (~25%). Above ~$30–50/MWh, or at high capacity factor, the penalty isn't recovered and efficiency wins. MED — derived, not a single published threshold
Counter-metric Same $/kg ≠ same value
An independent critique concedes the cost-per-kg is roughly a tie — but a high-efficiency machine makes ~2–2.5× more hydrogen per installed MW. If your binding constraint is capital return, land, or grid connection rather than raw $/kg, the efficient machine wins even at low capacity factor. Your thesis optimizes $/kg; make sure that's the metric that actually binds you. MED
Both are captive to their own fuel plants. They validate the architecture; they don't sell you the box.
USA · captive
UK · captive
The cheapest CAPEX is precisely the hardest to buy — the true cheap-first players are all vertically integrated into fuels. Sources: Terraform blog, Plural, Carbon Herald, rivan.com.
The real sub-$300/kW market. But that number is a stack in China, not a system on your site.
Here's the honesty flag that reframes your whole question: these cheap Chinese units are not notably inefficient. They run ~4.3–4.5 kWh/Nm³ DC (~48–50 kWh/kg) — competitive with Western alkaline. They're cheap because of chemistry, not efficiency sacrifice: nickel-and-steel electrodes, no platinum-group catalysts, no Nafion membrane. In the same tender, these OEMs quote PEM at ~$630/kW vs alkaline ~$210/kW — a ~3× gap driven by materials. HIGH
| Vendor | CAPEX $/kW | Efficiency | Largest module | Intermittency / turndown | N. America path |
|---|---|---|---|---|---|
| PERIC (718th Inst.) | ~$219 | ≤4.3 kWh/Nm³ ~47.8 kWh/kg | 1,500 Nm³/h stack (world's largest) | ~20–50% min load MED | Projects only |
| LONGi Hydrogen | ~$185–220 | 4.0–4.3 kWh/Nm³ Hi1 Plus, DNV-verified | 1,500 Nm³/h bath | ~20–30%; Kuqa ~50% floor | Uzbekistan, EU JV; no US |
| Sungrow Hydrogen | ~$200–300 LOW | ALK 4.5 · PEM 4.15 | ALK 1,000 · PEM 300 Nm³/h | ALK 25–110%, 5%/s PEM 5–110%, 10%/s best-doc'd | Nascent |
| John Cockerill / CJH | ~$200–220 CN US: multiples | ~4.3–4.5 kWh/Nm³ | US Baytown: 30 MW pressurized | "flexible + stable loads" | Baytown TX — IRA/45V |
| SANY Hydrogen | ~$235 | ≤4.4 kWh/Nm³ | 2,000 Nm³/h (E-series) | Cold start ≤35 min, hot ≤5 min | Domestic |
| Envision | n/a (sells NH₃) | — | 500 MW live @ Chifeng | Off-grid, AI dispatch + BESS buffer | NH₃ export |
The 10× stack gap shrinks to ~1.5–2.5× once you buy a system on your site.
The stack is only ~50% of installed cost for alkaline (~60% PEM, ~30% SOEC). The rest — rectifier/power electronics (~30%), gas separation, purification, water treatment, compression, civils — is largely technology-agnostic; you pay it either way. So the cheap-stack bet only pays if the stack is a large share of your installed cost and BoP quotes are comparable across vendors. HIGH
Sources: China Hydrogen (tender floor), World Bank 2026 (installed + stack shares), BNEF via ETN (cost rise).
This is where the cheap-alkaline bet actually breaks — and it breaks in your exact use case.
Field evidence Sinopec Kuqa, 260 MW — running under ⅓ capacity factor since 2023
The world's largest green-hydrogen plant chose Chinese alkaline for solar-intermittent operation. Its units can't safely hold output below roughly half load — as current drops, H₂-in-O₂ crossover climbs toward the explosive limit (trip at 2 vol%; LEL 4 vol%). So rather than throttle smoothly with the sun, the plant sheds whole modules on and off, and its annual capacity factor has landed under ⅓. The damage is twofold: under-utilized capital, plus the cold-start / thermal cycling that on-off module operation inflicts on cheap stacks. A capacity factor that low triples effective $/kg on every cost line — swamping any stack-price saving. (Note: the ~50% floor is these specific units' safe minimum; datasheet turndown is nominally lower — see below.) HIGH
| Chemistry | Min load / turndown | Ramp | Cold start | Cost driver | Cycling degradation |
|---|---|---|---|---|---|
| Alkaline | ~20–40% (Kuqa ~50% real) | minutes | ~35 min–hours | Cheapest — Ni/steel | Sensitive; 20–50 µV/h drift, delamination HIGH |
| PEM | ~5% | seconds | ~few min | Higher — Ir/Pt, Ti | Driven by potential cycling MED |
| AEM (emerging) | ~10–100% claim | PEM-like | not characterized | Aims low (no Ir) | ±3 µV/h claimed LOW |
One nuance in alkaline's favor: because H₂-in-O₂ accumulation is a dynamic process, a stack can dip below its safe steady minimum briefly without breaching the limit — so cycled operation is safer than sustained low-load. But its structural ~20–40% floor still means it leans on a battery or H₂ buffer to chase spiky solar. Sources: Asia Times, Wood Mackenzie, Oikonomidis 2023.
Your methanation reactor wants steady, clean, pressurized H₂. Cheap alkaline gives you none of those for free.
Every impurity that poisons a Ni Sabatier catalyst (sulfur <0.2 ppm, chlorides, HF) comes from your CO₂ stream, not electrolytic H₂ — so that's not the electrolyzer's problem. The electrolyzer-specific issues are O₂ crossover and KOH aerosol from alkaline, plus low pressure, plus intermittency into an exothermic reactor that wants steady flow. HIGH
| Dimension | Sabatier wants | Alkaline delivers | PEM delivers | Implication of going cheap |
|---|---|---|---|---|
| Purity | O₂-free, dry, S<0.2 ppm | 99.5–99.9% + O₂, KOH mist | >99.99% dry | Add demister + deoxo + dryer (±PSA) |
| Pressure | ~15–30 bar | ~1–10 bar | 30–50 bar | Add a compressor + its parasitic load |
| Steadiness | steady, thermally stable | 20–40% min, min ramp | ~5% turndown | Add an H₂ buffer / flexible reactor |
Every flagship e-methane plant proved intermittent alkaline works — but only with those add-backs:
Sources: IEA Bioenergy Task 44 (Werlte), McPhy (Jupiter 1000), MDPI Energies 18(11):2886 (alkaline purification trains), PEM vs alkaline overview.
Three real paths, depending on what binds you. None of them is "chase the cheapest stack."
If N. America / IRA-45V matters
If tight container + reactor stability
If cheap intermittent power + space to buffer
Watch, don't buy
The one-line reframe
The question isn't "cheap low-efficiency vs. expensive high-efficiency." It's "lowest $/kg delivered steadily over 20 years, at my real capacity factor, including the buffering my Sabatier reactor forces on me." On that metric, the cheap-alkaline bet wins only in the near-free-power, low-utilization, well-buffered corner — and loses badly exactly where it's marketed hardest: chasing intermittent solar at high nameplate scale.
For completeness — the market splits into genuinely-cheap vs. efficient-but-not-cheap. Hysata is the sharpest counter-example to your thesis.
| Company | Chemistry | $/kW | Efficiency | Buyable? | Position |
|---|---|---|---|---|---|
| Ohmium | PEM, modular | ~$775–1,000 stack | n/p | Yes — stacks | Mid-cost, dynamic, dispatchable |
| Electric Hydrogen | PEM | premium | 54 kWh/kg, 30 bar | 100 MW plants | Efficient, high-output; not cheap |
| Verdagy | large-area AWE | n/p | LCOH<$2/kg '28 | Module launched | Real-time load-matching |
| Hysata ⚠︎ | capillary-fed | not cheap | 41.5 kWh/kg ~95% system | Scaling | Ultra-efficient — opposite of your thesis |
| Supercritical | membraneless | <£1/kg target | 42 kWh/kg >220 bar out | Pre-commercial | Efficient + saves compression |
| Cipher Neutron | AEM | low (no Ir) | 41.75 kWh/kg | Emerging | Cheap materials + efficient hybrid |
| Advanced Ionics | symbiotic | <$1/kg target | waste-steam | Pre-commercial | Genuinely cheap; needs waste heat |
| TK Nucera | AWE (20 MW) | ~$800–1,000 | n/p | Yes — major OEM | Established, high-reliability |
Sources: Ohmium, EH2, Hysata, Supercritical, Cipher Neutron, Advanced Ionics, TK Nucera.
Every load-bearing number traces to one of these. Confidence tags flag where to lean and where to verify before citing externally.
Caveats to carry. The "$150/kW cheap stack vs $1,500–2,000/kW premium" framing is a stack-level comparison; verified installed-system costs are $800–2,500/kW. Efficiency is shown in kWh/kg (thermodynamic minimum ≈ 39.4 HHV / 33.3 LHV); vendor figures are DC-at-stack unless noted, so add ~10–20% for AC-at-plant. The installed-system figures already include generic gas purification — so the deoxo/dryer/compressor in §07 is the incremental alkaline-vs-PEM delta, not a wholly separate add-on. Efficiency is second-order only below the crossover (low CF + cheap power); above it, efficiency dominates LCOH again. The ~$15–20/MWh break-even and the "~10% efficiency overstatement / 1 mm separator" points rest on single or partly-paywalled sources — treat as MED, verify before external citation. Terraform's numbers are self-published and self-consistent but not independently audited. No Analog Devices electrolyzer spinout exists in public record — reported as not-found, not invented.
Prepared for Dan Wojno · Loa Carbon · research brief, not a procurement recommendation. Get vendor quotes on fully-installed $/kg and contractual turndown/warranty terms before committing capital.