The only industrial-scale gas fermentation plant operating anywhere makes animal feed from methane in Chongqing, China, and is rated at 20,000 tonnes a year. The food-grade side of the same technology category runs at 160 tonnes a year, in Vantaa, Finland, and generated EUR 0.1 million of revenue in the first half of 2026.
What separates them is not the underlying biology, which is microbial biomass fermentation on a gaseous carbon source in both cases. It is the feedstock, the end market, and the regulatory instrument required to sell the output — and the three move together. Hold the technology constant, vary the destination, and the capacity gap is roughly two orders of magnitude.
The assembled picture
Every figure below is design or nameplate capacity as stated by the operator. None of these companies publishes actual output.
| Company | Plant and location | Feedstock | Design capacity (t/yr) | Actual output | End market | Regulatory status | Source |
|---|---|---|---|---|---|---|---|
| Solar Foods | Factory 01, Vantaa, Finland | CO₂ + hydrogen (on-site electrolyser, some CO₂ from on-site DAC) | 160, rising to 230 during 2026 | Not disclosed; H1 2026 revenue EUR 0.1m, order book EUR 0.2m | Food (Health & Performance Nutrition) | Singapore novel food Sept 2022; US self-affirmed GRAS Sept 2024, FDA notification filed Sept 2025; EU and UK pending | Solar Foods, 20 Oct 2025 and H1 2026 report |
| Solar Foods | Factory 02, Lappeenranta, Finland | CO₂ + hydrogen | 3,200 phase 1, 6,400 phase 2, 12,800 phase 3 | Pre-construction; FID expected during 2026 | Food | As above; nothing built or permitted yet | Solar Foods, 16 Dec 2025 |
| Calysseo (Calysta / Adisseo JV) | Changshou, Chongqing, China | Methane from natural gas, plus oxygen and ammonia | 20,000 (phase 1) | Not disclosed; CEO states ~70% of nameplate is going to pet food | Feed — aquafeed, then pet food | FeedKind approved for sale in the EU, UK, Canada and China; US dog food anticipated for 2026 | Calysta, Jan 2021; AgFunderNews, Nov 2025 |
| Unibio / SIIG JV | Al Jubail, Saudi Arabia | Dry gas allocation from the Saudi Ministry of Energy | 50,000 announced, ambition stated at 300,000 | Not built; construction stated to start H2 2026 | Feed — aquaculture, animal feed, pet food | Uniprotein approved for aquaculture in Saudi Arabia, animal feed in the EU | Green Queen, 17 Mar 2026 |
Solar Foods: what “reached its productivity targets” actually means
On 20 October 2025 Solar Foods announced that Factory 01 had reached production parameters enabling it to run at its full design capacity of 160 tonnes of Solein a year. The release is specific about what was measured: productivity of 1 g/l/h and an energy efficiency value (O₂/CO₂) of 2.7, achieved after a summer maintenance break. Before that break the plant had run continuously for eight months. It started operations in April 2024. The heart of it is a single 20,000-litre bioreactor.
Note what the announcement does not say. It does not say the plant produced 160 tonnes. It says the plant reached the parameters that would let it. Design capacity and output are different numbers, and the release only gives the first.
The half-year report published on 11 August 2026 gives the closest available proxy for the second. Revenue for January to June 2026 was EUR 0.1 million, up 271% on a near-zero base. The order book stood at EUR 0.2 million, down from EUR 1.5 million a year earlier — an 87% fall. Operating loss was EUR 6.8 million.
Solar Foods has not published a Solein price, so what follows is a sensitivity rather than a calculation: at EUR 30/kg a EUR 0.2 million order book is under seven tonnes, at EUR 15/kg it is thirteen. On any plausible price the constraint in 2026 is demand, not fermenter volume. The CEO commentary agrees — stated second-half priorities include “selling production capacity of Factory 01” and “securing binding offtake agreements with large international CPG companies.” You do not list selling existing capacity as a strategic priority if it is sold.
Practical consequence: when a gas fermentation company announces it has hit full design capacity, the announcement is about the bioreactor, not the P&L. Ask separately for tonnes shipped and for the order book. Solar Foods, being listed on Nasdaq First North, publishes the second. Most of this category does not publish either.
Factory 02 is where the ambition lives
Factory 01 was never meant to be the business. Solar Foods’ own Factory 01 page describes its impact as “small” and calls the site a pilot and blueprint. The December 2025 site announcement sets out the real plan: Selkäharju in Lappeenranta, built in phases — 3.2 kilotonnes a year in phase 1, targeted operational at the end of 2028; 6.4 kilotonnes in phase 2 the following year; a possible third phase at 12.8 kilotonnes. The company states these volumes are based on production parameters already achieved, which is the load-bearing claim behind the whole scale-up.
On demand, Solar Foods signed MoUs and letters of intent with four international customers during 2025 covering 6,500 to 7,650 tonnes a year — roughly 100–120% of the full 6.4-kilotonne capacity, and not binding. On financing, Business Finland granted EUR 77.8 million in June 2026, split as a EUR 39.6 million grant and a EUR 38.1 million R&D loan under the company’s approved IPCEI notification, with approximately EUR 21 million still undecided. All of it is conditional on the FID, which is “intended to be made during 2026.”
So the food-grade side of this category currently has 160 tonnes of built capacity, an unbuilt 3,200-tonne phase 1 that is pre-FID, and a letter-of-intent book that only becomes meaningful if both convert.
Calysta: methane, aquafeed, and a quiet pivot to dogs
Calysseo — the joint venture between Calysta and Adisseo, which Calysta describes as a main subsidiary of China National BlueStar and which Shaw says is majority-owned by the state-owned enterprise Sinochem — switched on its Chongqing fermenter in October 2022, in the Changshou National Economic and Technological Development Area. At groundbreaking in January 2021 it was described as delivering 20,000 tonnes of FeedKind protein in its first phase, with a second phase to follow.
The process detail matters for anyone modelling this. Calysta grows a natural, non-genetically-modified strain of Methylococcus capsulatus isolated from hot springs in Bath, England, on methane from natural gas, with oxygen and ammonia. Fermentation runs above 40°C in 400,000-litre horizontal U-loop reactors. CTO Geoff Bryant told AgFunderNews the company runs at “nearly four” kilograms of biomass per cubic metre per hour, produces “over 1,000 kg of protein on a dry basis every hour” in those fermenters, and runs campaigns of 12 to 15 weeks. The China plant has two fermenters, each rated at 10,000 tonnes a year.
That is the only place in this category where enough numbers exist to check a nameplate. Four kilograms per cubic metre per hour across 400 m³ is 1,600 kg of biomass per hour, or roughly 14,000 tonnes per fermenter run continuously for a year — comfortably above the 10,000-tonne rating. The implied uptime assumption is around 70%, which for 12–15 week campaigns with periodic mechanical stops is a plausible design margin rather than a contradiction. It is also the sort of margin that never appears in a press release.
Two commercial facts published in November 2025 change how the plant should be read. First, pricing: Calysta CEO Alan Shaw stated that in aquaculture, “you’re not going to get better than $2,000 a ton for our product,” whereas in pet food “the price is almost double that.” Second, allocation: “for a 20,000 ton per annum nameplate capacity plant, we’re forecasting that 70% of the capacity will actually be directed into pet food.” Aquaculture, Shaw said, “was always going to be the low hanging fruit” — competitive fishmeal prices pushed Calysta towards higher-value markets.
At the same time Calysta closed its R&D labs and pilot plants in San Mateo, California and Redcar, Teesside, auctioning the UK equipment through BPI Asset Advisory in November 2025. Shaw’s framing: “We’re not a startup anymore; we’ve evolved into a manufacturing company.”
Practical consequence: the $2,000/tonne aquaculture ceiling is one of the very few disclosed price points in gas fermentation, and it sets the bar any CO₂-to-food player must clear if it wants a feed option. If your process cannot land near $2/kg for a 68%-protein feed ingredient, stop describing feed as a fallback market. Note also where the number came from — a CEO interview, not a filing, the same disclosure failure catalogued in the precision fermentation cost data gap.
The basis problem: 160 tonnes of what, 20,000 tonnes of what
The two headline tonnages are not measured on the same thing, and neither release says so.
Solar Foods publishes Solein’s composition as 80% protein, 6% fat, 10% dietary fibre, 4% minerals, explicitly as average values from dry weight. Calysta’s current FeedKind site states 68% crude protein and 9% crude fat. The older FeedKind Pet data sheet gives 71% at 5% moisture, and CTO Bryant described the organism as “about 70% protein.” Those three figures do not reconcile from public sources; 68% appears on the page most recently updated (March 2025) and is used below.
There is a second ambiguity in the phrase itself. “20,000 tonnes of FeedKind protein” is a brand name, not a protein-content statement. Trade coverage routinely compresses this to “20,000 tonnes of protein,” which is not the same claim.
| Basis | Solar Foods, Factory 01 | Calysseo, Chongqing | Ratio |
|---|---|---|---|
| Product tonnes as stated | 160 | 20,000 | 125× |
| Protein tonnes (80% and 68% respectively) | 128 | 13,600 | 106× |
| Protein tonnes at Solar Foods’ 2026 target of 230 t | 184 | 13,600 | 74× |
The basis correction is real and moves the number by about 15%. It does not rescue the comparison: the gap is two orders of magnitude however you measure it, and still one order of magnitude if Factory 02 phase 2 is built exactly as announced. Anyone quoting a single-figure ratio between food-grade and feed-grade gas fermentation without stating the protein basis is quoting a number they have not checked — the failure mode documented across the fermentation capacity map.
Practical consequence: demand three things from every capacity claim in this category — tonnes, protein and moisture basis, and end market. Two out of three compares nothing.
NovoNutrients: the CO₂-to-food route that ran out of runway
NovoNutrients is the control case for what happens when the food-grade CO₂ route meets the capital markets. Founded in 2017 by Brian Sefton and Russell Howard, the California company captured industrial CO₂ and combined it with hydrogen to make a protein ingredient branded Novotein for aquaculture, pet food and human nutrition. In July 2024 it closed an $18 million Series A led by Woodside Energy and co-led by CM Venture Capital. On 18 July 2025 it entered the Assignment for the Benefit of Creditors process — a California alternative to bankruptcy — and put its assets up for sale through Armanino Advisory, bids due 31 July.
Former CEO David Tze’s post-mortem is precise about the cause: “The technology works. We were proving that in our pilot phase. The challenge was capital intensity in a shifting investment climate.” The asset listing described issued US patents, non-GMO strains, bioreactor and scale-up trade secrets, and technology achieving 73% protein content plus carotenoids from waste CO₂. Total money raised is disputed: Green Queen put all-time investment at $27 million in July 2024 citing TechCrunch, and at $22 million in its July 2026 acquisition report. Both agree on the Series A; neither total is confirmable from a filing, so we treat neither as settled.
Biosphere, a Californian bioreactor company that emerged from stealth a few months before the ABC filing with a UV sterilisation platform, acquired NovoNutrients’ intellectual property, microbial strains and process expertise in mid-2026; terms were not disclosed. Biosphere had secured a $9 million grant from the US Army DEVCOM Soldier Center in May 2026, running three and a half years, for field-deployable bioreactors producing gas proteins for the Army. CEO Brian Heligman said the NovoNutrients platform “plays perfectly into our active Department of War contracts and gives us a strong foundation with an established FEL3 design of a loop reactor pilot plant.”
Read that sequence carefully. A CO₂-to-protein-for-humans company failed on capital intensity, and its reactor IP was bought by a company whose funded application is military logistics. That is not a food business absorbing a food business — it is the pattern seen across the cultivated meat supply layer: the hardware survives the thesis.
Practical consequence: if you are diligencing a gas fermentation asset sale, the question is not whether the technology works. Tze conceded it did. The question is which balance sheet is prepared to carry a first commercial plant, and whether that balance sheet has any reason to want food as the output.
The case against reading this as a failure of CO₂-to-food
Three counter-arguments deserve to be stated properly rather than waved at.
Feed is a deliberate beachhead, not a consolation prize. Calysseo was designed from the start for the Asian aquafeed market, described at groundbreaking as more than 70% of the world market. Feed clears regulatory approval faster than food, tolerates a crude-biomass downstream process — Bryant describes centrifuging to 12–15% solids and spray drying, with no purification step — and China’s food security policy supplied a state-linked partner willing to fund an industrial first-of-a-kind. That is a capital structure story as much as a technology story, and it is repeatable.
Methane is thermodynamically the easier feedstock, and Calysta says so directly. Methane is both the carbon source and the energy source. CO₂ supplies only carbon, so you must feed hydrogen as well — and once hydrogen, oxygen and CO₂ share a vessel you have a safety envelope to engineer around. Methane fermentation also runs above 40°C on a substrate that invading organisms cannot eat, which Bryant calls “almost sterile by design” and which is what makes 12–15 week campaigns possible. Long campaigns are the difference between unit economics that work and unit economics that do not. Bryant is equally clear this is not a green free lunch: Calysta uses natural gas today and wants synthetic methane from CO₂ and green hydrogen eventually, but says the cost and availability are not there. The CO₂-sourcing problem is shared — “the cost to clean up waste CO2 is very expensive.”
160 tonnes is a first-of-a-kind, and design capacity is not the ambition. Solar Foods scaled roughly a hundredfold from its pilot to Factory 01 and says it hit full design capacity earlier than planned. Its pilot already runs at 1.6 g/l/h against Factory 01’s 1.0, which is the basis for the 230-tonne uprate during 2026. Judging the category on Factory 01’s tonnage is like judging a chemical process on its demonstration unit. The fair test is whether phase 1 of Factory 02 gets a positive FID and gets built — and that test resolves inside the next eighteen months.
What we could not establish
- Actual tonnes of Solein produced at Factory 01 in any period. Solar Foods publishes design capacity, productivity in g/l/h and revenue, but no output figure.
- Actual output or utilisation at Calysseo Chongqing. 20,000 tonnes is nameplate. Neither Calysta nor Adisseo publishes production.
- Whether the Calysseo second phase is still live. The January 2021 Calysta release says only that “a second phase will bring extra tonnes of capacity.” Contemporaneous trade coverage reported 80,000 tonnes; we could not confirm that from a company source and have not used it. Calysta’s November 2025 comments point to Saudi Arabia, the UAE and possibly North America, not to a Chongqing expansion.
- The direction of Solar Foods’ O₂/CO₂ energy efficiency metric. Factory 01 reports 2.7 and the pilot 2.6, and the company describes bringing pilot parameters to Factory 01 — implying lower is better — but the release does not define the metric.
- Solein’s selling price per kilogram. Not published, which is why the revenue-to-tonnage inference above is a sensitivity rather than a calculation.
- A GRN number for Solein in the FDA GRAS inventory. The company confirms it filed on 16 September 2025; we could not retrieve the FDA inventory record to confirm a notice number.
- NovoNutrients’ pilot working volume in litres, the price Biosphere paid, and the company’s total capital raised (sources give $22 million and $27 million).
- The Unibio Al Jubail timeline. The single source available states construction begins in H2 2026 and completes by mid-2027, and separately that the partners aim to start commercial production in H1 2026. Those cannot both be right; we report the capacity, not a date.
What to watch
- Solar Foods final investment decision on Factory 02, by 31 December 2026. The FID is stated as intended during 2026 and the Business Finland money is conditional on it. A slip into 2027 pushes phase 1 past the stated end-2028 start.
- Factory 01’s uprate from 160 to 230 tonnes during 2026, committed to in the October 2025 release. The FY2026 results are where it either lands or does not.
- EU novel food approval for Solein during 2026 (applied 29 October 2021; EFSA questions answered February 2025 using Factory 01 data), and an FDA no-questions letter by end-2026 on the September 2025 GRAS notification. Both are the company’s own guidance.
- Whether Solar Foods’ order book recovers above EUR 1.5 million. It fell 87% year on year to EUR 0.2 million at 30 June 2026. Two consecutive halves at that level would say more about the food-grade market than any capacity number.
- US approval for FeedKind in dog food in 2026. Shaw said in November 2025 the company anticipated approval “next year.”
- Calysta’s next raise, “several hundred million dollars” for a plant of tens of thousands of tonnes, flagged by Bryant in January 2025 as in process. Where that plant is sited — Gulf or North America — tells you which feedstock economics the market believes.
- Construction start at Al Jubail in H2 2026. A 50,000-tonne methane-fed plant widens the feed-versus-food ratio, it does not close it.