> ## Content Index > Fetch the complete content index at: https://www.plasticshift.ventures/llms.txt > Use this file to discover other available public pages before exploring further. # Why the world’s best plastic-eating enzyme still cannot get a bank loan - URL: https://www.plasticshift.ventures/writing/why-the-worlds-best-plastic-eating-enzyme-still-cannot-get-a-bank-loan/ - Published: 2026-09-18T10:22:50.000Z - Updated: 2026-09-18T10:23:21.000Z - Author: Mateusz Wielopolski A French company called Carbios has an enzyme that eats plastic bottles and hands back the raw chemicals to make new ones. It works. It was published in *Nature* six years ago. Its first factory still does not have the money to be built. That is not a contradiction. It is the most reliable pattern in advanced recycling, and most funding in the field gets it backwards: the chemistry is the finished part. The plant around it, and the buyer at the end of it, are not. ### **What actually happens inside the tank** PET — the plastic in every water bottle and most polyester clothing — is a long chain of two alternating chemicals: terephthalic acid and mono-ethylene glycol. Think of a paper chain made from two colours of loop, repeated thousands of times. Mechanical recycling washes the bottles, shreds them, melts the flakes and forms them into something new. The chain stays a chain, and each pass leaves it shorter and more discoloured. Hence the trouble with clear bottles, food contact and dyed material. Carbios goes further back. It uses an enzyme — a protein that cuts one specific chemical bond and ignores everything else — to snip the chain into its two original chemicals. Purify those, string them together again, and you have PET chemically identical to virgin. Colour, labels and mixed grades stop mattering. The enzyme is impressive — an engineered cutinase originally found in leaf compost, which in the 2020 *Nature* paper broke down 90% of the plastic in 9.3 hours. The plant at Longlaville, beside Indorama Ventures' PET site in France, is designed for about 50,000 tonnes of post-consumer PET a year at an estimated €230 million. ### **Why the enzyme needs its lunch served a particular way** An enzyme works on a surface, and needs a loose chain segment it can thread into its cutting site. PET is not uniformly loose. Some regions are tangled and disordered; others are folded into tightly packed stacks called crystals. The enzyme pulls at the tangled regions and does almost nothing with the crystalline ones — it unpicks a loose knit, not a pressed weave. Post-consumer PET arrives partly crystalline, and polyester fibre more so. So before any biology happens, the waste is extruded and micronised: melted and quenched to freeze it in its disordered state, then ground fine to raise surface area. Carbios ran that step at tonne scale for the *Nature* work. It gets worse. The enzyme works best near 72 °C, which is also where PET re-crystallises — in Carbios' own data the substrate climbed to 24.7% crystallinity within nine hours. The plastic hardens against the enzyme while the enzyme eats it. Then there is the water. The reaction runs as a slurry at roughly 200 g of plastic per kilogram of mixture: you heat and circulate about 800 kg of water for every 200 kg of plastic, hold it near 72 °C for eight hours, then pull terephthalic acid back out of a dilute solution and crystallise it. Such a plant is mostly grinding, heating and water handling, with a bioreactor in the middle. ### **The enzyme is the cheapest part of the whole thing** The US National Renewable Energy Laboratory modelled these economics and landed on a minimum selling price of $1.93/kg for the recycled terephthalic acid, built on clean flake at $0.66/kg. Its summary of what moves that number deserves framing: feedstock cost, plant size, solids loading and yield are the main drivers — enzyme cost and residence time are not. A decade of protein engineering made the catalyst so fast and cheap it fell out of the sensitivity analysis. The newest published enzyme, TurboPETase, hits 98.2% conversion in eight hours — on pretreated flake. Fed non-pretreated powder at 27.6% crystallinity it performs far worse. The improvements keep landing on the step already solved. The number a buyer sees has not moved. Purified terephthalic acid in Germany averaged about $1,114 per tonne in the second quarter of 2026 — $1.11/kg. Against $1.93/kg that is a gap of roughly three quarters, and no enzyme closes it. **PULL-QUOTE CANDIDATE** *“The enzyme got a thousand times better. The grinder, the heat exchanger and the crystalliser did not.”* ### **What a lender actually asks** Carbios has €42.5 million in confirmed public funding, banks whose credit committees have approved their share, and a target of first production in the first half of 2028\. Commissioning was originally promised for 2025\. On 3 August 2026 the company said it would not reach financial close by 30 September. It ended 2025 with roughly €60 million in cash and expected to spend about €20 million during 2026 before touching the plant. None of that is about biology. A lender facing a first-of-a-kind plant asks three questions. Can you buy feedstock at a known price and specification for the length of the loan? Will someone sign, in writing, to buy the output above the virgin price? Will a machine that has never existed reach nameplate capacity? Carbios answers the first two in part. Nobody answers the third for a plant that has not been built. ### **Carbios is not the exception** Widen the lens and the shape repeats. Fraunhofer's October 2025 map of European chemical recycling counted 65 projects in the pipeline and 2,799 kt/a of planned capacity against 18 plants actually running — and nine projects totalling 819 kt/a already cancelled. Look at why they died. Viridor closed its plants in Oslo, Skive and Malmö in June 2026, saying demand for recycled material had weakened while cheaper virgin kept undercutting it, with policy failing to provide the certainty needed to continue. Not one word about chemistry. Plastic Energy's UK companies entered administration in spring 2026 while the Spanish pyrolysis plants carried on. Mura's project at Böhlen collapsed because Dow shut the steam cracker meant to take its output — the technology never got a vote. The counter-example sharpens it. PureCycle missed schedules at Ironton for years, and the problems were mechanical: filters, restarts, reliability. It did not fail. It raised $432 million net in June 2026 and reported second-quarter revenue of $4.5 million, up about 173% year on year. The survivor is the company that got a plant to run, not the one with the most elegant chemistry. None of these are technology failures. They are project-finance failures, and the variable is identical every time: a buyer willing to pay above virgin, and a plant that reliably makes what the model promised. What I cannot resolve: Carbios has never published a cost-per-tonne breakdown for Longlaville, and the NREL model is a US plant at a different scale and feedstock price. The direction is well supported. The exact share of cost sitting in pretreatment rather than downstream separation is inference, and anyone who claims to know it is guessing. ### **Four questions worth asking before any investment happens** - **Ask what state the feedstock must arrive in, and who pays to get it there.** Crystallinity and chain length, not tonnage. If the pilot ran on pretreated flake, the data does not describe your waste stream. - **Ask for solids loading, not conversion rate.** Conversion above 95% is table stakes; anything well below 200 g/kg means you are funding a water-handling business. - **Price the offtake against virgin, not against a decarbonisation story.** Around $1.11/kg is the number in your buyer's head, whatever is in the deck. - **Put the feedstock specification in the contract, not just the volume.** These routes genuinely win on coloured, opaque, multilayer and polyester-textile streams that mechanical recycling cannot take. A tonnage commitment without a specification hands your counterparty the only variable that matters. We keep funding the part that looks like an invention. The part that decides whether anything gets built is a grinder, a heat exchanger and a credit committee — and none of those three has ever been on a magazine cover.