Technical publication · Circular manufacturing and transparent business value
Research and decision question
At a publicly disclosed 50,000 t y−1 PET-waste nameplate, which combinations of feed PET fraction, availability, conversion, monomer recovery, residence time, solids loading, utilities and price assumptions determine whether an enzymatic PET-recycling concept clears a transparent economic gate? The second gate asks whether a governed reusable model merits its first-year implementation and adoption cost.
The boundary starts with PET-waste receipt at the modeled plant gate and ends with purified rPTA and rMEG in product storage. It includes internal recycle, wastewater, residues, consumables and onsite utilities. Collection before the gate, downstream repolymerization, product use and end-of-life are excluded unless a separate scenario expands the boundary.
This is a screening model for comparing explicit scenarios and locating evidence gaps. It cannot support final investment, financing, procurement, environmental, safety, permitting or plant-performance decisions without authorized feed, kinetic, equipment, utility, cost, product-quality, schedule and site evidence.
Wastewater, purge, non-PET residue, heat recovery, electricity, thermal utility, cooling and cleaning remain explicit side streams. A practical Acatian structure can use implemented generic operations—grinder, washer/mixer, heat exchanger, configurable reaction vessel, microfilter or rotary filter, crystallizer, filter dryer, distillation and hold tanks—without claiming a proprietary CARBIOS flowsheet or file compatibility.
Public evidence and data provenance
CARBIOS publicly describes feed shredding, melting and expanded granulation; enzymatic depolymerization in water; filtration and activated-carbon treatment; and separation of rPTA and rMEG. Its technology page states a planned 50,000 t y−1 PET-waste capacity [1]. A 24 September 2026 company release said project financing and due diligence remained underway, so this article makes no commissioning claim [2].
Tournier and colleagues reported at least 90% depolymerization over 10 h for their tested system at 200 g PET kg−1 suspension and 3 mg enzyme g−1 PET [3]. Those measurements are valid for that substrate, enzyme and experimental condition—not a transferable industrial guarantee.
Singh and colleagues modeled feed preparation, amorphization, hydrolysis and monomer recovery [4]. Murphy and colleagues evaluated a different fed-batch, pH-control and recovery route [5]. Their economics belong to their declared configurations and geographies; neither is transplanted into this example.
Nomenclature, balances and economic boundary
Use molecular weights 192.17, 166.13, 62.07 and 36.03 kg kmol−1 for the PET repeat unit, TPA, EG and two water molecules. Product mass legitimately includes incorporated water.
PET-01: MPET = MNAwPET.
PET-02: Mreacted = MPETX.
PET-03: MTPA = Mreacted(166.13/192.17)ηTPA.
PET-04: MEG = Mreacted(62.07/192.17)ηEG.
PET-05: MH2O = Mreacted(36.03/192.17).
PET-06: Vwork = (ṁPET/ws)τ/ρs; N = ceil[Vwork/(Vgrossfwork)].
PET-07: Q̇sensible = ṁslurrycp(Treaction−Tinlet)/3600. Add recovery, losses, reaction effects, agitation and cleaning separately.
CRF = i(1+i)n/[(1+i)n−1]. Annualized cost = FCI·CRF + Cfixed + Cvar. Gross margin = PTPAMTPA + PEGMEG + verified credits − feed cost − annualized cost. Keep TPA and EG allocation, currency year, location, ramp and financing explicit.
Reproducible synthetic worked example
Shared basis. The only company-specific number used is the publicly stated planned 50,000 t y−1 PET-waste nameplate. The synthetic screen uses 8,000 scheduled h y−1, 85% availability (6,800 running h y−1 for annual output), 90 wt% PET, 1,050 kg m−3 slurry, 20→65 °C, cp = 4.0 kJ kg−1 K−1, and 150 m³ gross vessels at 85% working fill. Instantaneous design flow is nameplate divided by 8,000 scheduled hours; availability reduces annual accepted mass, not that design flow.
The alternative produces 3,341.75 t y−1 more isolated monomers in this arithmetic screen, an 8.73% increase over Base, because its synthetic conversion and recoveries are higher. That is not a prediction. The same 642.86 m³ inventory occurs only because 50% longer residence time is exactly offset by 50% higher solids.
Vary PET fraction 0.75–0.95, availability 0.70–0.92, conversion 0.80–0.97, TPA and EG recovery independently, and solids jointly with residence time. Then test feed price or gate fee, enzyme, pH reagents, pretreatment power/heat, EG recovery energy, product quality/price, FCI, ramp, working capital, rate and project life.
No plant NPV, IRR or cost per kilogram is reported because public information does not establish site capital, enzyme price/lifetime, actual utilities, recovery duties, product contracts, ramp, financing or allocation. A transparent blank is more decision-useful than a false point estimate.
These are independent synthetic sensitivity ranges: Low and High mean the lower and upper numerical values, not combined favorable and adverse cases. They are not a ranked tornado result. Load authorized costs and product assumptions first; then report the threshold at which gross margin changes sign and rank drivers on the same currency-year basis.
Download the process, sensitivity, business-value and evidence ledger (CSV). Every row records its data class, source or assumption, formula ID, owner, evidence requirement and status.
Transparent software procurement case
This second model asks whether a recycler, engineering office, consultant or project developer should pay for and adopt a reusable process model. It does not value CARBIOS, predict a customer result or guarantee savings. The first-year cost boundary is subscription plus data preparation, model build, training/adoption, validation, external integration and maintenance. The live Acatian order page listed Professional at €590 per month on 3 October 2026; the synthetic first-year planning input is therefore €590 × 12 = €7,080 y−1. Reconfirm the commercial terms at checkout.
Bcapacity = Hdisplacedr freusea. This is engineering capacity, not cash. It has value only when a timestamped baseline shows the displaced task and the released capacity is productively reassigned.
Cyear1 = subscription + r(Hdata + Hbuild + Htraining + Hvalidation) + external integration + maintenance. Training is an adoption cost; data, build and validation are implementation costs.
Actual cash saving requires eliminated paid spend. Avoided cost is probability-weighted expected value, not booked cash. Throughput value requires accepted incremental product and contribution margin. Opportunity value is shown separately and excluded from hard ROI unless finance approves its conversion.
*Every value is synthetic. The High cash-saving input may be counted only after an invoice, purchase order or contractor spend is actually eliminated. Avoided-cost inputs require approved probabilities, consequence cost and attribution. Without Base's €5,000 avoided-cost hypothesis, Base is negative at −€4,992. Low needs 1,072 gross displaced hours—not attributed effective hours—to break even under its other assumptions. High is already above its synthetic cost only because conditional cash and avoided-cost inputs exceed that cost; remove them before relying on a zero-hour threshold.
Gross break-even displaced hours = max[0, (Cyear1 − other verified benefits)/(r freusea)]. Opportunity value remains outside this hard ROI. Synthetic examples are not customer outcomes or guarantees.
Attribution and realization evidence
Limitations and practical checklist
The reactor and heating screens omit mixing, rheology, fouling, heat recovery, losses, reaction heat, cleaning and hydraulic constraints. The public nameplate proves neither throughput nor economics. Literature performance is not an industrial guarantee. TEA changes with location, configuration, allocation, prices, financing and ramp. The result is neither a statement about an actual CARBIOS facility nor a recommendation to CARBIOS.
Technical FAQ
Define feed composition, pretreatment, enzyme and hydrolysis assumptions, separate TPA and EG recovery, utilities, residues, uptime, equipment, capital, operating cost, price basis, allocation, uncertainty and the evidence required for each input.
No. It is a publicly stated planned PET-waste capacity and is used here only to normalize a synthetic screening example. It is not evidence of achieved throughput, availability, yield or economics.
Hydrolysis incorporates water. One PET repeat unit consumes two water molecules to form one TPA and one EG molecule, so a correct mass balance includes the added water mass.
No. The company statement and the separate experimental or modeled studies use different substrates, enzyme systems, solids conditions, configurations and objectives. Each statement is valid only inside its own evidence boundary.
No. It is engineering-capacity value unless payroll or paid external spend actually falls. Productive reassignment, realization and attribution must be evidenced separately.
No. It is an independent screening blueprint based on public process architecture and synthetic assumptions. It contains no confidential plant data and is not a recommendation to CARBIOS.
Primary sources
CARBIOS. PET biorecycling technology. Current official process description and planned capacity; accessed 3 October 2026. Official company source.
CARBIOS. CARBIOS reports first-half 2026 results. 24 September 2026. Current public project-financing and due-diligence status. Official company release.
Tournier V et al. An engineered PET depolymerase to break down and recycle plastic bottles. Nature. 2020;580:216–219. doi:10.1038/s41586-020-2149-4.
Singh A et al. Techno-economic, life-cycle, and socioeconomic impact analysis of enzymatic recycling of poly(ethylene terephthalate). Joule. 2021;5:2479–2503. doi:10.1016/j.joule.2021.06.015.
Murphy NP et al. Process innovations to enable viable enzymatic poly(ethylene terephthalate) recycling. Nature Chemical Engineering. 2025;2:309–320. doi:10.1038/s44286-025-00212-y.