Paper that performs like plastic. Already selling.
Papkot is the bio‑mineral chemistry that finally makes paper replace plastic, on cost and performance, and it already sells at converters today. We are raising $35M at $116M pre‑money to scale it, and to build the agentic layer of papermaking: the layer that makes the chemistry impossible to remove.
Revenue first. The agentic layer on top. The reverse of every AI story.
↓The story starts with the reason paper kept losing.
Paper loses to plastic for one reason:
nobody solved the bond.
The spring‑hook: silica anchored to cellulose by silanol bonds (Si‑O‑C), elastomer springs between particles so the mineral layer flexes. Patent filed May 2026.
Every chemistry that gives paper its minimum viable properties, water resistance, grease barrier, strength, has to do two things at once: bond to the fibre, and cost next to nothing. Nothing did both, so the industry faked it. Two options only, plastic lamination or PFAS, both of them plastic wearing a paper costume. Papkot built the one with no plastic in it at all.
Papkot anchors bio‑mineral silica directly to the fibre through silanol chemistry, then locks it into something useful with a patent‑pending natural‑elastomer spring‑hook, so the layer flexes instead of cracking. Solve the bond once, and every product follows from the same mechanism: coatings, adhesives and glues, additives for papermaking.
Water‑based coatings from TRL 0 to TRL 8 in 14 months. PA™ from TRL 2 to TRL 8 in 6 months, accelerated by Papkot AI. One backbone, reused.
↓Nobody solved the bond. So we did. Here is who, and how the company is built.
The bond took seven years,
and a company built in reverse.
Every AI company is hunting for revenue; Papkot ran the playbook backwards and started with it. The chemistry already sells at converters and is climbing the paper value chain, tonne by tonne, on cost and on performance. The AI is built on top of that revenue to make the chemistry impossible to rip out. The revenue is already here. The AI is what makes it durable.
↓And once the bond held, it refused to stay one product.
Manuel Milliery
Founder, CEO & CTO · Shenzhen
Mathematics and AI, last as head of design for AI at Samsung. Papkot is his fifth company. He keeps starting them for the same reason: he cannot leave a broken thing alone, and plastic in food is a very broken thing. Seven years on, the silica finally bonds to the cellulose.
Solve the bond once,
and four families fall out of the same
mechanism.
Silica, MFC, fatty acids and a natural elastomer, recombined like building blocks across four product families and the entire paper value chain.
↓Where the chemistry runs decides whether it is a business.
Four families, one trap: off the machine, a demo.
On the
machine, a business.
There are two ways to coat paper. On‑machine means the chemistry goes on inside the paper machine itself, as the sheet is being made, about 7 m wide at roughly 1,200 m/min. Offline means a separate coating line afterward, about 1.5 m wide at roughly 200 m/min: a more than 25x area‑throughput penalty before capex, coating fees and handling. Offline proves people want the product. It can never make the product cost like plastic.
Solvent coating cannot run on a paper machine. The Series A breakthrough, 100% solvent to 100% water‑based, is what put the chemistry on‑machine at the size press, compatible with about 90% of the world’s coating equipment.
Paper cannot replace plastic unless it performs like plastic and costs like plastic.
Area throughput, indexed
A major Asian paper producer is moving a wrap machine to online Papkot coating today: one 4 m line is ~$42M/yr at capacity, versus $0.5–3M for the same coating offline. Worldwide, on the order of 100 specialty machines run wraps and baking paper (greaseproof, machine‑glazed and parchment, ~2–3 Mt/yr).
~$42M / machine × ~100 machines ≈ a multi‑billion‑dollar on‑machine coating pool. Capturing a small share of one segment is enough.
↓And the invoices did not wait for the machine.
Customers are already paying:
half of 2026 beat all of 2025.
Six months into 2026 customers have already paid us more cash than in all of 2025, on coatings alone. PA™ was a lab product in January. By June it is selling its first paid tonnes. The curve is bending, and the second engine has not even turned on.
Cash paid by customers: six months vs the whole prior year
Cash customers actually paid us, six months against a full prior year.
Half of 2026 already beat all of 2025, on coatings alone. The second engine, PA™, has its first mill past the on‑machine trial, savings confirmed, recurring volumes in negotiation. Two curves stacking, and the bigger one has just begun.
2026 coating deals
| Stage | Deals | Volume | Named anchors |
|---|---|---|---|
| Closed won | 1 | 200t | REWE bread bags, first commercial Vetrino™ reference in Europe |
| Negotiation | 9 | 741t | EU & APAC converters and a global specialty‑paper group |
| Proposal | 16 | 879t | Global packaging and tissue converters, EU and Asia |
| Qualified | 44 | 1,202t | Fedrigoni, plus EU and APAC converters |
| Scoped total | 70 | 3,022t | 2026 coating deal table, not PA™ maturity |
PA™ pipeline, speed to market
| Account | Reach | Scale surface |
|---|---|---|
| Flagship China mill | Paid China mill | On‑machine trial passed, savings confirmed on a few thousand tonnes, recurring volumes in negotiation. |
| Tier‑1 Asia paper group | Asia group | $3.0M one mill → $50M group |
| Major China paper group | China group | $2.5M one mill → $40M group |
| ANZ containerboard group | ANZ group | $4.0M one mill → $7.0M group |
| China packaging cluster | China cluster | $0.5M one site → $4.0M cluster |
| Global specialty‑fibre group | Plant network | $0.3M one plant → $2.5M group |
| EU independent mill | EU mill group | $0.1M one mill → $0.7M group |
The Asia group is post‑pilot and pre‑machine.
↓Coatings carry the revenue. Now watch the second engine.
That was engine one.
Engine two is validated, and pulled.
PA™ works inside the fibre, and ETS™ at the size press. Revenue is still small, but every trial has landed, and the demand came to us rather than the other way around: a leading wet‑end chemistry incumbent approached us on both.
PA™ molecular riveting: Si‑O‑C rivets and MFC gap‑filling reinforce the fibre network at the wet end. Source: Appita 2026.
27 industrial reels across 4 mills, 100% within spec. 300+ hours of mill trials, all successful. The chemistry is not fragile, it is proven on real machines.
PA™ proven on a commercial machine: on‑machine trial passed at 1,400 m/min, fibre savings confirmed on the first few thousand tonnes of production, recurring volumes in negotiation.
The recipe ran 50% long fibre, 50% short. Papkot replaced 100% of the short fibre, half the whole furnish, with powder milled from silicone‑coated glassine, the recycler’s nightmare. A negative‑cost waste stream became furnish. A production run, not a signed rollout, but the customer cut its furnish COGS by roughly 50%.
↓Two engines running. Now look at what they attack.
Two engines, four families:
four different lines on the
customer’s P&L.
They do not share one budget. Each competes against a different line on the customer’s P&L. The upstream pair plays the fibre game; the downstream pair displaces plastic and fossil systems.
| Family | Competes against | TAM 2028 | SOM 2028 | |
|---|---|---|---|---|
| PA™ (fibre) | the fibre bill itself | ~$415B P&L | $45–75M | Upstream |
| ETS™ (size press) | synthetic strength resins | $2.3–2.5B | $10–18M | Upstream |
| Coatings | extrusion‑lamination + PFAS spend | $11–13B | $25–45M | Downstream |
| Adhesives | plastic seal + barrier layers | $19–24B | $8–15M | Downstream |
| Consolidated | chemistry budget is only $23–26B; we sell a way to shrink the bigger fibre bill | $39–47B | $86–130M |
“PA™ is priced like an additive, but sized against the ~$415B fibre bill, the line 8 to 15x larger than the chemistry budget.”
↓Four markets, and one clock running on all of them.
No customer gets to wait:
qualification is slower than the
deadline.
A mill needs 12 to 18 months to qualify a new chemistry, and the rules forcing the switch, PPWR, the PFAS ban, rising EPR penalties, all land inside that window. That deadline pressure sits on the customer’s side of the table, and Papkot moves through qualification faster than anyone: the slow part is the data analysis, and Papkot AI makes it immediate, pulling the cycle under twelve months. The first customer’s results become the proof that closes the next. And this runs beyond on‑machine mills: the same pressure pulls the offline routes too, flexo, gluing, adhesives.
EPR spread, Belgium 2026: $5,126/t plastic‑coated paper vs $174/t recyclable.
Paper‑engineering graduates per year, 2000 to today. The experts mills lean on are retiring.
US states already restrict intentionally added PFAS in food packaging.
Certified, May 2024. Click to enlarge. Original shared on demand.
The Greenwashing Directive (EU 2024/825, from 27 Sep 2026) bans generic “recyclable” claims without substantiation. Papkot’s per‑reel data and CEPI v3 results are that substantiation, generated as a by‑product of normal production.
↓Regulation pulls the customer in. The data is what keeps them in.
Every qualified tonne leaves data behind,
and the data makes us
impossible to remove.
Papermaking got a data layer, decades of QCS sensors and dashboards. It never got a decision layer. The dashboard hands the problem back to the operator, who still picks the chemistry and turns the valve. And that operator is retiring.
Recipe archaeology, from a mill we sat with: 0.02% GPAM, 0.02% cPAM, 0.5% AKD and 1% ASA in one furnish, and nobody on site can say why all four are there. The only measurement is the customer complaint, a finished reel later.
Papkot AI is that missing layer. It reads the parameters, reasons over the full space from slurry to size press, and doses the chemistry, a human signing each move. We own the chemistry and the per‑reel dataset; the sensing and actuator surface is the partner half, and technical sessions with a leading control‑system vendor are underway. As that layer moves from advisory to dosing against a mill’s own per‑reel history, removing Papkot means giving up the only decision layer the mill has and handing the machine back to the operator who is leaving. The lock is that dependency, and it deepens with every reel.
Trust on a paper machine is staged, not assumed
Read the parameters. Manage the machine. Go autonomous.
The flywheel’s first turn, in one real mill: PA™ passed the on‑machine trial, the machine generates the per‑reel data, and that data is exactly what no incumbent can copy. The flagship China mill is where the loop has started.
↓A dependency that compounds. Here is its price, per mill.
Impossible to remove has a price:
we sell a bottom line, not a
product.
At commercial dosage (0.75 to 1.5%), PA™ costs the mill $9.5 to $35 per tonne of paper and returns $40 to $80 per tonne in fibre savings. Even at the bottom of the price band Papkot keeps a 50% gross margin, rising to about 73% at the top, before a single software dollar.
Pricing is set on value in use, not cost‑plus. Recurrence is mechanical: the chemistry is consumed on every tonne, so revenue scales with output, not with renewed sales effort.
Per tonne of paper, PA™ at commercial dosage
Blended gross margin trajectory
A single 300,000 t/yr mill cuts about $12M a year from its fibre bill with PA™, keeping roughly $8.5M after paying Papkot, about $4.8M of it durable net EBITDA. At a 6 to 8x paper‑industry multiple, that is $30 to 40M of enterprise value created for the customer, from one chemistry line. Once a mill has booked that number, we are the expensive thing to remove.
↓Value that size always draws an attack.
A bottom line that size draws attack:
six layers to crack at
once.
No single layer is the defense. A patent gets designed around, a trade secret gets poached, a certificate gets earned. The defense is that a competitor has to beat all six at the same time, and two of them, the dataset and the certifications, can’t be bought at any price. They only accrue while the chemistry is running in mills. Licences grant a right of use; ownership never transfers.
The spring‑hook bond
The silica‑to‑silica natural‑elastomer bond that defeats brittleness and carries function. Filed May 2026: 13 claims running from the composition to the finished article, a picket fence a competitor must clear at every rung, and the article claims expose even a converter buying infringing chemistry from a third party. A dedicated PA™ wet‑end filing is in drafting on the same chemistry.
The application formats
Cups, trays, baking paper, flow packs. The formats themselves are registered, on top of the trade‑secret chemistry that makes them.
Silica specification
Non‑intuitive particle characteristics commodity grades cannot replicate.
Formulation & process
Ratios, dosage windows, charge balance, drying profile, machine by machine.
Deployment intelligence
Papkot AI: machine‑specific optimisation, only learnable by running the chemistry at scale.
Regulatory certification
CEPI v3 at 98–99% fibre yield, SGS plastic‑free, food‑contact cleared on the commercial range.
↓And who built all six?
Built by eighteen people on $6.1M,
and built so it does not run
out.
A seed‑sized budget, spent like one. A $5M Series A from Fedrigoni and Anyma, $6.1M raised since inception. The entire commercial inflection came out of $1.9M of non‑salary spend, ~120 industrial trials, and Papkot AI, which together cut the invention‑to‑revenue cycle from 18 months to 6.
Manuel Milliery
Founder, CEO & CTO · 38
Owns technology, product direction and capital allocation. Controlling shareholder, based in Shenzhen.
Micaela Di Trana
President & COO · 53
20+ years at executive leadership level in listed and PE‑invested food multinationals, B2C and B2B, finance, category management, general management. Last six years in papermaking and converting (Fedrigoni Group) as VP Marketing, Innovation & Sustainability. Runs Papkot end to end: commercial, operations, finance, capital strategy.
70% of the 2023–26 venture cohort that died simply ran out of capital, a median of 22 months after the raise. Papkot already runs the system that prevents it: weekly cash tracking, a $12K authorisation gate, monthly re‑forecast, hiring released against evidence. That Series A record is the proof. The model: founder on technology and capital, President & COO on execution.
We hire for talent density, not headcount. The bar is high and turnover is deliberate: as the company grows, the team grows slower and denser, senior operators who have run real mills, not a payroll that scales with revenue.
Cap table after Series A
↓Discipline like that has a market price.
And that discipline is on sale:
more proven than the cohort, at
a third of the multiple.
| Company | Event | Revenue at round | EV / Revenue | Revenue / $ raised |
|---|---|---|---|---|
| Solenis | PE acquisition, 2023 | ~$3.8B (mature) | 1.4x | chemistry‑only floor |
| Cambium | Series B, Jan 2026 | pre‑revenue | n/a | <0.01x |
| Freeform | Series B, Feb 2026 | ~$7M ARR | 25.6x | 0.06x |
| CuspAI | pre‑Series B, 2025–26 | “millions”, pilots | >100x | <0.05x |
| Papkot | Series B, 2026 | ~$7.5M 2026E | 15–23x | 1.2x |
Papkot is the only company in the set with pure product revenue, shipped, invoiced and paid. Freeform’s ~$7M is ARR; CuspAI’s is unquantified pilots; Cambium’s commercial revenue is zero, all government R&D. Most proven revenue in the cohort, at the lowest entry multiple, and ~20x Freeform’s revenue per dollar raised.
We are raising to scale and compound value, rather than to sell. One signal of strength: six global category leaders, spanning paper chemistry, coatings, packaging inks and adhesives* came to us first, across supply, partnership and distribution. That optionality stays ours, to use later and on our terms.
Revenue / $ raised = revenue at round ÷ total capital raised to date. Papkot: ~$7.5M 2026E ÷ $6.1M raised ≈ 1.2x (comps shown at cumulative raise: Freeform ~$7M ARR ÷ $126M = 0.06x; CuspAI <0.05x; Cambium <0.01x). Revenue ranking at Series B eve: Papkot (pure product revenue) > Freeform (ARR) > CuspAI (unquantified pilots) > Cambium ($0 commercial, government R&D). Sources: Freeform (freeform.co, 2026); CuspAI (Sifted, 2025); Cambium (TechCrunch, 2023); Solenis shown as the mature chemistry‑only EV/revenue floor (PE acquisition, 2023).
↓Most proven revenue, lowest entry multiple. One thing left to buy: volume.
$35M to scale a business that already sells.
We are running this round on our own timeline.
~15–23x revenue, a third of the AI‑materials cohort at 25–100x (see the comps).
By 2028 (base, 50–60% conversion): $86–130M revenue, 55–60% gross margin, 8–12 PA™ and ETS™ mills plus coatings at scale.
$35M reaches cash‑flow breakeven by Q3 2027. No forced Series C: the round funds the company to self‑sustaining.
A focused data room from 18 August, signed term sheets by 15 September, close to follow. Lead allocation open.
We solved the bond and put it on the machine. What’s left is volume, not invention. Scale it with us.
Use of funds
Each line buys a milestone that re‑rates the next round: AI to operator‑approved closed loop, R&D to PA‑PFAS, deployment to signed mill contracts.
No manufacturing capex. ~200,000 t/yr toll capacity already operational.