The Wrong Scale-Up Risk

Disclosure: This is not financial advice. I own shares of $ADUR, and it represents ~30% of my portfolio, so I am biased. Long $ADUR | Not financial advice | DYOR.

Stocks covered: ADUR

In clean technology, the walk from a working pilot to a commercial facility is where capital goes to die. For a decade, equity investors have watched plastic pyrolysis companies present immaculate bench data, then hit reality at industrial scale: yields collapse, capex doubles, reactors coke up, and the equity gets recapitalized at a fraction of the last round.

That history is real, and the scar tissue is earned. It has produced a default assumption about Aduro Clean Technologies: scale-up is the existential risk, and the company deserves the same discount as everyone else who has stood at this particular threshold.

I think the market has the risk shape wrong. Not its existence. Its shape.

The specific failure mode that killed legacy pyrolysis at scale is structurally absent from Aduro’s process. What remains is conventional project risk: permitting, schedule, capital deployment, integration risk, but fundamentally different to underwrite. And on August 19, Aduro took the single largest step toward retiring the portion of that residual risk that actually sits on its own balance sheet.

Let me walk through why, and then through what would prove me wrong.

1. The physics: why thermal uniformity is the thing that breaks

Start with the mechanism of failure rather than the fact of it.

Pyrolysis reactors are oxygen-free; that’s the whole point of the process. Heat is applied externally, through the reactor wall or via a circulating heat carrier, into a molten polymer mass that happens to be an excellent thermal insulator. That is the problem. As you scale the vessel, the surface-area-to-volume ratio falls, and the distance heat must travel through a viscous, poorly conducting melt grows. You get thermal gradients. Material adjacent to the hot wall over-cracks and forms char and coke; material in the interior is under-converted. Product slate drifts, run length degrades, and the oil you get out is a variable soup requiring expensive downstream cleanup.

Mariusz Skonieczny, an independent investor and content creator who, like me, holds a position in the company, and whose work I’d treat as independent research rather than company disclosure, framed this well in a recent video, and I’m going to borrow his analogy because it’s the clearest version I’ve seen.

“Aduro’s plastic recycling technology is like cooking pasta in water... It will cook evenly in a small pot, a medium pot, a big pot, or many pots. Modularity means it will cook evenly.”
— Mariusz Skonieczny, MicroCap Explosions [3]

Baking a pizza depends on gradients. Boiling pasta doesn’t. A liquid medium at temperature transfers heat by convection and carries essentially no gradient; the reaction environment at the wall is the same as the environment at the centre.

Aduro’s HCT runs in liquid water, sub-supercritical, in the roughly 240–390°C band, with a proprietary catalyst and an in-situ hydrogen donor. The water is not incidental; it is the thermal medium, and it is the reason the reaction environment is uniform in a way a pyrolysis melt cannot be.

Here is the narrow claim I’ll defend: the dominant historical failure mode of scaled chemical recycling thermal non-uniformity and its downstream consequences does not apply to this process. The chemistry that works at 10 kg/hr is the chemistry that works at 1,000 kg/hr, because the variable that degraded with scale in the old architecture doesn’t degrade here.

Here is the claim I will not make: that scale-invariant chemistry gives you a scale-invariant plant. It does not. Mixing regime, residence time distribution, catalyst recovery and makeup, solids handling, corrosion behaviour under pressure at temperature, heat integration economics- none of that is settled by thermal uniformity. Those are engineering problems, and they remain.

Which is the entire point. Chemical risk and engineering risk are not the same asset class, and the market is pricing this name as though they were.

2. The Lego-block argument, and its one honest gap

A recurring error in capital markets is equating novel chemistry with novel engineering. Aduro’s chemistry is proprietary. Its plant is deliberately not.

Eric Appelman, Aduro’s CRO, has made this point directly:

“We have been able to assemble a process essentially from Lego blocks that have been around for so many years... We do not really need to build a 10,000-ton pilot plant because every builder knows perfectly well how you build those things bigger without running into trouble.”
— Eric Appelman, CRO of Aduro [1]

Sourcing note: this and the pilot-sizing rationale in Section 3 are management commentary from podcast interviews, not filings. I’m carrying them as such. They’re consistent with what the company has disclosed publicly, but they are not independently verifiable, and you should weight them accordingly.

The substance holds up. Extruders, pumps, heat exchangers, distillation columns, CHP units these are catalogue items with published scaling relationships, operating at ten to a hundred times Aduro’s FOAK capacity across the global petrochemical industry today. Nobody has to guess how a distillation column behaves at 10,000 tonnes a year. That collapses a large part of the FOAK premium: no single-source custom fabrication, no bespoke supply chain, no engineering guesswork on the balance of plant.

But the reactor is not a Lego block. It is the one genuinely novel piece of hardware in an otherwise conventional plant, and it is exactly where the residual risk from Section 1 lives. Anyone telling you the standardization argument covers the whole facility is overselling it.

So the real question isn’t whether Aduro can buy standard equipment. It’s whether anyone credible will take the novel unit and integrate it into a standard plant to a bankable cost estimate.

That’s the question that got answered this month.

3. The pilot size: pragmatism, not evasion

Before I get to Saipem, the bear objection that comes up most often.

The NGP pilot in London, Ontario runs at 10 kg/hr, roughly 240 kg/day. The FOAK at Chemelot is designed for about 1,000 kg/hr, or ~10,000 tonnes a year. Critics reasonably ask why the company didn’t build an intermediate 100 kg/hr unit to bridge the gap, and read its absence as avoidance.

Management’s answer, per interviews, is unglamorous: 240 kg/day was the largest reactor that would safely fit in the existing lab. Building bigger meant a new site, new permits, and a 12-to-18-month delay to generate data that standard scaling relationships already provide. They chose speed.

I find that credible precisely because of Section 2. If the balance of plant is catalogue equipment and the reactor chemistry is scale-invariant, an intermediate unit burns capital and calendar to produce very little novel engineering information. If either premise is wrong, the intermediate unit was the right call and skipping it was a mistake. The argument is coherent, but it is load-bearing; it depends entirely on the two claims above holding.

4. Saipem: the residual risk gets a counterparty

On August 19, 2026, Aduro announced the selection of Saipem S.p.A. for Early Works and Services supporting the FOAK facility at Chemelot. Early works are funded from existing cash, and the company expects no additional financing. The scope covers Process Design Package review, optimization of critical equipment packages, preliminary utility integration and capital cost refinement, under a structured stage-gated approach that opens into FEED, detailed engineering, procurement, construction and start-up.

Be precise about the instrument. Aduro’s own forward-looking language describes this as a Conditional Letter of Award. It is an entry point into a staged process, not an unconditional lump-sum EPC contract, and it does not commit Saipem to build anything. Treat it as what it is.

With that said, I think this is the most under-read item of Aduro’s year, and it lands directly on the gap I identified in Section 2.

Saipem is a tier-one industrial contractor. Its core competency is integrating process technology into executable, cost-bounded industrial facilities — including novel units, which is a substantial part of what they get paid for. When you hand a firm like that your Process Design Package and ask them to tear into your capital cost estimate, you are doing three specific things:

You are importing execution capability you do not possess. Aduro is a chemistry company with a pilot plant and roughly a C$10 million annual burn. It does not have in-house industrial project delivery. Saipem does. The engineering competence required for the build now sits with an organization that has done this repeatedly at far larger scale.

You are subjecting the capex number to adversarial review. This is the part I’d emphasize most. FOAK cost overruns are the single most common way clean-tech equity gets destroyed, and they usually originate in an internally generated estimate that never met a contractor’s pricing discipline. A capital cost refinement performed by the firm that would execute the work is a fundamentally different number from one built by a management team that wants the project approved.

What this does not do: it does not eliminate engineering risk. It transfers the residual to a counterparty with a track record of managing it, and it puts that residual under external scrutiny before committing capital. That’s the honest framing, and it’s still a material re-rating of the risk profile, but “de-risked” and “solved” are different words, and only one of them is defensible.

One open question worth flagging: Aduro has not disclosed whether Saipem is the same “leading global EPC firm” behind the March 2026 non-binding licensing-package MOU. The company has kept that counterparty unnamed. I’m not going to assume they’re the same entity, and neither should you.

5. What the NGP actually proved and what it didn’t

Two things get conflated constantly here, including by people who are long. They are separate claims from separate disclosures, and they deserve separate treatment.

The yield claim. On June 9, 2026, Aduro reported that the NGP pilot ran 47 continuous hours on recovered polypropylene at 86% liquid hydrocarbon yield, with roughly 35 hours at steady state. That’s the first public empirical validation of the yield claim under sustained continuous operation rather than batch conditions. It matters less as a number than as a category change: the machine turns on, stays on, and does what the deck said. Note the feed: this campaign was pure polypropylene. Mixed and contaminated feedstock campaigns are ongoing, and their results are not yet public.

The cracker-compatibility claim. This is a different disclosure entirely: the November 20, 2025 press release covering pilot-scale steam cracking trials at a European facility, conducted with a global organization that designs, licenses and services large-scale steam cracking operations. HCT oil was processed as produced, without dilution or pre-treatment, on a mixed waste plastic feed. Ethylene and propylene yields were reported as comparable to fossil naphtha, with a substantially lower boiling range than comparable chemical recycling oils.

That claim drives the economics, because pyrolysis oil is olefinic and cannot enter a cracker without hydrotreatment—a step that peer-reviewed work puts at roughly $400 to $1,200 per tonne, depending on geography. Removing it drops the minimum economic plant scale from the 100,000–200,000 tonne threshold pyrolysis requires to roughly 25,000 tonnes, which is what makes modular deployment work at all.

Carry Aduro’s own hedging with it. The company’s language is “potential” and “may be used,” and it refers to this particular Aduro product. One batch, one feed, one unnamed counterparty. It is a genuine validation milestone. It is not a finished commercial specification, and anyone citing it as one is going beyond what the company said.

CEO Ofer Vicus has been direct that the pilot-to-industrial jump is the largest physical step in the company’s life, and that the FOAK does not need to be the optimum on day one; the goal is steady-state continuous commercial operation, with refinement coming in units two and three. I think that’s the right posture, and it’s more honest than most of what this sector puts in front of retail.

6. What would prove me wrong

I’ve argued that the market is mispricing the shape of Aduro’s execution risk. That is not a claim that the risk is small. Here is what I’m actually watching, and what would break this thesis.

Permitting at Chemelot. Management has itself identified permitting as the most significant gating risk to the FOAK schedule. Nothing in this article addresses it. Standardized equipment and uniform thermodynamics do not move a Dutch environmental permit. The Ebert HERA engagement is the vehicle; visible permitting progress is the first concrete de-risking event, and its absence through year-end would be a genuine problem.

Schedule divergence. The company targets mechanical completion in 2H27 with commissioning in 2028. Sell-side models carry that later; Ladenburg’s runs to late fiscal 2028 with first revenue in fiscal 2029. Analysts are embedding a six-to-twelve-month slip multiplier, and they are usually right to. If you’re underwriting the company timeline without that haircut, you’re underwriting the optimistic case.

The instruments are still conditional. The Saipem award is a Conditional Letter of Award. The offtake is a non-binding LOI. The licensing arrangement is a non-binding MOU. The AstroTurf and Ortessa arrangements are MOUs at the evaluation stage. The thesis is conversion of that funnel into executed contracts, and none of it has converted yet. FEED authorization is a clean, falsifiable marker. If FEED hasn’t been authorized by mid-2027, the schedule has slipped materially regardless of what anyone says on a podcast.

The reactor could still surprise. I’ve argued thermal uniformity travels with scale. Catalyst performance over extended runs, fouling behaviour on contaminated feed, and materials durability at pressure and temperature remain open questions that only sustained operation can answer. The mixed-feedstock NGP campaigns are where this gets tested. Poor results there would directly damage the Section 1 argument.

And the standing bear case hasn’t changed: scale-up failure at FOAK, no commercial license signed by end of 2027, or repeated dilutive raises at falling prices. Any of those and I’m wrong. This remains a pre-revenue microcap whose valuation depends on belief about 2030. Position size accordingly.

The verdict

Scaling custom machinery running on chaotic thermodynamics is a monumental, often insurmountable risk. That risk killed the pyrolysis cohort, and the market is still pricing it into this name.

Scaling standard chemical engineering components around a controlled, water-based reaction with a tier-one contractor inside the data room reviewing the process package and the capex estimate is a different problem. It is a project execution and capital deployment problem. It can absolutely still go wrong. Permits get delayed, schedules slip, and contractors find the number was too low. But it is the kind of problem thousands of industrial facilities solve every year, and it shouldn't carry the same discount as an unsolved chemistry problem.

That distinction is the whole thesis. In March, you were underwriting chemistry. Today you are underwriting a project schedule. The Saipem mandate is what marks the transition, and I don’t think the market has repriced for it.

Sources

Tier 1 — Company disclosure and filings

• Aduro Clean Technologies, “Aduro Selects Saipem for Engineering and Procurement Support for First-of-a-Kind HCT Facility at Chemelot,” GlobeNewswire, August 19, 2026
• Aduro Clean Technologies, pilot-scale steam cracking of plastic-derived Hydrochemolytic oil, GlobeNewswire, November 20, 2025
• Aduro Clean Technologies, NGP pilot continuous operating campaign results, June 9, 2026
• Aduro Clean Technologies Corporate Presentation, March 2026

Tier 2 — Independent research and sell-side

• Roth Capital Partners, Initiating Coverage, Buy, July 27, 2026
• Ladenburg Thalmann (Jon Hickman), coverage updates 2026

Tier 3 — Interviews and community commentary (management commentary; hedge accordingly)

• [1] How to Commercialize Technology with Eric Appelman of Aduro, The Chemical Show:
• [2] Aduro CEO Ofer Vicus: The 86% Yield Run, Stock Therapy with Penny Queen:
• [3] Is Scaling Aduro’s Biggest Risk?, MicroCap Explosions (Mariusz Skonieczny independent investor, holds a position):
• [4] Why is Aduro’s pilot plant so small? Will their tech actually scale?, Amateur Investing

This is not financial advice. I own shares of $ADUR and it represents ~30% of my portfolio, so I am biased. Long $ADUR | Not financial advice | DYOR.