Catalyst N° 047 of 125 16 Jan 2025

FOAK tales

with Mario Fernandez, head of Breakthrough Energy Catalyst, Breakthrough Energy

In this note

01The question
02The answer
03The argument
04What you need to know first
05Details worth keeping
06Claims worth citing
07Where it’s contested

The question

What does it actually take to get a first-of-a-kind climate project financed, and why is that still the hardest step?

The answer

Because “proven” and “de-risked” are two different standards, and only the second one buys infrastructure capital. Closing that gap is mostly commercial and organizational work rather than technical: walk a specific scale-up path, run the steps in parallel because venture runway will not allow them to run in sequence, build your own plant instead of licensing the technology, write offtake contracts with flexible pricing and forgiving terms, and assemble capital at both the project and the parent-company level. There is no hurdle rate that makes a first-of-a-kind financeable on its own.

03The argument

Fernandez’s framing organizes everything else. These companies have technology that is proven, in the sense that a pilot generated data and the original scientific thesis held, but not de-risked in the eyes of large-scale infrastructure capital, meaning infrastructure funds and project finance banks. Those investors behave the way they do for a structural reason: there are multi-billion-dollar funds with 15-year track records that have never lost money on any project they invested in. That is the inverse of venture math, where nine failures are paid for by the tenth, and it produces a rigid checklist. A long track record for the technology. Hours of operation. A solid engineering, procurement and construction scheme. Long-term contracts. And enough scale to absorb hundreds of millions of dollars, because these funds are not looking to make a single $100 to $200 million investment and stop. New technologies fail that list on nearly every line, and no amount of scientific validation substitutes for it.

The path he proposes out of that is specific: pilot, then a demonstration five to twenty times larger, then a first-of-a-kind another ten to twenty times larger again. The demo’s job is not really scale, it is completeness. It has to run end-to-end, feedstock all the way through to the product a customer will actually buy, at spec, so customers can test the output. What he calls half-hearted demos prove one component and leave the integration unproven, and integration is precisely the part nobody has done before. He also wants six months or more of continuous operation, because things degrade when you run them continuously in ways a pilot never reveals, and those degradations are what you engineer around in the first-of-a-kind design. But that sequence collides with how these companies are funded, and the collision is the heart of the episode. A venture round buys roughly two years of runway, and you have to start raising six months or more before the money runs out, so it really buys twelve to fifteen months of progress. Waiting for six months of demo data before beginning to develop the first-of-a-kind is time nobody has. So the activities have to overlap: you develop the project while betting the demo will work, because the alternative is worse. Fernandez says companies that go strictly sequentially can take three to five years to reach their first-of-a-kind. The same logic drives a decision he says boards resist, which is to deploy your own technology rather than license it. Two years earlier he was hearing from chief executives whose boards wanted to avoid the capital spending and sell a license instead. Across 300 to 350 projects Catalyst has looked at, the companies that build their own plants are the ones that get there, while the licensing route typically burns two or three years and ends with nobody willing to take the risk on something never built.

The commercial construct is where he has watched companies destroy themselves. Demos should not carry long-term offtake at all; a demo is a money-losing research and development exercise, and no customer wants that risk anyway. Kann adds the incentive version of the same point: once a customer is waiting for product, you can no longer shut the demo down for a month to retool, and the demo’s whole purpose is learning. At the first-of-a-kind, the recurring mistake is pricing an offtake before engineering is done, on optimistic capital cost assumptions, then discovering during front-end engineering that capital cost is twice as high, with no way to go back and ask the customer to double the price. The remedy is to price and structure the offtake flexibly: off the final capital cost number, off an agreed margin, off an agreed project return, or as a collar with a base price plus shared upside indexed to the market. Kann pushes hardest right here, and the pushback is worth keeping. Absent a first-of-a-kind premium, which can show up as a green premium, or capital that is not entirely return-seeking, the first plant is inherently the most expensive one the company will ever build, and companies are typically already pricing at the customer’s maximum willingness to pay. He says he does not know whether they can actually win that flexibility. Fernandez concedes the limit and answers only partly: the demo is what lets you put parameters around the capital cost, so the flexibility band can be narrow rather than open-ended. He then insists that contract terms matter as much as price, because stringent volume obligations and delivery dates can put a company under if commissioning slips, and commissioning at a new scale is hard even in conventional technologies.

None of that produces a number, which is his answer on economics. The risk of a first-of-a-kind will mostly never be compensated by a return, so what makes an infrastructure investor comfortable is downside protection assembled across the entire contract set: does the feedstock contract run as long as the offtake, is the power purchase agreement durable, what are the offtake obligations, how good is the customer’s credit, who operates the plant. On top of that sits the platform motive. The investor puts a few hundred million into the first project because it wants to put a billion into the next two or three, often taking options on them, and it is the eventual sale of a platform at a multiple, as happened in wind and solar, that makes the arithmetic work. Project-level debt, meanwhile, he simply has not seen. Debt shows up at the corporate level as venture debt or corporate debt at low leverage, nothing like the 80% to 95% available on a solar project, and most money is raised at the parent company. His contention is that a correctly structured project should support 20% to 30% bank leverage, and he is candid that this is theory: in practice it has not been working yet. His larger conclusion is sharper than the financing mechanics. The asset class that would fund these projects, blending project-level positions with parent-company exposure through warrants or upside sharing, does not exist today, and while capital is an issue, the binding constraint is the skill set required to assess and de-risk first-of-a-kind construction and operation.

04What you need to know first

De-risked, as distinct from proven: Proven means the science worked and a pilot produced data. De-risked is the infrastructure investor’s standard: operating hours, a contractable construction scheme, long-term contracts, and a credible path to repeat projects. Almost every difficulty in this episode lives in the gap between the two.
Offtake: The long-term contract under which a customer commits to buy the plant’s output. Its price, its term and its delivery obligations are what make a project bankable or unfinanceable, and Fernandez argues the terms matter as much as the price.
Front-end engineering: The staged engineering process that turns a rough idea of what a plant costs into a number you can contract against. Fernandez puts the earliest stage at plus or minus 50%, which is why an offtake signed before that work is done is dangerous.
TopCo and ProjectCo: The parent company versus the separate entity that owns a single plant. Money can be raised at either level, and Fernandez’s argument for blended structures is that a first-of-a-kind creates enormous value at the parent, since without it the company dies, so an investor who only takes project-level risk is not being paid for what it is actually enabling.

05Details worth keeping

On modularity, Kann asks whether “numbering up,” building many copies of a proven unit instead of one large plant, is genuine risk reduction or just what infrastructure investors find palatable. Fernandez says it is real, and locates the benefit in execution risk: engineering design, procurement and logistics, construction, operations and the size of the capital raise. He qualifies it twice. Integration is still hard, so it is not a cure-all, and for some technologies, reactors being his example, bigger genuinely is better and descends the cost curve faster. Where the whole plant cannot be modular, modularize components. He notes a billion-dollar project carries risks a $100 to $200 million one does not, drawing on cogeneration projects he built around a turbine deployed thousands of times where plenty still went wrong.
Offtake term depends on the product. Catalyst helped structure what he describes as the first ten-year fully bankable offtake for electrofuel jet fuel, with American Airlines motivated by access to future product in a market it sees as supply-constrained. Cement is a spot commodity where getting anyone past five years is very hard; jet fuel is spot-traded too, but airlines can take a longer view. He thinks investors will give some credit for merchant exposure, so twenty years is not always required, but one- and two-year offtakes are not useful.
Selling the plant instead of the product looks attractive and has a trap in it. A first-of-a-kind customer will not bear construction risk, so the company builds on its own balance sheet and hands over only after an acceptance test. Kann points out there is no construction financing product for a first-of-a-kind the way there is for utility-scale solar, and procurement starts eighteen months ahead, so this opens a cash hole most of these companies cannot fill. A structure Fernandez has seen is a long-term offtake carrying a buyout option, so the company is still covered if the customer does not buy.
Capital cost misses have no consistent pattern, and he says so directly. He puts much of the blame on the post-pandemic engineering and construction market, where firms that used to offer lump-sum turnkey contracts no longer do so even for conventional technologies. On one project Catalyst committed to that was later canceled, the prices of four individual systems were well understood, but the balance of plant, the integration and the steel and cement it consumed, was not sized correctly because nobody had done it before.
The power contract is an underrated exposure. His example: a long-duration storage project whose economics depend on arbitraging negative-priced power might only be offered three years, because the utility knows a new transmission line is coming that will erase the opportunity. A three-year power contract under a much longer offtake is exactly the mismatch an infrastructure investor prices.
The Infinium example. Eighteen months earlier the company had an idea to convert a gas-to-liquids plant into an electrofuels facility, with no offtake, no feedstock, no power contract, no commercial construct and no construction plan. The work was a ten-year offtake in place of the one-to-three-year terms on offer, feedstock and power protections an infrastructure investor would require, a team structure, a construction plan with contingency, and capital cost properly sized for the brownfield conversion. Brookfield then committed over a billion dollars to the platform.
The Rondo example is the clearest case of commercial rather than technical innovation. The company had more than a dozen projects worldwide and no clarity on what it was selling: some customers wanted to buy the machine, some wanted it built and then handed over, some just wanted the steam. The answer was steam as a service. Rondo carries the capital cost of attaching its thermal battery to an industrial process and sells steam at a monthly price, bearing the risk of charging the machine. That converts a large day-one capital decision into a utility-style monthly bill, which matters because the buyer at a food, beverage or utility customer is often someone with twenty or thirty years at the company taking a personal risk on an unproven technology. Three first-of-a-kind projects were funded, with a joint venture with the European Commission and the European Investment Bank allowing a larger package that combined Catalyst funding with European Investment Bank venture debt.

06Claims worth citing

All figures as stated on 2025-01-16. These are patterns from one program’s deal flow rather than a survey, and the financing conditions described are the part most likely to have moved.

Infrastructure funds with 15-year track records that have never lost money on any project they invested in. The size figure is garbled in the transcript, reading as “70, 60, 70 billion,” so it is unclear whether he means funds of $60 to $70 billion each or that much capital in aggregate. The never-lost-money point is the load-bearing part and is stated clearly. Fernandez
Scale steps: pilot to demonstration five to twenty times, demonstration to first-of-a-kind another ten to twenty times. Kann offers the engineering rule of thumb of roughly one order of magnitude per step and Fernandez agrees, widening the range at the pilot-to-demo stage. Kann and Fernandez
Six months or more of end-to-end demonstration operating data is what the Series C and Series D investors funding first-of-a-kind projects want to see. Fernandez
Catalyst has looked at 300 to 350 projects in the past couple of years. What he draws from that base is qualitative, that the companies which deploy their own technology are the ones that succeed, rather than a measured success rate. Fernandez
Companies that run the steps sequentially can take three to five years to reach first-of-a-kind, against “a lot less time” when the steps overlap. The comparison is stated loosely and it is not clear whether three to five years is the total elapsed time or the added delay. Fernandez
A venture round buys about two years of runway, or twelve to fifteen months of real progress once you allow six months to raise the next one. Kann
The earliest engineering estimate stage is a plus or minus 50% capital cost number. Fernandez
Project-level debt on a first-of-a-kind: he has not seen it. Corporate-level venture or corporate debt exists at low leverage, far from the 80% to 95% available in solar. He contends 20% to 30% bank leverage should be achievable with the right contractual structure, and states plainly that in practice it has not been working yet. Fernandez
Catalyst structured what he describes as the first ten-year fully bankable offtake for electrofuel jet fuel, with American Airlines as the customer. Fernandez
Cement offtakes beyond five years are very hard to obtain; one- and two-year offtakes are not helpful in any product. Fernandez
Brookfield committed over a billion dollars to the Infinium platform. Fernandez
Rondo: more than a dozen projects in its pipeline, three first-of-a-kind projects funded, financed by Catalyst funding plus European Investment Bank venture debt through a joint venture with the European Commission. Fernandez
First-of-a-kind is the remaining valley of death, with the rest of the climate tech capital stack in reasonable shape despite the market shift. Kann

07Where it’s contested

Kann’s pushback on flexible offtake pricing is the sharpest exchange, and it does not get fully answered. He says outright that he does not know whether companies can win the pricing flexibility Fernandez recommends, given that they are already quoting the customer’s maximum willingness to pay. Fernandez concedes there is a limit to the green premium and responds with the demo as a way to narrow the capital cost band, which addresses the size of the problem rather than the negotiation.
The host frames this as a playbook and the guest keeps resisting that. Kann’s opening describes an industry standardizing processes and financing structures, and he asks repeatedly for rules of thumb. Fernandez declines the single number on returns, says the answer is a combination of factors, and ends by putting partnership above his own published list of twelve keys, saying that without partnership between companies, current investors and future investors it is very hard to get done. The material supports common patterns more than it supports a standard process.
Modularity is not universal and he says so. Integration remains hard, it is not a cure-all, and for some technologies bigger is genuinely better.
The financing conclusion is explicitly aspirational. The blended asset class he describes does not exist today in his own words, and his 20% to 30% leverage figure is a contention that has not yet worked in practice.
He gives a clean “no pattern” answer where one would be convenient. Asked whether companies consistently underestimate the same capital cost items, he says there is no consistency, and points at the engineering and construction market rather than at the companies.
The evidence base is one program’s deal flow. The lessons come from projects Catalyst reviewed, funded or, in at least one case, committed to and then saw canceled. Both speakers have a stake in these structures being seen to work, which the disclosure above records.