Climatetech startups need strong techno-economic analysis

Start with what the model is for, because that decides how to build it. Thiel’s framing is that a techno-economic analysis earns its keep from the moment an idea is still being mulled over, when it answers a crude question: could this compete at all? Refined, it becomes a roadmap. If the thing is not economic today, the model says what it would have to achieve to become economic, and which design decisions move affordability and which do not. For a company with a handful of engineers, that ranking is the value. The practical consequence: run the analysis backwards from the competitive threshold rather than forwards from what is achievable now. Thiel does this with synthetic fuels: priced at today’s clean hydrogen and captured CO2 costs the answer is absurd, so instead ask what the inputs must cost for the fuel to compete unsubsidized. Pushed for numbers, he offers hydrogen around a dollar per kilogram and CO2 at $100 to $200 per ton, with the CO2 necessarily atmospheric or biogenic or the fuel is not carbon neutral.

The first discipline is inputs, and its logic is that optimism has a budget. A novel technology already requires believing something unbuilt can be built, so every optimistic input compounds a bet you were already making; Kann’s rule is to isolate the magical thinking to the technology leap itself, the part within your control. Ball’s standing example is the company whose process runs on electricity and assumes very cheap power and full-time operation at once. Those are close to incompatible, because cheap green power means tethering yourself to a resource that runs perhaps 30% of the time for good solar or 50% for good wind, and the levelized cost of generating electricity is not the delivered price a customer pays. Her second is chemical: organic molecules look cheap to organic chemists, but yield and purification sit inside the boundary, and the practical floor is near ethylene, around a dollar per kilogram. Thiel’s is free waste heat, flagged against his own instincts as a thermal engineer: it is abundant and genuinely tempting, but wrapping a large heat exchanger around a long run of moderate-temperature flue gas can cost more than the recovered energy is worth. Ball’s fix is not to ban the optimistic number but to ask what would have to be true for it, then range it. If you are only in the money in the rosiest case, that is the finding.

The second discipline is where you draw the box, because cost lives in the system rather than the component. Total installed cost of a facility can be two, three, four times or more the core equipment, so being 50% better than state of the art on a component that is 20% of system cost barely moves the delivered number. Ball’s case is modular ammonia, where the clever low-pressure, low-temperature reactor is real but the hydrogen and nitrogen supplies upstream of it dominate cost and scale down badly; draw the boundary around the reactor and you miss the business. Kann adds that this worsens as a market matures, since the widely quoted sub-$100 per kilowatt-hour battery cell price is not a system price, and in utility-scale solar the module is a minority of project cost, the rest being labor, interconnection and permitting. It cuts the other way too: a couple of points of efficiency in an already 90%-efficient electric vehicle drivetrain component matter a lot, because every wasted kilowatt-hour has to be carried in the most expensive part of the vehicle. Components are not unimportant; only the system tells you which ones are. The same widening applies to the benchmark, which has to be the right object and a moving target. Ball’s version is comparing your levelized production cost against a market selling price, which is apples to oranges: the price includes delivery and a margin, and delivery is not always small, since her team found ammonia distribution reaching twice production cost. Kann’s addition is that in a commodity market the number to beat is not the incumbent’s current price but its floor, its ongoing operating cost, because an incumbent facing a cheaper entrant will cut toward that floor and undercut you anyway. And the floor moves: the late-2000s thin-film solar companies priced themselves against crystalline silicon as it then was, and silicon fell faster and further than anyone expected, so they arrived out of the money. Thiel applies the same test to new storage chemistries, where the question is not whether you beat lithium iron phosphate today but whether you beat it in 2035.

The last discipline is choosing the metric, and Thiel scopes that claim carefully: he is not arguing against research on efficiency or catalysis, only that from a venture perspective the improvement has to move the economics. If the model says input costs dominate, a better single-pass CO2-to-methanol catalyst is good science and a hard investment story, because hydrogen and CO2 are where the money goes. Kann extends it past cost to what the customer buys: a weeding robot optimized for capital cost may lose to one that covers the field faster, and 99% mineral recovery is worthless if the leaching kinetics are too slow for the mine’s downstream capacity. Which is why, Thiel argues, the work needs commercial and technical people together, since the economics half of techno-economics is a claim about customers. Then comes the counterweight, Kann’s own pet peeve, which cuts against everything above: seed-stage models arrive with four decimal places on numbers nobody can know. Ball’s worry is that counting pump power and valves is precisely how you miss the one driver that decides the company. Thiel’s is that early designs are still in flux, so detailed work on one subsystem gets thrown out when something learned elsewhere changes the design. His sequencing rule resolves the tension: error bars and sensitivities early, detailed rabbit holes only once the high-level design is fixed. The effort belongs in the width of the analysis rather than its depth.