Catalyst Live at SF Climate Week

The opening segment sets up everything that follows. Kann’s premise, which is his own and offered without data, is that consumer companies account for a disproportionate share of climate tech’s rare generational outcomes. Schroepfer answers that a consumer product has to clear two separate bars, not one. It has to be a product people love, and the problem it solves has to be big enough to carry a business. He offers his own counterexample rather than a hypothetical: Meta’s Portal video device cleared the first bar and missed the second, because people who bought it adored it and a dedicated video-chat device was never a large enough market. Chaset’s version is grounded in electricity specifically. Energy is invisible and nobody interacts with it, so the products that have worked either delivered something tangible, an induction cooktop that boils pasta water in three minutes instead of seven, a car that is simply better than the one it replaced, or delivered no experience at all and pure economics instead. Kann names those two archetypes cleanly, and points out that residential solar belongs to the second: what unlocked it was the lease and power purchase agreement, which produced savings on day one and no consumer experience whatsoever.

From there both guests reject the idea of leading with the climate benefit, and Schroepfer goes further than the question asked. He argues the category label itself confuses the analysis, because at the end of it somebody is paying you, and they pay because the thing saves them money or makes their life easier. Chaset’s field answer is the same shape: selling residential solar and storage in the East Bay, the pitch was resilience first, for households losing power fifteen times a year and a fridge full of food with it, then bill savings. Customers who led with wanting to be green were a small minority, happy to have the benefit but not buying for it.

That criterion is what the hype game then runs on, and it is where the two split systematically. Chaset’s over-hyped calls all rest on constraints outside the product. Data center load growth is over-hyped not because the demand is absent but because connection capacity is limited and expanding it runs through utilities that need regulatory authority and rate cases, on timescales designed to be slow; Kann narrows the claim to the buildout rather than the demand and Chaset accepts the narrowing. Carbon removal is over-hyped because it asks consumers with short horizons to pay for a benefit twenty years out, which means governments have to buy it, and governments were pivoting away. Schroepfer’s calls run the other way, on what capable organizations can engineer around. On data centers he insists on saying both: over-hyped in the sense that US demand growth is over-attributed to them when manufacturing and electric vehicles are a large part of it, and under-hyped in the sense that chips are being bought and powered up and well-resourced companies are being extremely creative about finding power. On carbon removal he calls it a pure economics problem, with nature-based routes such as enhanced rock weathering, ocean alkalinity enhancement and biochar getting under $100 a ton and the market opening when they do; he answers Chaset’s demand problem directly, saying consumers will never care but governments will in the long run. Fusion is the exception that proves the pattern rather than breaking it. Both guests call it under-hyped, but Chaset’s agreement arrives wrapped in his own constraints: he doubts the timescale on which these companies can build and deliver, and thinks local communities will take a while to accept that a fusion plant is safe, at least in the United States.

Their fullest convergence is on artificial intelligence, and it is instructive because the reasoning is the same shape as the consumer argument. Both call artificial intelligence applied to climate over-hyped. Schroepfer’s test is to draw one pie chart of where a problem’s value and difficulty actually sit and overlay a second of what AI genuinely solves, and to check whether they overlap. For materials discovery they mostly do not: AI is good at finding a new material and the hard, expensive part is manufacturing it cheaply at scale. He contrasts this with cancer therapeutics, where finding the right molecule is the hard part and scale-up is comparatively straightforward. Chaset’s version is blunter, that AI is not uniquely suited to climate problems, because these are businesses solving ordinary business problems and AI helps exactly to the degree it would help any business doing the same thing. Neither position depends on being sceptical about AI, and Schroepfer says plainly that he is not.