The Money Didn't Leave Climate Tech. It Left the Part That Matters Most.

The headlines say 2025 was a strong year for climate tech. US climate tech venture investment reached about $29 billion, the third-highest year on record. Global venture and growth investment climbed 8% to $40.5 billion after two years of decline.

Look one layer down and the story inverts. Per Silicon Valley Bank’s Future of Climate Tech report, ten late-stage deals captured 28% of all US investment. Per Sightline Climate’s 2025 investment trends analysis, seed investment fell roughly 20%, Series A deal counts fell 22%, and total deal count hit its lowest level since 2020. Early-stage venture’s share of total climate capital has fallen from around 20% in 2021 to under 8%.

The money did not shrink. It walked up the stage ladder, away from exactly the moment when a technology decides whether it becomes a company or stays a paper.

Why the early stage is the part that matters

Every hard technology passes through a phase where it is too proven for a research grant and too risky for growth capital. People call it the valley of death, and the name has been around long enough that we treat it as geography: a permanent feature of the landscape that companies must simply survive.

It is not geography. It is a market failure with a known shape. The technical risk at that stage is real but tractable; what is missing is an institution willing to retire it. Venture investors are increasingly unwilling. They now ask early-stage founders for signed offtakes and proven unit economics, which is to say they ask companies to arrive already de-risked. That is a rational response to a difficult decade. It is also an abdication of the one job venture capital exists to do.

So if equity will not retire early technical risk, something else has to. This is where I stop being a commentator and start being a witness.

The model I watched work

In 2018 I co-founded the Shell GameChanger Accelerator Powered by NREL and directed it through 2020. I have written elsewhere about what that program taught me, so I will keep the summary short. GCxN gave selected startups access to national laboratory facilities and technical experts, took no equity, and used a corporate partner to keep the work pointed at real markets. The companies we selected have gone on to raise more than $1 billion, and I want to be careful with that sentence: GCxN was one contributor among several in those companies’ journeys. Selecting them early, and retiring the specific technical risks a national lab is uniquely equipped to retire, is the contribution I will claim.

The public data says the model generalizes. The Department of Energy’s Lab-Embedded Entrepreneurship Program embeds entrepreneurial fellows inside national labs for two years. Its alumni have started 153 companies, created over 2,300 jobs, and raised $2.73 billion in follow-on funding. Those are remarkable numbers against modest program budgets, and they come from precisely the stage of the pipeline that private capital is abandoning.

Here is the uncomfortable arithmetic: as early-stage equity retreats, programs like these stop being a nice supplement to the capital markets. They become the load-bearing structure of American energy innovation. We are not funding them like load-bearing structure.

Three lessons for anyone building one

Having built one of these programs and studied several others, I would offer three design principles to anyone standing one up inside a lab, a university, or a foundation.

1. Selection is the product

An accelerator’s diligence process is worth more than its check, and this is doubly true when there is no check. The market eventually reveals whether you picked well, and it reveals it after you can no longer influence the outcome. Build a selection process you would defend in five years, because in five years someone will ask.

2. Corporate partners de-risk markets, not just technology

A national lab can prove a technology works. Only a customer can prove someone wants it. The most valuable thing our corporate partner contributed was not money; it was the constant, sometimes uncomfortable question of who would buy this and why. A pilot with a strategic partner de-risks a company in ways no amount of laboratory validation can.

3. Design for the graduation, not the cohort

The temptation in program design is to optimize the experience: the curriculum, the demo day, the cohort culture. Those matter, but they are inputs. The output is what companies raise, build, and hire after they leave you. Instrument the program around graduation outcomes from day one, or you will spend your annual reports describing activities instead of results.

The founders are still coming

Nothing in the funding data suggests a shortage of people willing to start hard companies. Applications to lab-embedded programs remain deeply oversubscribed. The technologies keep coming out of university labs and national labs at the same rate they always have. What has changed is the width of the bridge between the lab bench and the first commercial dollar.

Private capital has decided, at least for now, that building that bridge is someone else’s job. Fine. Then the someone else, the labs, the universities, the foundations, and the public programs, should recognize what they have become and resource themselves accordingly. The lab-embedded accelerator is no longer an experiment. It is infrastructure.