What the data says about scientific founder discovery, and the emerging frontier.
An analysis of 224 Activate ventures and 292 fellows (2015–2025), built from public data. The aim isn't to flatter the program; it's to find the signal that's actually there, report the parts that aren't, and draw the operational implications. Methods and every source are in Data, methods & sources.
Who Activate funds is remarkably consistent
From their own bios, 88% of fellows hold a PhD, and the training pipeline is concentrated: MIT, Berkeley, and Stanford together account for nearly a third of all university mentions. The discipline a founder trained in maps cleanly onto the space they build in, electronics ventures come from electrical engineering, energy storage from chemistry and materials, industrial biotech from biology. This isn't a pedigree rule to enforce; it's a baseline that tells you what a strong candidate looks like, and a sourcing key, to find founders for a space, look where its science is done.
There is a clear, data-derived "typical founder" profile (doctoral scientist, deep cited research, a specific discipline per space). It works as a screening filter, not an arbitrary bar.
The most important finding: discovery works, selection doesn't (yet)
The single most useful result is a dissociation between two things people usually conflate. A founder's research footprint is a strong discovery signal: their pre-founding publications visibly foreshadow the venture, often two to four years early (Ryan DuChanois's membrane and separation work preceding Solidec's electrochemical chemical manufacturing; Bilen Akuzum's battery-materials research preceding Aepnus's battery-production process). That means you can find these scientists before they've incorporated, the earliest possible point of contact.
But that same footprint is not a selection signal. Splitting the portfolio at the median founder-citation count, the two halves win federal non-dilutive funding at essentially the same rate (about 38% vs 36%); the dollar gap that looks larger is driven by a handful of outsized DOE awards, not a broad effect. At this sample, research depth does not separate the outcomes.
Use research footprint to find founders, not to rankthem. It's an excellent radar and a poor scoreboard, and conflating the two would quietly bias selection toward citation counts that don't actually predict success.
The outcomes validate the picks, independently
Whatever selects these founders is working at the portfolio level. Matched against federal records, the ventures have captured $227.2M in non-dilutive funding across 46% of companies, led by the Department of Energy, before any equity round. For an equity-free nonprofit that has grown revenue from $3.7M to $26.7Msince 2019, that external validation of pick quality is the whole ballgame, there's no markup to hide behind.
The frontier is detectable bottom-up, and there's science-level whitespace
Rather than measuring Activate's 16 pre-defined sectors, you can let the literature surface what's accelerating. Ranking ~4,500 fine-grained research topics (including arXiv and bioRxiv preprints) by growth in publication share, the fastest-rising deep-tech areas include perovskite materials, advanced batteries, electrocatalysts, and energy-harvesting materials, and Activate already has fellows publishing in all of them. That's a good sign: the program is tracking the real frontier, not last decade's.
More interesting are the rising areas where no Activate fellow publishes yet, currently Concrete and Cement Materials, Cancer Genomics, Advanced Photocatalysis. These are the earliest sourcing leads a discovery function can act on, emerging science before it's a named category, surfaced by the data rather than chosen in advance.
Founder discovery should run two clocks: the slow one (who's succeeding in our existing spaces) and the fast one (what science is emerging that we have nobody in yet). The second is where being early pays.
Two more structural reads
Hubs are distinct theses, not branches. New York over-indexes on carbon management (+33 points vs the portfolio) and climate; the communities specialize, so sourcing and sponsor cultivation should be hub-specific. And new fields form at intersections: the densest pair in the portfolio is chemistry × climate, a reminder that the next venture often sits between two disciplines, not inside one.
What I'm not claiming
The momentum figures are keyword-based, so the rankings are sound but the exact multiples aren't; the founder analyses run on 112 resolved profiles across 100 companies and are descriptive, not predictive; the emerging-topic list is a candidate surface with real tagging noise, not a ranking to act on blindly; and I've deliberately not inferred any identity demographics, the signal here is scientific and career depth, used to widen discovery rather than narrow it. The credibility is in stating that plainly.
The strategic version of this is the Point of View; the live evidence is the dashboard.