Activate's real advantage is finding exceptional scientific founders early. Here's what the data says about that frontier, and how I'd turn it into a system.

Built independently from public data: Activate's own companies and fellows directories, USAspending, OpenAlex, SEC EDGAR, and IRS 990s. 224 ventures and 292 fellows, cohorts 2015–2025. Figures are directional where noted; the rankings are the signal.

The portfolio is not static, and that's the point. Between the 2024 and 2025 cohorts the mix swung hard toward climate (2% → 32% of the cohort) and electronics & connectivity (8% → 24%). Zoom out and the decade tells a clearer story: chemistry and materials remain the backbone, but the action is at the intersections, chemistry × climate, carbon × climate, computing × electronics. New fields are forming between disciplines, not inside them. A talent engine that reads last cohort's mix is already a year behind; the job is to read the acceleration.

Cross each field's research momentum (its growing share of global publications) and federal funding momentum against Activate's own presence, and a short list of whitespace falls out: energy storage and transportation & mobilityare both research-hot and funding-hot, yet sit in the single digits of recent cohorts. Some of that is deliberate (thesis fit), but it's exactly the list a discovery function should be interrogating every quarter, not as a target, but as a question: are we early, or are we missing it?

Activate's picks win money on their own. Matched against federal records, the portfolio has captured $227.2M in non-dilutive funding across 46% of ventures, DOE-led, before a dollar of equity. That is an independent, public validation of pick quality, and it's the kind of outcome signal a selection system should be tracking on every cohort, not discovering years later in a press release.

The founders themselves have a clear shape: 88% hold PhDs, concentrated at MIT, Berkeley, and Stanford, with deep, cited research that visibly precedes the venture, and the discipline maps onto the space (electronics ← electrical engineering, energy storage ← chemistry and materials). That's not a pedigree filter to enforce; it's a baseline that tells you what a strong candidate looks like, and, more usefully, helps spot the exceptional scientist who doesn't fit the usual mold.

Activate is an equity-free nonprofit that has grown revenue from $3.7Mto $26.7M since 2019, roughly 7.2×, funded by philanthropy and government, not venture capital. The model only works if the science selection is exceptional, because there's no portfolio markup to hide behind. That makes founder discovery and selection the literal engine of the institution, and it makes a continuously-improving, data-driven version of it a genuine moat.

1. 1. Instrument the engine

   Stand up exactly this pipeline as the internal source of truth: every applicant and fellow with their research footprint, and every cohort tracked against real outcomes (non-dilutive funding, raises, exits). Make the loop visible.

2. 2. Source the whitespace

   Run the frontier radar quarterly and pair it with the discipline → space key, so sourcing is pointed at accelerating fields and the scientist profiles that succeed there, with hubs treated as the distinct theses they already are.

3. 3. Close the selection loop

   Correlate selection-time signals with outcomes to learn what actually predicts a strong fellow, then feed that back into reviewer rubrics, carefully, with the bias guardrails the mission demands. The goal is a system that gets better every cohort, not a fixed scorecard.

4. 4. Productize the intelligence

   Package the frontier read as an offering, internal decision support for selection and sponsor cultivation, and an external Frontier Report that makes Activate the authority on where hard tech is heading.

None of this is a black box. The momentum figures use keyword-relevance search, so trust the ranking over the exact multiple; the founder sample is 112 resolved profiles across 100 companies and the data is descriptive, not a predictive model; and I've deliberately avoided inferring identity demographics, the right signal is scientific and career depth, used to widen discovery, not to gate it. The credibility is in the sourcing and the caveats, not in pretending the data is cleaner than it is.

Explore the live data behind every claim in the dashboard, or see exactly how each figure is computed in Data, methods & sources.