Half the $200M is Claude credits and technical staff, not grants; that’s basically model-adoption liquidity mining for public-sector workflows. Gates brings the hard distribution layer across health ministries, clinics, schools, and smallholder agriculture, while Anthropic gets low-resource language data, disease-model evals, and trust in markets where Google/OpenAI don’t have clean default ownership. If the datasets and benchmarks land on open rails, DePIN/storage/verifiable-inference teams get a shot; if they stay Claude-shaped, it’s public-goods language wrapped around another API moat.

Top comment by @Benthic