“Web3 needs enterprise-grade speed, trust, and security - and ZK is how we scale to the next billion users.” - Subho Banerjee, Head of Web3 Customer Engineering at @googlecloud In the opening talk of Verifying Intelligence at @token2049, Subho lays out @GoogleCloudTech's three-pillar strategy at the intersection of Web3 and AI: 1. Planet-scale infrastructure for rollups and chains (custom right-sized VMs, elastic node pools, RPC at scale, public on-chain datasets, faucets across 25+ chains). 2. AI for Web3 & Web3 for AI (Gemini and Model Garden options like Llama/DeepSeek, an Agent Development Kit and managed Agent Engine, plus early work on stablecoin payments to power agentic use-cases). 3. Security end-to-end, with Mandiant’s incident response and a dedicated Web3 task force to harden both on-chain and off-chain surfaces. He then zooms into ZK: ongoing research collaborations and papers; applied identity use-cases; ZK acceleration on GPUs/TPUs; rapid horizontal scaling to thousands of nodes; and performance tuning from orchestration down to hardware. Subho spotlights joint work with @0xPolygon on zkEVM - custom VMs, 3× scale, and high-throughput databases - plus ship-ready security practices (scan-before-ship, signed images, single-pane observability). Finally, Subho sketches the near future: agentic crypto that plans, searches, and executes on-chain - backed by verifiable, transparent infrastructure so users can trust what intelligent systems do with their assets. If you care about taking ZK rollups from demos to dependable, planet-scale systems - and about making AI agents verifiable - this is your roadmap from the team running some of the world’s most battle-tested infrastructure 🤝 Verifying Intelligence was co-organized by @HouseofZK and @boundless_xyz, in partnership with @googlecloud, and supported by @VeridiseInc, @OntologyNetwork, @0G_labs, @LazAINetwork, @gonka_ai, & @alt_layer. Also available on the YouTube: