Not sectors. Enabling substrates. The picks-and-shovels of the next era. Each is independently a multi-trillion-dollar opportunity. The most defensible companies of the next decade will sit at the intersections.
Verticalized AI solving mission-critical problems with defensible data moats.
KPMG estimates global agentic AI spend at $50B in 2025, forecast to $155B by 2030. The durable winners will be domain specialists combining foundation-model leverage with proprietary data, regulatory expertise, and direct customer integration — not horizontal AI tooling, which the foundation-model providers will absorb.
Highest-confidence subcategories: vertical agents in regulated financial workflows, industrial-process AI in energy and manufacturing, agent-orchestration tooling, and trust/safety/MRM infrastructure for high-consequence deployments. The Venture Humanism orientation Alsop Louie articulates — accountable AI with safeguards built in — is a competitive sourcing advantage.
Next-generation architecture and hardware for massive-scale inference, agentic workloads, and energy-efficient compute.
The 2024–2026 inference build-out is structurally larger than the training build-out it followed. Power, latency, and cost-per-token are the binding constraints, each creating investable categories: novel inference accelerators, networking and interconnect, datacenter cooling, silicon-photonics, and chiplet architectures.
We have a particular sourcing edge here. One of the highest-multiple investments in the partnership's pattern library was an early bet on specialized deep-learning silicon — made before the post-2017 AI wave made it a consensus category, and acquired by a major semiconductor incumbent. Compute Technology rewards investors who identify the right architectural bet before the workload becomes consensus. We have done this work once. The pattern transfers directly to what we are sourcing now in inference accelerators, photonics, and post-quantum-ready silicon.
The financial layer of the Evernet. Programmable money, instant rails, hardware-rooted identity, post-quantum cryptography.
This is the family where the regulatory, technological, and competitive landscape has shifted most dramatically in the last 36 months. The GENIUS Act ended the decade of US digital-asset regulatory ambiguity. FedNow and RTP have crossed the threshold from pilot to production. Stablecoin market cap reached $317B by April 2026; stablecoin payment volume hit $9T in 2025. Visa and Mastercard have launched agentic-commerce products. The financial layer is becoming the connective tissue of the entire portfolio.
What we are sourcing, in rough order of conviction: agentic commerce infrastructure (identity, authorization, payment, dispute, reversal plumbing for AI-initiated transactions); vertical AI for regulated financial workflows; stablecoin and tokenization infrastructure; cross-border instant-payment rails; FinTech-for-hard-tech (capital-structure innovation for energy, climate, manufacturing, aerospace); trust, safety, and accountability for AI in finance; spatial / Evernet-native financial interfaces; hardware-rooted identity for agentic commerce; and post-quantum cryptography for financial infrastructure.
The picks-and-shovels of the financial layer are infrastructure-grade, bank-customer-validated, regulator-friendly companies positioned at the seam between today's rails and tomorrow's. An active portfolio company at the heart of this thesis has settled over $17 trillion to date, processes roughly $30B daily, and is live with 7 of the 9 largest global banks — the shape of what we are hunting for in Fund I.
The largest allocation, because energy powers everything else.
AI inference demand is restructuring grid economics. Onshoring of advanced manufacturing requires reliable industrial power. Electrification of transport and heating is moving the demand curve up structurally. The IEA and major utilities are converging on a view that global electricity demand will grow faster in the second half of the 2020s than at any point since the 1970s.
Highest-confidence subcategories: grid-scale storage; behind-the-meter generation for AI workloads; advanced nuclear including SMR and fusion-adjacent; and industrial-process decarbonization where capital-intensive hard tech meets novel financing. Two architectural ideas anchor our sourcing here: that the gap between in-service combustion efficiency (~30–40%) and physically-permitted efficiency (~62–67%) is one of the largest untapped optimizations in the global economy; and that ~72% of all global primary energy is lost as waste heat — heat is treated as a byproduct, when it should be treated as a resource. Two of the three thermal-system architectures named in those theses are in the portfolio.
Frictionless, high-bandwidth collaboration between people and intelligent systems. Where the Evernet renders for the user.
The hardware story has shifted in 2025–2026. Smart glasses are the breakout form factor — Ray-Ban Meta crossed 2M units with sales tripling in Q2 2025; the smart-glasses category was up an estimated 211% in 2025 per IDC — while bulky mixed-reality headsets remain enterprise-only. Investable opportunities are not in headsets but in the substrates that make next form factors work: optics, displays, sensors, and the agent / wallet / identity software layers that turn the device into a useful surface.
Where this family meets the financial layer: agentic wallets, payment surfaces native to glasses and AR, identity-and-attestation hardware for agent-on-behalf-of-human commerce. Where it meets Compute: edge-inference silicon. Where it meets Materials: photonics and display substrates. HMI is small by allocation, large by interlock.
Foundational, slower-cycle, asymmetric. Breakthroughs in materials science, bio-engineering, and manufacturing inputs.
The smallest allocation by design — slower-cycle, asymmetric bets — but the family produces some of the highest-multiple outcomes when they work. A decade of work in batteries, nanomaterials, solar, and aerospace gives the partnership a specialized sourcing network.
Most interesting subcategories: battery and storage materials (cross-pollinates with Energy); novel substrates and packaging for compute (cross-pollinates with Compute); bio-engineered manufacturing inputs displacing energy-intensive incumbents; and agricultural and food-system materials — nanocomposite coatings, biopolymers, and barrier technologies that reduce food waste, extend shelf life, and decarbonize agricultural supply chains. Food loss accounts for an estimated 8–10% of global GHG emissions. One portfolio company in this space recently signed an exclusive global distribution agreement with the leader in postharvest quality — the kind of slower-cycle compounding outcome materials investments produce when they work.
Most venture funds organize around sectors, which forces founders into a category. We organize around substrates, which lets the most ambitious founders compose across categories. Three illustrative interlocks:
Energy-efficient inference silicon. Datacenter siting decisions driven by power availability. Behind-the-meter generation specifically for AI workloads. The most interesting compute companies of the next decade will have an energy thesis baked in.
Vertical agents in financial workflows, in supply-chain coordination, in industrial-equipment fleet management. AI as the new connective tissue between systems that have not previously talked to each other.
Solid-state batteries. Advanced photonics for AR glasses. Novel substrates for high-density compute. Places where a materials-science breakthrough makes a previously infeasible product suddenly feasible.
Meet the team built for this moment.