AI Infrastructure
Hardware and systems powering AI workloads — GPUs, servers, networking, cooling.
Active institutional exposure flipped from a net outflow of $21.5 billion in Q1 2026 to a net inflow of $14.5 billion in Q2 2026. 51.9% of companies in this theme saw increased active positions in Q2 2026.
Ecosystem Map
HBM, DRAM and flash optimized for AI workload bandwidth demands. Companies that design and sell memory and storage products; excludes companies that merely use memory as a component in their own chips or systems.
High-speed interconnects, InfiniBand and ethernet fabrics inside AI clusters - the links between GPUs, servers and racks within a facility.
Rack-scale AI compute systems and integrated AI server platforms. Companies that design, manufacture or integrate complete servers and racks; chip and component vendors belong in their component sub-themes and qualify here as enablers at most.
Liquid cooling, immersion cooling and thermal management for high-density AI compute.
Graphics and tensor processors purpose-built for AI training and inference.
Optical transceivers, DSPs, lasers, silicon photonics and co-packaged optics for moving AI-cluster data. Distinct from AI Networking (fabric switching) and from Fiber & Connectivity (metro, long-haul, subsea).
Bottleneck Tracker
Bottleneck details are part of the Pro plan.
Recent Discoveries
Recent discoveries are part of the Pro plan.