−5.0%
since Sep 29
The idea
I don't think any single AI lab wins the whole stack, so I'd rather own the toll booths than bet on one model. This basket holds neutral infrastructure that gets paid regardless of who's ahead, plus the rails agents need to transact and verify identity, a spread of chip makers so no single architecture bet dominates, and application companies rebuilding around agents instead of getting replaced by them. It keeps a cash buffer and spreads risk across many names rather than concentrating in one. I'd rethink this if one lab actually captured the entire stack top to bottom.
What it owns
- Infrastructure that processes AI workloads regardless of which model provider or chip vendor wins, collecting fees on volume and usage rather than picking sides. · CloudflareIt sits between users and every model provider, collecting fees on AI traffic flows that move through its edge network and gateway products regardless of which lab powers them., MongoDBIts document database and vector search capabilities serve as a default foundation for AI applications regardless of which model they query, with growth depending on real production deployments., DatadogIt monitors AI workloads in production across any model or infrastructure, collecting monitoring fees from companies running agents and inference at scale., ElasticIt powers retrieval search and vector storage for AI applications that fetch relevant documents before generating answers, competing against dedicated startups and cloud provider built-in options., TwilioIt owns the messaging and voice infrastructure that AI agents use to reach customers, positioning it to capture volume growth from agentic automation across industries.30%
- Payment and identity systems designed for machines to transact with each other, either through new crypto rails or existing financial networks adapted for autonomous agents. · CoinbaseIt operates the on-ramp and stablecoin rails for machine-to-machine payments that don't require a human bank account, positioning itself as the infrastructure for autonomous agents transacting with each other., OktaIt verifies agent identity and permissions for autonomous systems acting on behalf of humans, with growth tied to this becoming a standalone revenue category rather than a feature., CircleIts regulated stablecoin settles instantly between machines without a bank in the middle, positioned as the native money for machine-to-machine commerce if regulation favors compliant issuers., VisaIt processes agent payments over existing card networks and builds credentials designed for autonomous transactions, capturing volume if machine commerce grows on traditional financial rails rather than crypto.24%
- Companies that profit when AI compute buyers build multiple sources of chips and processors instead of depending on a single vendor. · Advanced Micro DevicesIt sells data center AI chips as an alternative to a single vendor, winning as hyperscalers commit capital to diversifying their inference compute sources., BroadcomIt builds custom AI chips and networking fabric for hyperscalers that want their own silicon instead of buying off-the-shelf processors, profiting when large customers expand proprietary chip programs., Arm HoldingsIt collects royalties on chip designs from every major builder of custom AI silicon, profiting directly from the diversification of inference compute away from any single vendor., QualcommIt sells processors that run inference on phones and laptops where no model lab controls the chip, growing if on-device AI becomes a real reason for hardware upgrades.22%
- Platforms and tools that help developers and merchants prepare their products, data, and systems for agents to interact with them directly. · ShopifyIt provides the structured product data and checkout APIs that shopping agents need to transact with independent merchants, staying relevant if it remains the default transaction point rather than getting cut out by agent platforms., SnowflakeIt lets companies run any AI model against their own data warehouse instead of shipping data to a model lab, capturing value when enterprises treat their data center as the hub for AI rather than depending on external platforms., GitLabIt manages the full pipeline around code when agents write and deploy it, with exposure to whether customers expand seats and usage as agent coding spreads or contract as tools replace headcount.14%
- Cash10%