−5.1%
since Oct 15
The idea
I don't think any single AI lab wins the whole stack, so I'd rather own the layers that get paid no matter who's on top. This basket holds neutral rails that collect fees on AI traffic and data, tools that let agents pay and verify identity, a spread of chipmakers so no one vendor lock-in wins, and application companies rebuilding around agents instead of getting replaced. No single name dominates, and I'm holding a slice in cash. I'd rethink this if one lab actually captured the entire stack top to bottom rather than the ecosystem staying fragmented.
What it owns
- Infrastructure that AI applications need regardless of which model provider or chip vendor wins. · CloudflareIt sits between users and model providers, collecting fees on AI traffic regardless of which lab succeeds, and its edge computing tools position it to capture inference as it moves closer to end users., MongoDBIts document model and vector search capabilities make it a natural database layer for AI applications regardless of which model they call, and the take needs Atlas to show genuine adoption by production AI apps., DatadogIt monitors production AI workloads for any company running inference, collecting observability fees independent of which model sits underneath, and the take requires AI workloads to represent a growing portion of its usage., ElasticIt provides search and vector retrieval infrastructure that most useful AI apps depend on before sending queries to models, and the take requires it to defend that retrieval layer against specialized vector databases and cloud provider alternatives., TwilioIt owns messaging and voice channels that agents use to reach customers, plus customer data enabling personalization across any model, and the take requires AI-driven message and voice volume to show in its growth rather than flow through the labs' own channels.30%
- Payment and identity systems designed for machine-to-machine commerce and autonomous agent verification. · CoinbaseIf autonomous agents transact with each other, crypto rails eliminate the need for traditional bank accounts, and the take depends on agent payments actually moving through its infrastructure rather than remaining theoretical., OktaWhen agents act on a person's behalf, they need identity verification and permission validation, and the take assumes agent identity becomes a real revenue-generating product category rather than a conference keynote., CircleIts USDC stablecoin enables instant machine-to-machine payments without banking intermediaries, and the take depends on USDC circulation growth and regulatory favoritism toward compliant stablecoin issuers over offshore alternatives., VisaIt operates the existing card payment infrastructure that agents could use for commerce and is building agent-specific payment credentials, serving as a hedge if autonomous agents prefer traditional rails over stablecoins.24%
- Chip makers and designers positioned to benefit when AI compute buyers avoid dependence on a single vendor. · Advanced Micro DevicesIt competes directly against Nvidia by selling data center AI chips, and the take requires large inference buyers to commit real spending to its MI-series as an alternative rather than consolidating on a single vendor., BroadcomIt designs custom AI chips for hyperscalers seeking independence from standard processors and supplies the networking infrastructure connecting those clusters, with the take dependent on major customers continuing to expand custom silicon programs., Arm HoldingsIt licenses chip designs to nearly every company building AI silicon, from device makers to hyperscaler custom chips, collecting royalties on the proliferation of diverse AI computing hardware., QualcommIt supplies processors for phones and laptops where inference runs locally outside any lab's control, and the take assumes on-device AI becomes a genuine hardware upgrade driver where it maintains leadership.22%
- Platforms built to help enterprises deploy agents against their own data and transaction systems. · ShopifyIt provides merchants with machine-readable product data and checkout systems designed for shopping agents, and the take requires it to remain the default transaction point for agents engaging independent merchants rather than being bypassed., SnowflakeIt brings AI models to enterprise data rather than requiring data shipment to model providers, and the take needs its AI products to drive consumption growth as enterprises treat their warehouse as the system of record instead of relying on lab platforms., GitLabIt owns the code platform where agents are already performing real work, and the take assumes the surrounding pipeline of review, security, and deployment matters more as agents write larger portions of code.14%
- Cash10%