Data as of 8/30/2026. This product is a basket created by personnel of Agora Indexing Technologies LLC or its affiliates and is provided for informational purposes only. It is not a financial index, financial benchmark, or IOSCO-compliant product, and is not administered by Tilt Indices LLC. Product performance is shown for informational purposes only and does not constitute investment advice or a recommendation to buy or sell any security.
Composition
Themes
T1Power producers exposed to AI data center electricity demandT2Electrical grid equipment manufacturers (transformers, substations, transmission)T3Gas turbine and power generation equipment suppliersT4Nuclear energy operators and reactor developersT5Uranium producers and nuclear fuel suppliersT6Data center infrastructure providersT7Industrial infrastructure beneficiaries of AI capital expenditureT8Grid modernization and transmission infrastructureT9AI infrastructure buildout without model-winner riskT10Utilities serving large-scale hyperscaler and data center customers
Paper basket composition
as of 2026-06-10Long power producers and electrical infrastructure, not AI software
The market still talks about AI as a software story. I think Q3 is increasingly a power availability story.
The key bottleneck is no longer training models. It’s deploying enough compute at scale.
Every new AI model announcement ultimately turns into:
- More GPUs
- More data centers
- More electricity
- More transmission equipment
The entire stack above software has become the constraint.
My thesis
The biggest surprise of the AI cycle is that AI is behaving less like software and more like industrial infrastructure.
The market expected the value capture to look like:
AI applications → AI software → cloud providers → chips
Instead, we’re getting:
Electricity → grid equipment → data centers → chips → everyone else
The reason is simple:
You can copy software features quickly.
You cannot build a 500 MW power connection quickly.
A transformer, transmission line, gas turbine, substation, or nuclear reactor approval is far harder to replicate than another chatbot.
The core idea is that AI is transitioning from a software boom into an infrastructure boom, and infrastructure providers are often the last part of the value chain to be fully appreciated by the market. Given your existing interest in uranium and industrial exposure (URA, XLI), this is also one of the few AI-related theses that doesn’t require betting on which model or chatbot wins.