I am trying to follow one system from end to end: what infrastructure is being built, whether people and businesses are actually using it, whether that use is creating durable value, and what growing AI demand means for the resources needed to sustain it.
How AI-optimised compute differs from conventional cloud infrastructure, what the available workload and power data show, and why megawatts alone do not tell us how much AI capacity is truly spare.
Explore the buildout →A recurring monitor of adoption, deployment, payment, usage depth and realised value — the public evidence we can use to test whether economically meaningful demand is strengthening.
Open the demand monitor →Is downstream AI demand growing fast enough to narrow the gap with the infrastructure being built and committed?
Public data do not give us a clean, recurring view of AI-capable capacity that is live, committed, available and actually utilised. So the gap cannot be measured honestly as one number today. These pages track the strongest observable evidence on either side instead.
The infrastructure explainer builds the upstream mental model: conventional versus AI-optimised compute, the changing workload mix, power demand and the limits of vacancy data.
The monitor follows downstream signals over time: whether businesses are adopting AI, deploying it more deeply, paying for it, using it more intensively and realising measurable value.
The sustainability work did not begin as a separate argument against AI growth. It asks what happens if that growth continues: can unnecessary computation be reduced, can resource use become more visible, and can people use AI more intentionally without losing usefulness?
Build → Use → Value → Resource impact. These are different projects, but increasingly parts of the same inquiry.