Peregrinations | Building the Future

Featured analysis

The Mercurial Muse

Nine months ago, we called the talent spending rational. Last week, the first model shipped. Now comes the harder question.

Expected nextLaunches

Our range is 2028–2029. The curve below shows where named forecasters sit.

'262026

'272027

'282028

'292029

'302030

'312031

'322032

Grid, electricity, nuclear, energy economics for AI compute.

GPUs, TPUs, ASICs, HBM, packaging — and the racks, networking, and cooling that turn them into a single computer.

Frontier labs, model launches, training, scaling.

Chatbots, coding terminals, cloud agents, knowledge work — the apps that run on top of frontier models.

Humanoids, autonomous vehicles, embodied agents — AI that has to act in a world that pushes back.

Explore METR capability limits, safety enablement ladders, and multi-source prediction models for the path to sovereign superintelligence.

Model the raw physical-to-digital translation from megawatt capacity directly to token throughput across state-of-the-art server fabrics.

Granular analysis of key physical bottlenecks and corporate battlegrounds.

Nine months ago, we called the talent spending rational. Last week, the first model shipped. Now comes the harder question.