Sunday, August 9, 2026

TL;DR: Neocloud valuations depend on controlling scarce GPUs and contracts while outsourcing much of the underlying physical infrastructure.

The piece maps the awkward contractual stack between six neoclouds, GPU suppliers, and converted bitcoin miners that often own the harder assets. It asks who retains the economics when the most valuable intermediaries may mainly hold accelerators and leases.

TL;DR: Labor scarcity is industrializing data-center construction, but vendors' competing definitions of modularity obscure the real economics.

It shows how labor constraints are turning construction into manufacturing: precast walls, prewired electrical rooms, and even truck-delivered data halls. Its 60-plus-gigawatt vendor map then tests whether modularity claims survive contact with deployment economics, including higher equipment content per megawatt.

TL;DR: Agentic coding does not eliminate engineering bottlenecks; it moves them into judgment, quality, and organizational design.

Lamanna reports that Microsoft engineers are moving from hand-writing code to designing and directing an “agentic coding machine.” His operational lesson is sharper than “adopt AI”: teams must rebuild decision-making and quality systems around the new production model.

TL;DR: Gavin Baker argues AI infrastructure fundamentals remain stronger than public-market prices imply, though financing and regulation still matter.

Baker builds the bull case from hyperscaler cash flow, GPU spot prices, cloud demand, and memory-contract game theory rather than stock momentum. Patrick O'Shaughnessy also probes what could break it, from financing strain and regulation to changing infrastructure demand from open models.