The theses we're tracking.

POWER — CONSTRAINED

Power is a binding constraint on the AI buildout — in many regions, the tightest one

Through 2027, electrical infrastructure — grid interconnection, high-voltage transformers, and generation — is a binding constraint on AI data-centre expansion in many regions, working alongside compute and capital rather than replacing them as the limiting factor.

Last reviewed 17 July 2026Read the full thesis →

ELECTRICAL EQUIPMENT

Electrical equipment is the picks-and-shovels layer capturing AI-buildout value

Power-equipment makers capture durable, high-visibility value from the AI buildout — multi-year backlogs and lead times confer pricing power and earnings visibility through ~2030.

Last reviewed 15 July 2026Read the full thesis →

CAPEX FINANCING

AI-infra capex financing is structurally fragile

The AI buildout is increasingly funded by debt and private credit with circular vendor-financing and a duration mismatch — fragility that could force a slowdown if AI revenues fail to service the debt.

Last reviewed 15 July 2026Read the full thesis →

HBM MEMORY

Memory, not just logic, is the AI-compute chokepoint

HBM is draining DRAM wafers faster than logic scales — the memory wall is the AI bottleneck few price.

Last reviewed 17 July 2026Read the full thesis →

NUCLEAR / SMR

Firm carbon-free power is the AI build-out's scarcest input

Data-centre load doubles by 2030 and firm nuclear is the scarce answer — restarts now, SMRs later.

Last reviewed 16 July 2026Read the full thesis →

DEFENCE / AUTONOMY

The incremental defence dollar is shifting from primes to drones and counter-UAS

Ukraine rewrote the rules: value in defence is moving from primes to attritable drones and counter-UAS.

Last reviewed 15 July 2026Read the full thesis →