All Insights

Most AI inference runs through three cloud providers and one or two chip manufacturers — a single point of failure smaller than a mid-sized corporate IT department, for the entire global AI stack. Why AI dominance is no longer the real question, and what building resilience around it actually requires.

A hospital running AI-assisted radiology for a decade doesn't have the radiologists to revert — they were never hired, never trained. Why the fallback to pre-AI operation isn't a step backward in time for the most dependent industries. It's a step off a cliff.

Almost all AI inference runs on chips from one dominant manufacturer, fabricated in a handful of facilities in one region — a single-threaded dependency underneath the entire global AI stack. Five credible failure paths, each with real historical precedent, and the one most dangerous because it wouldn't look like failure until it's too late.

AI has quietly become the third layer of critical infrastructure, after power and internet — but unlike those systems, it has no equivalent resilience standards, regulatory oversight, or stress-testing. A sector-by-sector risk map of what breaks first, and why the fallback keeps disappearing.

SMMT's 5-year bull case reaches 500-1000% — but carries a real scenario of losing nearly everything. LLY, despite an 800% run since 2019, now trades at 38x earnings with a base case barely ahead of the S&P 500. A side-by-side scenario analysis across three completely different risk profiles.

A tech company's marginal cost of serving one more customer is fractions of a penny. A pharma company's cost scales with every pill manufactured, shipped, and monitored. That single structural difference explains more of the tech-pharma performance gap than any other factor — and it's not the whole story.

Institutional-quality pharma research requires no Bloomberg terminal — it requires knowing that FDA Medical Reviews, R&D Day slide decks, and ClinicalTrials.gov protocols exist, and reading them critically. A complete map of every public source serious pharma investors use, and a step-by-step research workflow you can apply to any company.

The press release always quotes the CEO — but the AstraZeneca-Daiichi Sankyo ADC deal, arguably pharma's most value-creating business development move of the decade, was found and built by the BD team before Soriot ever saw it. A map of who actually decides what gets built, and why the highest-cost mistakes rarely make headlines.

Pascal Soriot rejected an 87 billion euro takeover bid in 2014, betting the company's future on a single compound that became Tagrisso. David Ricks calls his leadership approach "bumpers on the bowling alley." Eight pharma CEOs, profiled by leadership archetype — and why matching the right archetype to the right strategic moment matters more than any individual decision.

Lilly and Novo Nordisk fight head-to-head over GLP-1 patients. AbbVie and JNJ never share a patient, yet compete intensely at the payer negotiation table. Two companies can be in the same disease and not compete at all — a five-part framework for understanding what's actually driving pharma competitive dynamics.