AEACapital Research

Independent student research — not an investment firm or financial advice

AEA Capital Research

Selected Work

2026 editionResearch / cases / record

A working selection, not a victory lap.

These are the pieces that best show the practice: an argument made from primary material, a decision with a countercase, a tool that exposes its assumptions, and a correction that stays public.Read the full record →

Lead dossier / AI infrastructure

The Capacity Test

Before an AI-infrastructure story becomes an investment claim, four things must be true: demand must be observable, capacity must be delivered, financing must hold, and the economics must reach cash.

Research dossierPrimary-source framework
Read the dossier →

The desks

Four kinds of work.
One standard of proof.

01 / Research

The AI Demand
Evidence Ladder

A five-layer test for separating an appealing narrative from changed operating reality.

Read the framework →02 / Investment case

AMD: the
second-source thesis

A live decision record: the premise, countercase, re-underwrite test, and follow-up.

Open case file →03 / Filed work

Intel’s segment
economics

The segment disclosure that made the divide beneath the earnings headline visible.

Read the reporting →04 / Public record

The work that
didn’t work

Open losses, material errors, and the process changes that followed—kept beside the wins.

Open The Record →

From the files

The pieces worth returning to.

Capital allocation

The AI Capex Reality Check

What hyperscaler spending needs to earn back through cash generation and return on invested capital.

Power & infrastructure

Inside PJM’s Capacity Crunch

The reliability constraint under the data-center buildout—and why the marginal megawatt matters.

Field research

Data Center Alley

A physical map of Northern Virginia’s infrastructure corridor and its public-market exposures.

Published record

Where the numbers have to agree.

These are the pages where narrative gives way to a dated, reconcilable record.

Month-end tear sheet / August 2026The book, position by position.

Thirty-two positions, real cost bases, end-of-month closes, and the cap arithmetic that was corrected after the fact.

Read the tear sheet →
Ongoing performancePerformance Tracker

Published return series against stated benchmarks, with its limitations kept visible.

Open tracker →Monthly letterThe Month the Site Went Stale

An audit that put three out-of-date claims ahead of performance discussion.

Read the letter →

The instrument panel

Tools that make
the assumption visible.

ValuationReverse DCF

Start with what the market price already requires.

Portfolio constructionRisk X-Ray

See the concentrated exposures a ticker list hides.

Portfolio constructionConcentration Premium

Test whether several positions are really one bet.

The selection principle

Publish the argument, show the work, and leave the revision where readers can find it.See wins, losses & corrections →