Thirty Years of Technology Investing (Fundamental Edge)

TL;DR — A 180-page year-by-year history of US tech stock returns, 1996 to October 2026, written as a CIO teaching a junior analyst to tell which kind of year they are in.

Released free by Fundamental Edge on October 18–19, 2026, as the first sector in Brett Caughran’s “30 Years of Investing” project, which aims to give a human analyst or an AI agent the “seen this story before” experience of many cycles by working through the record (see How Do You Train A… (tweet) for his theory of analyst training). The data grid is calendar-year price returns, without dividends, for the 300 largest US-listed tech companies each January 1, priced to October 16, 2026; the period quotes come from a dated news archive (FactSet StreetAccount, Quartr transcripts).

It is organised in seven parts: orientation and the full 31 years in one sitting; Mania and Bust (1996–2002); the Quiet Rebuild and the Crash (2003–2009); Zero Rates and the Platform Decade (2010–2019); the Liquidity Spike and the Multiple Event (2020–2022); the Concentration Regime (2023 to October 2026); and method notes. Each era opens with a short essay, and each year gets a 2–4 page chapter: cap-weighted and median return, the ten largest companies, the biggest winners and losers against their sub-sector, the price-to-sales multiple and rate backdrop, the dated “tape” of what people were saying, and lessons labelled JUDGMENTAL or MECHANICAL. The eight mechanical lessons report hit rates in a 1996–2011 fit window and a 2012–2026 test window.

The central finding is that the leaders and the median stock usually disagree: the cap-weighted return beat the median by more than 10 points in 15 of 31 years, and the median led in only four (2003, 2010, 2013, 2016). It distinguishes three kinds of bear market (2001 a demand shock with wide dispersion; 2008 and 2022 liquidity events in which every top-ten name fell; 2000 and 2002 in between); notes that trough recoveries in 2003 and 2009 saw 89–93% of names rise, led by distressed small chip and hardware names, while 2023 broke the pattern (median +27%, cap-weighted +60%, NVIDIA +239%); and finds that neither momentum nor reversal works as a strategy (last year’s top winners beat the next year’s median in 11 of 30 years, last year’s losers in 8). Price-to-sales hit 7.43x in January 2026, the highest in the grid, as the real 10-year yield rose from −1.0% (end of 2021) to 2.9% (September 2026). It names NVIDIA, AMD, First Solar, Netflix, Micron and Apple as the cases any rule should be tested against.

What it is: A practitioner’s teaching text, not research, and candid about its limits. It was drafted with model assistance and with hindsight (“every page was drafted with the answer key open”); the follow-through figures count survivors only, though Part VII reruns them with every delisted name as a total loss; the companion data volume with full tables is not included; and Caughran calls it “version 1.0”. Each year’s rate-and-valuation reading is offered as a starting point, not an answer: a rising real yield with a rich multiple preceded both the worst year in the grid (2000) and four strong ones (2014, 2019, 2023, 2025).

Where this came from

• Linked by Brett Caughran in a reply to his X post on training digital hedge-fund analysts, October 19, 2026.
• Publicly available: the “Download the PDF” link on Fundamental Edge’s research library.

180 pages. A copy is archived locally against link rot; the header links the original source.

Connections

• How Do You Train A… (tweet) — the post that released it.
• Brett Caughran — author (Fundamental Edge, Analyst Academy).
• Pattern Recognition — the book’s purpose: experience gained by working through the record.
• Hedge Funds — written for hedge-fund analysts (and the AI agents Caughran is training).
• Best Ideas… (tweet) — saved the same day; another look at what actually drives fund returns.