Texas Teachers’ CIO Warns AI Capex Boom Echoes Past Busts

Jase Auby, CIO of the $225.3 billion Teacher Retirement System of Texas, at an investment committee meeting, comparing AI infrastructure spending to five prior US infrastructure buildouts — and warning that at ~3.5% of annual US GDP, AI capex already dwarfs all five.

Connections

Jase Auby of the Teacher Retirement System of Texas draws a historical comparison between the current AI buildout and the canals, railroads, electrification, US interstate highway, and early-2000s telecom/fiber booms. The cautionary pattern across them: massive overbuild, long waits for demand to catch up, and widespread bankruptcies before revenues materialized. He names Nvidia, Microsoft, Apple and Taiwan Semiconductor Manufacturing Company (TSMC) as the largest equity positions in TRS’s portfolio as of August 2025, with Nvidia alone at 2.45% of total equity.

The article also cites Monte Tarbox (NYC Retirement Systems, $327B) turning away managers to diversify from AI-related systemic risk, and Jagdeep Singh Bachher (UC Investments) calling AI “the biggest technological revolution at the fastest pace in history.”

Capital Allocation angle: Auby notes generative AI hit ~53% adoption within three years of ChatGPT’s release — faster than the internet or PC — and remains cautiously optimistic it will look more like the Eisenhower interstate buildout (government-funded, tolerant of decades-long payback) than the railroad or fiber busts.

Related reads: Where’re All The AI Chips? (Ed Zitron) takes the bearish case further, arguing hundreds of billions of NVIDIA GPUs are warehoused and uninstalled.

Notes

Jase Auby’s five historical analogies for AI capex:

1. Canals — overbuilt, many operators went bankrupt before rail displaced them
2. Railroads — massive overbuild, speculation, collapse (the classic Gilded Age bust)
3. Electrification — decades before widespread commercial adoption
4. US Interstate Highway System — the positive comparator: federal funding tolerated a long payback, and demand eventually justified the investment
5. Telecom/fiber buildout (2000s) — canonical tech overbuild; dark fiber sat unused for years

Key quote: “It’s a cautionary tale for the period that we’re in now, where you have some very large companies, some of the largest companies in the S&P 500, that are very much engaged in the build out of artificial intelligence.”

At the time of the article, Anthropic had just raised $65B in a Series H at a $965B post-money valuation, with investors including the Ontario Teachers’ Pension Plan and the Qatar Investment Authority — an illustration of how large pension funds are caught in the same dynamic Auby is warning about.