Reading the AI bubble — a syllabus for both sides of the argument

Whether AI is a bubble, a transformation, or both at once is the most consequential disagreement in 2026 — and the one most poorly served by the surrounding discourse. Every week the Twitter-and-Substack cycle produces another piece that confidently argues one direction while citing the same handful of poorly-understood charts. A serious view requires reading both sides at length, plus the historical baselines neither side cites enough. This page is a reading list for that effort, organized by stance and by source depth: the bubble-skeptics making the case that we're in a 2000-style top, the bull-case people taking the diffusion-of-capability angle seriously, the historical references worth knowing before staking out a position, and the publications tracking the actual on-the-ground economics of compute, model deployment, and corporate AI revenue. One deliberate omission: doomer-vs-accelerationist arguments about AI safety are not the same conversation as the bubble question and would muddy this one — they get their own syllabus.

A reading list for taking the AI-is-a-bubble question seriously, from both directions.

The bubble case

  • Ed Zitron — Where's Your Ed At?
    The most aggressive bubble-skeptic on the beat. Sometimes overstated; often correct on the specifics, especially OpenAI's revenue math. Read the longform AI revenue analyses; skip the rhetorical flourishes if they don't land.
    webpage
  • Gary Marcus — Marcus on AI
    Marcus has been right about LLM limitations for years and somewhat wrong about implications. Read for the technical critiques of scaling and benchmarks; treat the prognoses as one data point among many.
    webpage
  • Cory Doctorow — Pluralistic
    Doctorow's Pluralistic newsletter covers AI enshittification with the same framework he applies to platforms. The political-economy angle most AI critics miss, and the writer most willing to be specific about who profits from each cycle.
    webpage
  • AI's $600B Question (David Cahn, Sequoia)
    Sequoia partner's 2024 essay calculating the implied infrastructure-ROI gap. The reference document for 'where will the revenue actually come from?' arguments. Notable because the author is inside the bubble he's questioning.
    webpage
  • AI Snake Oil (Narayanan & Kapoor)
    Princeton researchers' newsletter (and book). The most consistent practitioners-of-machine-learning critique of AI hype, with primary sources. Best entry point for the academically-credentialed skeptic position.
    webpage

The bull case (or skeptic-of-skeptics)

  • Ben Thompson — Stratechery
    Thompson's framework treats AI as the next platform layer; his strategic analyses of OpenAI, Microsoft, Google, and Anthropic are the most-cited bull-case sources in tech business writing. Subscription required for the full archive; some pieces are free.
    webpage
  • Dwarkesh Patel — podcast & blog
    Patel's long-form interviews with frontier-lab principals (Karpathy, Dario Amodei, Demis Hassabis, Sutton) are the canonical insider record. Whether you find them bullish depends on whether you find the subjects credible. Either way, required listening.
    webpage
  • Andreessen Horowitz
    Andreessen's a16z is the largest VC bull on AI. Their content is partisan but worth reading to understand what the capital actually believes about deployment timelines, agent infrastructure, and consumer adoption.
    webpage
  • Marginal Revolution (Tyler Cowen)
    Cowen treats AI as the most important economic development since electricity. One of the few places that covers AI economics with the rigor of a working economist, and one of the few willing to keep changing his mind in public.
    webpage

Historical baselines

  • Devil Take the Hindmost (Edward Chancellor)
    Chancellor's history of speculative manias, from tulips to dot-com. Required baseline for anyone making bubble claims with confidence. The patterns rhyme more than the present-tense pundits acknowledge.
    webpage
  • Manias, Panics, and Crashes (Kindleberger)
    The classic financial-history text on speculative episodes. Cited in essentially every 'is this a bubble' argument worth reading. Worth reading once cover-to-cover before forming a strong opinion.
    webpage
  • Carlota Perez — Technological Revolutions and Financial Capital
    Perez's framework explains how technological revolutions reliably produce financial bubbles as part of their deployment. The pattern that 'this time is different' arguments need to dispute, not ignore. Read the book if you'll only read one history-of-technology source.
    webpage
  • W. Brian Arthur — increasing returns
    Santa Fe Institute economist whose work on increasing returns explains why some technology bubbles are followed by genuine winner-take-most outcomes. Critical for distinguishing bubble-then-bust from bubble-with-eventual-payoff patterns.
    webpage

On-the-ground economics

  • The Information
    Paywalled but cited everywhere. Their AI reporting on Microsoft, OpenAI, and Anthropic revenue figures is the most-trusted source of leak-driven business reality. Subscribe if you want to read the source material everyone else is paraphrasing.
    webpage
  • SemiAnalysis (Dylan Patel)
    Covers AI infrastructure with the technical rigor you'd expect from an analyst whose readers may be short-selling NVIDIA. The serious reference for compute economics, supply-chain bottlenecks, and lab capex.
    webpage