How AI conquered the US economy: A visual FAQ

AI productivity and who captures the gains

• Several commenters note rising revenues for AI and hardware firms but little visible uplift for customers or the broader S&P 490, suggesting “shovel sellers” are winning.
• Explanations offered: delayed payoff due to implementation lags, weak managerial ability to leverage GenAI, limited competition allowing vendors to capture most surplus, and uncertainty preventing smaller firms from investing.
• Some argue AI may mainly reduce costs rather than raise revenues, and that it’s unclear yet whether promised productivity gains are materializing at scale.

Bubble, hype cycles, and historical analogies

• Many liken the moment to the late‑1990s internet or 19th‑century railway manias: real, transformative technology plus capital overreach and eventual bust.
• The consensus is “both revolution and bubble”: long‑term impact likely large, but current capex and valuations may be unsustainable and overly concentrated in a handful of firms.
• Others think an AI crash might be more muted than dot‑com because AI hasn’t permeated everyday jobs and consumption to the same degree.

Model progress, limits, and technical skepticism

• Debate over GPT‑5 and recent models: some see only incremental improvements since GPT‑3.5 and suspect LLMs are hitting a performance plateau; others cite blind tests and benchmarks showing meaningful gains, especially for coding and reasoning.
• There is concern that current architectures may be approaching an inherent ceiling, with no clear path from today’s systems to reliable “agentic” automation or AGI.

Actual usefulness and user experience

• Split views: some say AI meaningfully boosts productivity (especially for coding, drafting, research) if treated like an error‑prone assistant whose work is checked.
• Others report no compelling use cases in their own work, or that AI integrations are degrading products (e.g., customer support, search, content “slop”).
• Trust is a recurring problem: hallucinations, inconsistent quality, and lack of incentives or accountability make many reluctant to rely on outputs beyond brainstorming or first drafts.

Market concentration, capex, and sustainability

• Commenters highlight that most growth and capex come from a small S&P “top 10,” with AI‑related spending dominating GDP contributions while the “S&P 490” stagnates.
• Some see this as a power‑law dynamic in a mature, low‑growth economy; others as a dangerous concentration where AI/data‑center investment crowds out more broadly productive uses (e.g., housing, manufacturing).
• High GPU and data‑center costs, plus unproven business models, fuel fears of an eventual AI capex bust once growth or pricing power falter.