Google boss says AI investment boom has 'elements of irrationality'

Original Article ↗ Hacker News Discussion ↗

Is AI a bubble – and where are we in it?

  • Many see classic bubble signs: massive capex, circular funding between hyperscalers, GPU vendors, and model labs, plus obvious disconnect between revenues and valuations.
  • Others argue we’re closer to “1995 dotcom” than 2000: useful tech, still early, bubble likely to grow before it bursts.
  • Several emphasize that a bubble pop doesn’t mean AI is worthless, just that prices will be violently re-rated.

Utility vs. economics of AI

  • Multiple anecdotes of big productivity gains (especially in web dev and coding) – “week of work to 15 minutes” type stories.
  • Critics counter that current tools are heavily subsidized, prices are arbitrary (e.g., sudden 10× price hikes), and inference likely isn’t really profitable once full costs are included.
  • Consensus in the thread: AI clearly has value; unclear whether current valuations and pricing reflect sustainable unit economics.

Winners, losers, and systemic risk

  • Big tech (Google, Microsoft, Meta, Amazon) seen as able to eat large AI losses; startups and pure-play labs (OpenAI, Anthropic, Oracle’s AI push) viewed as fragile and acquisition targets “for cents on the dollar” after a crash.
  • Nvidia is a focal risk: if customers fail or GPUs obsolete quickly, both earnings and broad index funds could get hit hard.
  • Concern over SPVs and securitized datacenter debt ending up in pensions and insurance portfolios, echoing pre‑2008 structures.

Labor, productivity, and inequality

  • Disagreement whether AI is already displacing workers or just correcting COVID over‑hiring. Some report real team‑size reductions tied to AI; others say execs are simply using the hype as cover.
  • Fears that broad automation plus offshoring push more people out of “good jobs,” concentrating consumption in the top 10% and increasing pressure for UBI.
  • Others cite history (mechanization, PCs) to argue productivity shocks eventually create new industries and roles.

Product quality, trust, and user behavior

  • Strong backlash against “AI in everything”: degraded search results, intrusive copilots, low‑value summaries, and hallucinations eroding trust.
  • At the same time, many non‑technical users reportedly treat chatbots as near‑omniscient and are shifting queries away from traditional search.
  • This tension—real utility vs. unreliability and over‑promising—is seen as a key driver of both enthusiasm and skepticism.