Apple Will 'Watch Everything Burn' When the AI Bubble Bursts

Original Article ↗ Hacker News Discussion ↗

Skeptics of the current AI boom argue that large language model companies are burning cash on infrastructure and subsidized tokens in a way that resembles a financial bubble, with uncertain paths to sustainable profits once true costs are passed on to customers. Others counter that rapid capability gains, strong developer adoption, and enterprise willingness to pay high per-employee fees indicate durable demand, even if valuations eventually correct. Apple’s comparatively cautious approach—focusing on on-device AI and avoiding massive cloud bets—is framed either as shrewd risk management that will age well if the bubble pops, or as a strategic gamble that could leave it trailing if centralized “frontier” models become a lasting platform.

Perceived AI Bubble and Profitability

  • Many agree current AI valuations look bubble‑like, but disagree on severity: from “violent crash with recession risk” to “dot‑com‑style correction but lasting tech.”
  • Skeptics argue:
    • AI vendors are structurally unprofitable: huge capex, expensive inference, opaque securitized debt comparable (in size) to pre‑2008 mortgage products.
    • Even if inference margins improve, sunk infra and debt may never be repaid; some capex will be stranded.
  • Optimists counter:
    • Enterprise willingness to pay (e.g., high per‑developer token budgets) shows clear business value.
    • Revenue growth at major labs and falling per‑token costs suggest a path to sustainable margins, especially with usage‑based pricing.

Usefulness of LLMs for Software Development

  • Views are polarized:
    • Some developers report large productivity gains, say they can’t imagine working without AI, and cite increasing reluctance to participate in “no‑AI” studies.
    • Others (notably systems/C++ developers) find LLMs poor at code generation, acceptable only for review or boilerplate, and feel claims of 10x boosts are overblown.
  • Several commenters stress “it depends” on skill, domain, and workflow; tools are seen as amplifiers rather than replacements.

Assessment of the Critic’s Analysis

  • Supporters praise deep dives into revenue, capex, and circular financing; see the critic as one of the few systematically challenging hype.
  • Detractors say:
    • Arguments mix solid forensic finance with misunderstanding of technical trajectories.
    • Numbers and sources are cherry‑picked; when rebutted, goalposts shift rather than errors acknowledged.
    • Claims about AI’s lack of business value, plateaued capabilities, and unsolvable hallucinations are called outdated or overstated.

Token, Subscription, and Cost Economics

  • Debate over whether subscriptions are fundamentally loss‑making:
    • One side: theoretical “maxing out” per‑user limits far exceeds subscription revenue, implying subsidy.
    • Other side: average usage is likely far below caps (gym‑membership analogy), many users underutilize tools, and routing across cheaper models will expand margins.
  • Concern that if prices rise or novelty fades, casual users may churn, leaving only heavy power‑users and forcing tighter limits.

Apple’s AI and Hardware Strategy

  • Some see Apple’s relative restraint on cloud AI as savvy:
    • Avoids massive AI capex and debt.
    • Focuses on NPUs, on‑device models, and “private cloud” compute; aims to sell the end devices while others fight over commoditized tokens.
  • Others argue:
    • Apple misread or lagged on gen‑AI, fumbled its own models, and is now dependent on external providers.
    • DRAM and fab scarcity, driven partly by AI demand, already hurt Apple via higher BoM costs; benefits from any AI bust come only after short‑term pain.

On‑Device vs Cloud, Security, and Search

  • Some predict powerful on‑device models plus secure off‑device inference, driven by privacy and rising token costs.
  • Others doubt typical users care enough about data locality to resist cheap, centralized services.
  • A minority highlights “national security” and the strategic importance of domestic AI and data centers; another group calls this a pretext for subsidies and control.
  • Several note that AI is rapidly becoming a new search interface; incumbents with distribution (especially search defaults) may capture much of the value.

VR / Vision Pro and Apple’s Track Record

  • Vision Pro discussed as a parallel: seen either as:
    • A cautionary tale of Apple misjudging demand and releasing an expensive, niche device to appease investors.
    • Or evidence Apple can walk away from overhyped categories (VR, cars) without betting the company.
  • This history colors views on whether Apple will “watch AI burn” from the sidelines or has already missed a key platform shift.