Intel's Humbling

Original Article ↗ Hacker News Discussion ↗

Intel’s long reign as the dominant CPU maker is increasingly challenged by TSMC’s foundry model, ARM-based designs, AMD’s resurgence, and Nvidia’s AI-focused GPUs. Commenters trace Intel’s troubles to years of strategic and cultural missteps—from missing mobile and underinvesting in R&D to botched process-node transitions and failed acquisitions—while also noting how Apple Silicon and cloud ARM servers exposed x86’s power-efficiency gap. There is cautious optimism that CEO Pat Gelsinger’s push to rebuild Intel’s manufacturing leadership and become a competitive foundry could succeed, but many doubt whether it can regain clear technological or economic leadership.

Apple Silicon and Intel Laptops

  • Many commenters see Apple’s switch to its own ARM chips as a major inflection: far better battery, thermals, and noise than recent Intel MacBooks.
  • Disagreement on whether Apple “sandbagged” Intel Macs with poor cooling:
    • Some argue Apple consistently under‑cooled machines, not as a plot but a design preference for thin/quiet.
    • Others say the same chassis ran dramatically better once M1 was dropped in, showing Intel’s power issues.
  • Debate on actual perf/W: some claim Ryzen and certain Intel mobile parts can match or beat M1 in similar TDPs; others cite benchmarks and battery life showing 2–3× better perf/W for Apple Silicon.

Intel’s Strategic Missteps

  • Long list of perceived errors:
    • Selling off its ARM (StrongARM/XScale) line just before smartphones took off.
    • Betting on Itanium/IA‑64 instead of AMD64 and having to adopt AMD’s ISA.
    • Massive but unproductive acquisitions (e.g., McAfee, Altera; frequent failed M&A).
    • Missing mobile, underperforming in GPUs/AI, and complacency while holding an x86 monopoly.
  • Spectre/Meltdown era cited as a trust and architecture warning that Intel downplayed.

Manufacturing, Nodes, and TSMC

  • Consensus that Intel’s process stumbles (10nm/Intel 7 delays) exposed design weaknesses.
  • TSMC portrayed as learning from Intel’s mistakes: cautious equipment buys, flexible capacity (shifting 5nm to 3nm and backfilling 7/6nm), strong foundry culture.
  • Discussion that node labels (“nm”) are now mostly marketing; performance, yield, and cost matter more than nominal size.

Competition from AMD and ARM

  • AMD seen as having seized the performance crown in many segments, especially servers (EPYC) and desktops, with Ryzen as the turning point.
  • ARM seen as a serious, long‑standing threat: phones since ~2007, now cloud (AWS Graviton, Oracle, others), and Apple’s Macs.
  • Some argue Intel “crushed” ARM earlier; others counter Intel failed in mobile despite heavy subsidies (“contra‑revenue”).

Intel’s Recovery Prospects and Foundry Pivot

  • Mixed views:
    • Optimists note Gelsinger’s aggressive roadmap (Intel 4 → 3 → 20A/18A, backside power), new fabs in US/EU, and visible talent re‑acquisition.
    • Skeptics highlight missed roadmaps, limited Intel 4 volume, reliance on TSMC for parts of Meteor Lake, and doubt Intel can match TSMC’s cost/yield or become a trusted high‑volume foundry.
  • Some see CHIPS‑act subsidies and national‑security concerns as a buffer; others fear politicization and “Boeing‑style” financialization.

AI, GPUs, and Future Architectures

  • NVIDIA’s GPU‑centric AI stack viewed as an entrenched ecosystem; alternatives (TPUs, NPUs, non‑GPU accelerators, Intel Arc) exist but haven’t displaced it.
  • Debate whether unified, CPU‑centric matrix hardware could eventually undercut discrete GPUs; current sentiment: software and tooling lock‑in make that uncertain.