Pat Gelsinger was wrong for Intel

Original Article ↗ Hacker News Discussion ↗

Assessment of Gelsinger’s Tenure

  • Many see him as “right person, 10 years too late”: technically capable, honest about Intel’s problems, but inheriting a near‑unsalvageable situation (10nm disaster, fab lag, loss in mobile/GPUs).
  • Supporters credit him with:
    • Ending buybacks, cutting dividends (too slowly for some), and pushing “5 nodes in 4 years” culminating in 18A.
    • Massive fab investments (Arizona, Ohio, Europe) and winning early foundry customers and PDK usage from big cloud vendors.
  • Critics argue:
    • Revenue, margins, and cash flow cratered on his watch; headcount rose ~20% despite falling revenue.
    • 18A/20A delays and cancellations, poor Arrow Lake launch, Raptor Lake instability, weak Arc and Gaudi execution, and no major external 18A wins yet.
    • He over‑promised (e.g., impact of Larrabee) and under‑delivered, while staying highly media‑visible.

Board, Culture, and Governance

  • Strong sentiment that Intel’s board—heavy on finance and “Boeing‑style” executives—prioritized dividends, buybacks, and short‑term stock price over engineering.
  • Some say he was constrained (e.g., couldn’t kill the dividend sooner); others counter that a CEO who truly insisted could have forced the issue or resigned.
  • Broader complaint: decades‑long cultural rot; entitlement, finance‑driven thinking, under‑valuation of software and VLSI design, and “no patience” for multi‑generation bets.

Foundry Strategy and Corporate Structure

  • Split vs. integrated:
    • One camp: Intel must fully separate fab and product; otherwise AMD, Apple, Nvidia, etc. won’t entrust their crown‑jewel designs to a competitor.
    • Other camp: only a vertically integrated “design + fab” Intel can justify the CHIPS‑funded investment and compete with TSMC.
  • Consensus: Intel is caught in a bind—design side subsidizes an unprofitable foundry; yet foundry must succeed for Intel to matter long‑term.

GPUs, AI, and Software Ecosystem

  • Many argue Intel’s biggest strategic misses were:
    • Abandoning Larrabee/Xeon Phi and failing to persist for a decade the way Nvidia did with CUDA and GPUs.
    • Repeating a pattern of killing promising, non‑core projects that weren’t instantly high‑margin.
  • Strong thread that Nvidia’s real moat is software and developer experience, not “bean counting” or pure fab tech; skepticism that Intel’s culture can support a CUDA‑class ecosystem.

Customer Trust, Tofino, and Product Discipline

  • The Tofino cancellation is seen as emblematic: Intel acquired a strong product, promised a roadmap, then killed it early and even skipped its own EOL policy.
  • This feeds a widely shared view that:
    • Intel repeatedly abandons partners and ecosystems (Larrabee, Knights, mobile, Tofino).
    • Roadmaps are overly complex (“Lake” soup) with frequent cancellations, node switches, and slips, eroding confidence.

Broader Themes: Capitalism, MBAs, and Comparisons

  • Frequent comparisons to:
    • Nvidia (founder‑led, long‑horizon, software‑centric).
    • AMD (lean, patent‑ and design‑driven, fabless).
    • VMware, Google, Microsoft, IBM: how incumbents either cannibalize themselves successfully or ride declines.
  • Several commenters tie Intel’s problems to:
    • “Late capitalism”: short horizons, financial engineering, index‑fund‑dominated ownership, and boards insulated from technical reality.
    • The danger of non‑technical or MBA‑centric leadership in deep tech, though some note technical leaders can also ignore business realities.
  • Unclear in the thread whether any CEO could realistically reverse Intel’s trajectory under current market and governance constraints.