Interview with Jim Keller, a chip expert lauded for his work at AMD and Apple, on ARM vs x86 vs RISC-V, managing 1,000 people at Intel, and plans at Tenstorrent
Dr. Ian Cutress / AnandTech : Tweets: @iancutress , @anandtech , @0xcats , and @hkanji . Thanks: @anandtech Tweets: @iancutress : Jim Keller guide to engineering #18: New to the company? Obliviously argue with Gordon Bell. But only if you're right. https://www.anandtech.com/... https://twitter.com/... @anandtech : In semiconductor design, Jim draws attention. He's worked at DEC, AMD, Apple, Tesla. In this interview, @IanCutress talks to Jim the person, rather than simply Jim the engineer, to ask about his approach to work, life, and balancing it all https://www.anandtech.com/... Tip @techmeme https://twitter.com/... @0xcats : This interview is just a must read. So many insights and diverse topics from engineering, management, people skills and life in general. https://www.anandtech.com/... Hussein Kanji / @hkanji : There are no bad employees, there are just bad employee job matches https://www.anandtech.com/... Thanks: @anandtech
Context & Ripple Effects
This interview is Jim Keller's first extended sit-down since leaving Intel for Toronto-based AI chip startup Tenstorrent in January, and it picks up where AnandTech's 2018 conversation with him running Intel's Silicon Engineering Group left off — only now he speaks as an outsider to the company where he managed roughly 1,000 people. The throughline of his career, from DEC through Apple, AMD, and Tesla, makes his read on ARM vs x86 vs RISC-V more than armchair commentary.
First-order effects
- Keller's public architecture comparisons hand Tenstorrent a credibility signal no marketing spend could buy, right as the startup tries to establish its AI chip roadmap.
Second-order effects
- A figure with Keller's track record choosing an AI silicon startup over an incumbent reinforces the hardware talent moat concept — other senior architects face the same fork between big-company scale and startup equity, pressuring AMD, Apple, and Intel to compete on mission rather than salary.
Third-order effects
- If repeated departures of marquee designers — including his earlier exit from AMD after three years — keep flowing toward AI-focused startups, instruction-set choice shifts from inherited default to a deliberate strategic decision, weakening x86's structural lock-in.
The trend: Veteran CPU architects are increasingly trading incumbent x86 and ARM roles for leadership positions at AI silicon startups, with open architectures like RISC-V as their lever.