Q&A with AMD CTO Mark Papermaster on the company's bet on 7nm instead of 10nm as Intel's chips face delays, plans to win market share in the server space, more
Dylan Martin / CRN :
Context & Ripple Effects
In August 2018, AMD's process-node gamble was still a live question: Papermaster is defending the decision to skip 10nm entirely and go straight to 7nm at TSMC, precisely because Intel's own next-node launches were slipping. The related coverage shows how that bet compounded — by November, Papermaster was detailing the chiplet architecture pairing 7nm CPU dies with 14nm IO dies that made early 7nm economically viable.
The arc since then validates the framing of this interview: AMD moved from explaining the node bet to executing it, with the CEO talking up a fast production ramp of EPYC server chips a year later (quickly ramping EPYC production to hold the advantage created by Intel's stumbles), and by 2026 Papermaster claiming AMD as the first customer for TSMC's 2nm process (first in line for TSMC 2nm) — the same foundry-first strategy this 2018 conversation laid out.
First-order effects
- Intel's delayed next-generation launches hand AMD a timing window in servers, where EPYC can win sockets on process-node advantage alone before Intel's roadmap catches up.
- Server buyers gain a credible second source for high-core-count CPUs, shifting procurement conversations that had been single-vendor by default.
Second-order effects
- TSMC's 7nm capacity becomes a strategic asset rather than a commodity purchase — whoever secures leading-edge wafers sets the pace of the x86 server race, raising the stakes of AMD's foundry partnership.
- Intel is forced to compete on its own execution timeline instead of dictating the industry cadence, pressuring its node roadmap and pricing in the datacenter segment.
Third-order effects
- If the pattern holds, fabless design plus a committed foundry relationship becomes the structural template for competing in leading-edge compute — a trajectory the corpus traces from the 2018 7nm bet through AMD's claimed first-customer status at TSMC 2nm in 2026.
- Server market structure shifts toward sustained two-vendor competition, where share changes are driven by node transitions and supply commitments rather than incumbent lock-in.
The trend: Leading-edge x86 competition is being decided by foundry partnerships and node-transition timing, with AMD's TSMC-first strategy turning process risk into its primary weapon against Intel.