Anthropic Moved Claude to AMD Chips in Three Days
Anthropic moved Claude onto AMD chips in about three days instead of six months to a year. An in-house AI model called Mythos wrote the needed code itself.
Anthropic, the company behind the Claude chatbot, says it moved Claude onto AMD's computer chips in about three days, a switch that normally takes engineers six months to a year. The work was done by Mythos, Anthropic's own AI model, which wrote nearly all the specialized software the new chips needed. Tom Brown, Anthropic's co-founder and chief compute officer, says Claude is now trained and run on four kinds of chips: Nvidia's, AMD's, Google's TPU chips, and Amazon's Trainium chips.
The claim, made in an investor presentation and not independently verified, strikes at Nvidia's biggest advantage: CUDA, the software that has long made switching away from Nvidia's chips slow and expensive for developers. If a major AI lab can move to a rival's chips in days rather than months, picking a chip supplier stops being a months-long commitment. That weakens one justification for the premium prices Nvidia charges for its chips.
Source: the-decoder.com
Free course
Stop reading about AI — start building with it
The free Claude Code course: your first site, tool or game — no coding. No upsells, no cross-sells — nothing to buy here.
Start free →
Author
Evgenii Arsentev
PhD · Founder & CEO, Dad4Kids
Articles · Latest articles