Patil started earlier this week, succeeding co-founder Sam McCandlish, who will now resume the role of chief architect, focusing on pre-training and large-scale model training.
Both individuals will now report to Anthropic’s president Daniela Amodei.
The leadership shift includes changes to the company’s technical structure, incorporating product-engineering with infrastructure and inference teams. As CTO, Patil will supervise inference, infrastructure, and other engineering units.
This significant move comes as the company experiences mounting infrastructure competition from other AI leading players, including Meta and OpenAI. Meta revealed plans to spend up to $600 billion on U.S. infrastructure by 2028.
On the other hand, OpenAI has joined forces with Oracle on large-scale projects such as Stargate. Although Anthropic’s own spending is less clear, the company experiences pressure to balance scale, power, consumption, and speed.
Demand for Claude products has already strained Anthropic’s systems, fostering new rate limits for Claude Code in July.
Heavy users are now capped at 240–480 hours of Sonnet per week and 24–40 hours of Opus 4, depending on usage levels.
Patil brings a vast experience of up to 20 years of leading from roles at top-notch companies, including Stripe, Oracle, Amazon, and Microsoft.
Amodei appreciated his track record in building enterprise infrastructure, while Patil stated he was “thrilled” to be a part of Anthropic at a “pivotal moment in AI development.”
