Anthropic Evaluates Custom AI Chip Design to Combat Hardware Shortages and Fuel Claude’s Growth

Anthropic explores designing its own AI chips as revenue hits $30B. Learn how the Claude creator plans to tackle the global hardware shortage.

By: AXL Media

Published: Apr 10, 2026, 5:41 AM EDT

Source: Information for this report was sourced from CNBC

Strategic Diversification of the Silicon Supply Chain

Anthropic, the San Francisco based laboratory behind the Claude chatbot, is investigating the feasibility of designing its own proprietary AI chips. According to industry sources, the company is responding to a persistent global shortage of the specialized hardware required to train and deploy increasingly sophisticated large language models. While the project remains in its infancy, the move signals a potential departure from total reliance on third party silicon providers. A spokesperson for Anthropic declined to provide an official comment on these preliminary internal discussions.

Scaling Revenue and Infrastructure Investment

The push for custom hardware follows a period of explosive financial growth for the startup, which saw its run rate revenue climb from $9 billion at the end of 2025 to over $30 billion by April 2026. This surge in demand for the Claude model has necessitated a massive expansion of computing power. Earlier this week, Anthropic committed to a $50 billion investment aimed at strengthening United States computing infrastructure, highlighting the scale of the capital intensive requirements needed to remain competitive in the generative AI sector.

Deepening Partnerships with Google and Broadcom

Despite the exploration of independent designs, Anthropic continues to deepen its existing technical alliances. The company recently signed a long term agreement with Google and Broadcom to leverage and co develop Tensor Processing Units. Anthropic currently utilizes a hybrid approach, running its software on a mix of Google’s TPUs and Amazon’s custom chips. A transition to in house design would likely be a multi year endeavor, as the company has yet to assemble a dedicated engineering team specifically tasked with semiconductor architecture.