Anthropic Eyes Custom AI Chip Design as Claude's Demand Hits New Highs

Artificial intelligence company Anthropic is exploring the possibility of designing its own chips, as the race to secure AI computing power intensifies across the industry, according to a Reuters report.

The plans remain in early stages. Anthropic has not committed to a specific chip design, nor has it assembled a dedicated team for the project. The company may ultimately choose to continue purchasing chips rather than building its own.

The move comes as demand for its AI model Claude has surged sharply in 2026. Anthropic's annualised revenue has crossed $30 billion, a steep jump from around $9 billion at the end of 2025.

Currently, Anthropic relies on a mix of chips to power Claude, including tensor processing units (TPUs) developed by Google and custom chips from Amazon. Earlier this week, the company signed a long-term agreement with Google and Broadcom, which helps design the TPUs, as part of a broader $50 billion commitment to strengthen US computing infrastructure.

Building a Chip Doesn't Come Cheap

Designing a custom AI chip is a costly undertaking. The price tag can be at roughly half a billion dollars, covering skilled engineering talent and rigorous manufacturing quality control, the report added.

Notably, Anthropic is not alone in this pursuit. Meta and OpenAI are also working on proprietary chip designs as the tech industry moves to reduce its dependence on third-party suppliers.

Meta, meanwhile, has separately signed a $21 billion cloud infrastructure deal with CoreWeave, the Nvidia-powered data centre company, on top of a prior $14.2 billion agreement. Meta's total commitment to CoreWeave now stands at $35.2 billion.

Anthropic's Cybersecurity Ambitions

Beyond chips, Anthropic has also been expanding its footprint in cybersecurity. The company recently launched Project Glasswing, an initiative that brings together major technology firms and critical infrastructure organisations to test an advanced, unreleased AI model, capable of identifying software vulnerabilities at scale. The AI model named 'Claude Mythos Preview' already identified thousands of zero-day vulnerabilities. It supposedly uncovered critical flaws across major operating systems and browsers, including a 27-year-old bug in OpenBSD, a 16-year-old issue in FFmpeg, and a chain of Linux kernel vulnerabilities that could enable full system control.

The project will focus on practical security applications. Anthropic earlier said partners will use the model for tasks such as local vulnerability detection, black-box testing of binaries, endpoint security and penetration testing of foundational systems.

The broader goal is to help secure a significant portion of the global cyberattack surface.

Anthropic is supporting the initiative with funding, credits, and expanded access. The company said it will provide up to $100 million in model usage credits for Project Glasswing and related efforts, along with donations to open-source security groups.