Anthropic Forms Chip Team to Cut Claude's Compute Costs

A New Push Into Custom Silicon

Anthropic is assembling an in-house chip design team dedicated to building custom silicon for its Claude family of AI models, according to reports confirming the company's expansion into hardware engineering 12. The move signals a deliberate shift away from total reliance on third-party chipmakers as Anthropic looks to secure more direct control over the infrastructure that powers its models.

Why Anthropic Wants Its Own Chips

The rationale behind the initiative centers on compute scarcity and cost efficiency. Building proprietary chips could help Anthropic reduce the compute constraints that have become a persistent bottleneck for AI labs racing to train and serve increasingly large models 2. Custom silicon tailored specifically to Claude's architecture could also improve processing speed and scale efficiency, allowing the company to run inference and training workloads more economically than relying solely on off-the-shelf accelerators 1.

This is a familiar playbook in the AI industry, where the cost of renting or purchasing general-purpose GPUs has become one of the largest line items for frontier labs. By designing chips in-house, Anthropic could better tailor hardware to the specific mathematical operations and memory patterns that Claude models rely on, potentially squeezing more performance out of every dollar spent on compute.

Hiring at a Premium

To build out this new hardware division, Anthropic is reportedly offering salaries as high as $485,000, an aggressive figure that underscores how fiercely AI companies are competing for scarce chip design talent 1. Engineers with experience in semiconductor architecture, particularly those who have worked on AI accelerators, are in high demand across the industry, and top labs are increasingly willing to pay premiums typically associated with elite machine learning researchers rather than hardware engineers.

Possible Manufacturing Ties to Samsung

Alongside the hiring push, Anthropic is said to be in early discussions with Samsung about potential manufacturing arrangements 1. While details remain limited, any partnership would place Anthropic alongside other major AI players that have sought direct relationships with chip foundries rather than depending exclusively on intermediaries. Such a tie-up would also reflect Samsung's broader ambitions to expand its foundry business by courting AI companies eager for dedicated fabrication capacity.

What It Means for Claude's Future

Taken together, the reports point to Anthropic preparing for a future in which its own hardware, not just partnerships with existing chip suppliers, underpins the scaling of Claude models. If successful, custom silicon could translate into faster response times, lower operating costs, and greater independence from the volatile supply chains that have constrained AI compute availability. The effort remains in its early stages, but it marks a notable step for Anthropic as it moves beyond software and model development into the far more capital-intensive world of chip design.