Anthropic is reportedly looking more closely at the hardware behind its AI systems. According to The Information, the company is in talks with Samsung Electronics about manufacturing a custom AI chip, even as it continues to emphasize the importance of existing suppliers.
An early-stage chip effort
The reported project is not yet a finished plan. The chip does not have a detailed design, and Anthropic is still working through basic questions about what the component would do and how powerful it would need to be.
That matters because a custom AI chip is not just another procurement choice. It would require Anthropic to define a workload, decide where the chip fits in its infrastructure, and determine whether the benefits justify building a dedicated hardware path.
For now, the company is keeping the effort in perspective. Anthropic told The Information that chips from AWS, Google, and Nvidia remain central to its strategy. It also declined to comment on any chip roadmap of its own.
Why Samsung Electronics matters
The reported talks with Samsung Electronics point to the manufacturing side of the question. Anthropic is not described as having a completed chip ready to build. Instead, the source article says the company is exploring manufacturing while the design and performance goals remain unresolved.
That distinction is important. A company can investigate production options before it has settled every technical detail. In practical terms, those conversations can help shape what is possible, what tradeoffs may exist, and how a future chip effort might be organized.
Still, the source does not say that Anthropic has committed to producing a chip with Samsung Electronics. It says the discussions are underway and that the project remains early.
Signals inside Anthropic
Even with the company playing down the effort, there are signs that chip work is becoming more than a casual idea. Anthropic has brought on chip engineers, including Clive Chan, an early member of both Tesla's and OpenAI's custom chip teams.
Clive Chan is expected to build out a dedicated chip group at Anthropic. That does not prove that a specific product is imminent, but it does suggest the company is preparing to evaluate custom silicon with more internal expertise.
For an AI company, that expertise can matter before a chip is ever built. Engineers with custom chip experience can assess whether a design target is realistic, how a chip might support AI workloads, and where existing chips still make more sense.
The economics behind custom AI chips
The broader logic is straightforward: whoever can build and run AI infrastructure more cheaply keeps more of the revenue. The source article frames custom chips as one way companies try to reach that goal.
AI infrastructure is expensive, and the hardware layer can shape the economics of running models. If a company can tune silicon to its own workloads, it may be able to reduce reliance on general-purpose options or improve the efficiency of specific tasks.
That is the strategic appeal, but it does not remove the risk. A custom chip only helps if it matches the company's actual needs. Anthropic is reportedly still deciding what its chip would do, which means the most important technical and business questions remain open.
A wider industry pattern
Anthropic would not be alone in examining custom silicon. The source article notes that OpenAI recently unveiled "Jalapeño," its first in-house inference chip, built with Broadcom.
AWS, Google, and Meta also run custom silicon tuned for AI workloads. That places Anthropic's reported interest within a broader shift among major AI and technology companies: hardware is becoming part of the competitive strategy, not just a supply chain issue.
At the same time, Anthropic's public posture is cautious. The company is not presenting a custom AI chip as a replacement for AWS, Google, or Nvidia chips. Based on the source, the more accurate picture is a company exploring its options while continuing to depend on established AI hardware providers.
That balance may define the next phase of Anthropic's hardware strategy. A custom chip could eventually give the company more control over cost and performance, but the reported effort is still at the stage of questions, hiring, and manufacturing discussions rather than a defined product launch.