Anthropic, the developer behind the Claude family of AI models, is in talks with Samsung Electronics to manufacture a custom artificial intelligence chip, according to multiple reports. The discussions, still in their early stages, reflect a growing push by top AI labs to move beyond off-the-shelf hardware and tailor chips to their specific workloads.
Early-stage discussions center on 2nm technology
The project remains exploratory. Sources cited by Bloomberg said Anthropic is evaluating Samsung Foundry's 2nm process technology and advanced packaging capabilities, though the company has not yet defined exact performance targets, server integration plans, or begun detailed design work. Anthropic is also in talks with several chip design firms, indicating the supply chain is still being assembled.
Existing Samsung relationship provides a foundation
The reported discussions build on a deeper tie. In May, during Anthropic's Series H funding round, the AI firm named Samsung Electronics, SK hynix, and Micron as strategic infrastructure partners. Samsung's unique position as both a memory and logic chip foundry makes it a natural candidate for manufacturing any future Anthropic-designed processor. Industry observers have long speculated about such a partnership given Samsung's ability to supply both memory and custom logic.
Custom silicon becomes an AI industry priority
Anthropic's interest in proprietary chips follows a wider pattern. OpenAI recently partnered with Broadcom to launch a custom inference processor called Jalapeño, which the company claims offers better performance-per-watt than competing hardware. Anthropic itself has been building in-house silicon expertise - it recruited Clive Chan, who previously helped establish OpenAI's custom chip initiative. Despite these moves, the company has signalled that its core computing strategy will continue to lean on Amazon Trainium, Google TPUs, and NVIDIA GPUs for the immediate future.
Why this matters for IT and Development
Custom AI chips shift how infrastructure teams think about workload placement, cost, and performance tuning. As hyperscale AI firms design their own processors, IT and development professionals will need to adapt their deployment and optimization practices for hardware that may differ significantly from traditional GPU clusters. For those working directly with Anthropic's models, structured Claude AI Training can sharpen skills in using and integrating the technology. Gaining cross-stack knowledge through AI for IT & Development training can help teams bridge the gap between emerging silicon and production AI workloads.
Your membership also unlocks:
