Anthropic is in early discussions with Samsung Electronics about producing a custom artificial intelligence chip, according to a report from The Information. The project remains in a preliminary phase, with no detailed design or final specifications yet determined. Anthropic is still working to define what the chip would do and how much computing power it would need.
Company downplays chip effort
Anthropic played down the significance of the reported initiative. The company told The Information that chips from AWS, Google, and Nvidia remain central to its strategy. Anthropic declined to comment on any chip roadmap of its own.
Despite the cautious public stance, there are clear signs that a custom chip project is underway. Anthropic has hired chip engineers, including Clive Chan, an early member of both Tesla's and OpenAI's custom chip teams. Chan is expected to build out a dedicated chip group at Anthropic.
Part of a broader industry trend
The move follows a pattern seen across the AI industry. Companies that can build and run AI infrastructure more cheaply are able to keep more of their revenue. Custom chips offer a path to that efficiency.
OpenAI recently unveiled "Jalapeño," its first in-house inference chip, built in collaboration with Broadcom. AWS, Google, and Meta all run custom silicon designed for AI workloads. AWS uses Trainium and Inferentia chips for training and inference, Google relies on its Tensor Processing Units (TPUs), and Meta uses its MTIA family of chips.
Stay updated
Get the day's AI and automation news in your inbox. No spam, unsubscribe anytime.
Background on the companies
Anthropic was founded in 2021 by former OpenAI employees, including siblings Dario and Daniela Amodei. The company focuses on developing large language models and safety research. Its Claude model family competes with OpenAI's GPT series and Google's Gemini.
Samsung Electronics is one of the world's largest semiconductor manufacturers, operating a foundry business that produces chips for external clients. The company has been investing in advanced fabrication processes to compete with TSMC.
Nvidia remains the dominant supplier of AI training chips, particularly its H100 and B100 series GPUs. Any custom chip from Anthropic would likely be aimed at inference rather than training, given the complexity of competing with Nvidia in that area.
What comes next
The talks with Samsung are reportedly still at an early stage. No production timeline or volume commitments have been disclosed. The outcome will depend on design decisions, manufacturing costs, and the broader availability of foundry capacity.
For now, Anthropic's position is that its chip efforts, if they materialize, will complement rather than replace its use of existing hardware from Nvidia, AWS, and Google.
Related on Neura Market: