Anthropic confirms in-house AI chip team for Claude
Anthropic s working on putting together an internal silicon team to design chips for Claude
Anthropic has confirmed it is building an in-house team to design custom AI chips for its Claude models, marking a major step in the company's efforts to develop its own computing infrastructure as demand for generative AI continues to surge.
Following earlier reports on the development of custom semiconductors by the AI startup, the firm has finally confirmed its intentions publicly for the first time.
Anthropic expands into AI chip design
An Anthropic spokesman told Business Insider that the firm is working on putting together an internal silicon team to design chips for Claude.
This information also came out in the recent recruitment process, wherein the company is hiring engineers for a "custom silicon team". The requirement is that the engineer must have experience of moving semiconductor designs from concept to production, reflecting Anthropic’s intention of creating its own commercial-grade AI hardware.
The post pays between $320,000 and $485,000.
It is becoming more appealing for companies specialising in AI to develop custom AI chips as a means to cut down on their reliance on Nvidia, whose graphic chips dominate the AI accelerator market.
By controlling the hardware, these companies can optimise their own AI chips, which will work better in the future at a lower cost of hardware.
On the other hand, developing semiconductors is among the costliest tasks in the industry.
Jalapeño, which OpenAI has specifically built for AI, was revealed by OpenAI in June in collaboration with Broadcom. Meanwhile, Meta is also gearing up to launch its next-generation AI chip, and it appears that Mistral is considering designing its own chip.
It has been previously revealed through Reuters' reporting that Anthropic was exploring the option of building its own chips, and according to information from The Information, it was discussing manufacturing collaborations with Samsung Electronics.
-
Switzerland’s new registry faces cyberattack fears ahead of launch
-
Google Messages users report old private texts being sent by mistake
-
3 reasons Apple Pencil won't come to iPhone Ultra
-
Think your company is innovating with AI? You may be wrong
-
Temu ads used 73% fake influencers, research finds
-
EU designates ChatGPT, Reddit and Roblox for tougher new rules
-
AI sentiment turns negative at work, says Glassdoor study
-
AI cyber risk becomes global finance’s biggest threat