Hardware & InferenceBusiness & Market 🇷🇺 05.08.2026 19:04

Anthropic Begins Assembling Team to Develop Its Own AI Chips

AnthropicAnthropic OpenAIOpenAI Google DeepMindGoogle DeepMind MetaMeta
Anthropic has started forming a department to develop its own AI accelerators, aiming to design models and hardware in parallel for fast and efficient operation, according to Business Insider. The company is reportedly considering Samsung as a manufacturing partner. Anthropic seeks to scale beyond partnerships with AWS, Google, Nvidia, and AMD to meet demand for its Claude models.
Anthropic has begun forming a team to develop its own AI accelerators, intending to develop models and hardware in tandem to ensure fast and efficient operation, as reported by Business Insider. It was previously known that Samsung is a candidate to manufacture these chips. Anthropic decided to create its own accelerators due to high demand for Claude models. The company has signed agreements with AWS, Google, Nvidia, and AMD for access to computing hardware, but these partnerships are insufficient to truly scale operations and fully meet service demand. Anthropic is not the first AI lab to pursue custom silicon: in June, OpenAI introduced the Jalapeño accelerator, developed with Broadcom, for AI inference workloads; Google DeepMind has long used TPU accelerators, and Meta is preparing to produce its own MTIA chips. Anthropic is seeking engineers with chip design experience for its 'custom silicon design' department. No official comments from the company have been made.
Abbreviations
AWS = Amazon Web Services — Amazon Web Services
TPU = Tensor Processing Unit — тензорный процессор
MTIA = Meta Training and Inference Accelerator — ускоритель обучения и вывода Meta
Source: 3DNews — original
Our earlier posts on this topic ↓
Fresh news