Anthropic is in talks with Samsung Electronics regarding the production of custom AI chips, involving Samsung's 2nm process and packaging business, further adding competitive pressure on Nvidia's stock; Nvidia currently holds approximately 74% of the AI chip market share. Anthropic has hired Clive Chan, a former member of OpenAI's early custom chip team, but the project remains in early stages, with final design and manufacturing not yet initiated.
Anthropic in talks with Samsung for chip production, primarily using 2nm process
According to reports, Anthropic's discussions with Samsung focus on Samsung's 2nm process and packaging business; the packaging business is responsible for connecting different components of the processor, a critical step in advanced chip manufacturing. Anthropic has recruited Clive Chan, a former member of OpenAI's early custom chip team, reflecting the company's systematic efforts to build out its chip development team.
These discussions are still in an early exploratory phase, with no final design or manufacturing work officially started; Anthropic has not disclosed specific details of its chip development roadmap.
Anthropic's multi-compute strategy: six platforms retaining architecture and the positioning of Samsung custom chips
Anthropic told The Information, "Amazon AWS Trainium chips, Google's Tensor Processing Units (TPUs), and Nvidia's GPUs will remain core to our compute scaling strategy."
Anthropic is also considering using Microsoft (NASDAQ:MSFT) hardware and chips from UK startup Fractile, with the custom Samsung chip as a potential supplement to its diverse compute options. Anthropic's goal in seeking multiple options is:
Nvidia GPUs: existing core compute infrastructure
Google TPUs: tensor processing acceleration
AWS Trainium: Amazon's self-developed AI training chip
Microsoft hardware: enterprise cloud computing
Fractile chips: UK startup's inference chips
Samsung custom processor (if talks are completed): potential 2nm process option
AI chip competitive landscape: Nvidia's market share at 74%
Nvidia (NVDA) currently holds approximately 74% of the AI chip market, while multiple tech giants are accelerating their in-house chip development: Alphabet has TPUs, Amazon has Trainium, Meta Platforms (NASDAQ:META) and Microsoft are each developing their own chips; OpenAI partnered with Broadcom (NASDAQ:AVGO) in 2024 to design the Jalapeño inference chip, aimed at running large language models more efficiently.
Taiwan Semiconductor Manufacturing Co. (NYSE:TSM) remains a leader in advanced chip manufacturing, while Samsung continues to face challenges in yield rates as it catches up to TSMC's N2 process. Additionally, Samsung Group and SK Group recently announced plans to invest $518 billion over the next decade to build four memory chip factories in South Korea, with both SK Hynix and Samsung participating in Anthropic's latest funding round.
Frequently Asked Questions
Why is Anthropic seeking Samsung to manufacture custom AI chips instead of continuing to use Nvidia?
According to reports, Anthropic's goal is to establish diverse compute options to reduce reliance on any single supplier; the talks regarding Samsung's 2nm process and packaging business are early exploration, with Nvidia GPUs explicitly listed as core to its compute scaling strategy, and custom chips as a potential supplement rather than a replacement. Specific decisions are subject to Anthropic's official announcements.
What is Nvidia's current share of the AI chip market, and is it affected by this news?
Nvidia (NVDA) currently holds approximately 74% of the AI chip market. On Thursday's close, Nvidia stock rose 0.7%, Broadcom (AVGO) rose 0.7%, and TSMC (TSM) rose 3.5%; specific stock movements are influenced by multiple market factors, subject to real-time trading data.
What are Samsung's current positioning and challenges in the AI chip market?
Samsung Electronics (KS:005930) has a 2nm process and packaging business and participated in Anthropic's $65 billion funding round in May, with Google also considering adding Samsung to the supply chain for future TPU components. However, Samsung continues to face technical challenges in yield rates as it catches up to TSMC's N2 process; Samsung Group also announced a $518 billion investment over the next decade to build four memory chip factories in South Korea.
