Cerebras unveils AI chip for chatbots; TSMC fabricates
Cerebras Systems introduced new server hardware, including its WSE-3 Turbo chip, to accelerate AI chatbot queries. Taiwan Semiconductor Manufacturing Company (TSMC) produces the large chips.
Cerebras targets AI inference with new system
Cerebras Systems, a US-based AI hardware firm, launched new server hardware designed to speed AI chatbot queries. The company introduced its CS-4 system on Tuesday, featuring dinner-plate-sized chips. These chips target AI inference, the computing process that generates responses in chatbots like Anthropic’s Claude.
Cerebras competes with Nvidia in this specialized AI hardware segment. The CS-4 system uses three of Cerebras' large chips, which the company states improves performance.
TSMC fabricates 5-nanometer chips
The CS-4 system uses Cerebras' Nexus server architecture. This design incorporates pluggable modules that house the large chips. Cerebras' chips gain a speed advantage from their size, avoiding energy consumption and slowdowns associated with moving data between multiple smaller chips.
Taiwan Semiconductor Manufacturing Company (TSMC) fabricates these chips using its 5-nanometer manufacturing process. The new CS-4 machine became available in the third quarter of this year.
Efficiency and future throughput goals
Cerebras also designed the CS-4 for easier data centre setup. Chief Technology Officer Sean Lie said at a San Francisco media briefing that the system has 50 per cent fewer components, speeding data centre construction. The server rack includes the WSE-3 Turbo chip and new networking components. Mr Lie stated these components accelerate data movement between chips.
Cerebras plans another chip and server generation in 2027. CEO Andrew Feldman expects the company to deliver 600 megawatts of computing power by end-2027. Mr Feldman added that future chips will be four times faster and offer 20 times more throughput by the end of 2027.
Asian data centres may see cost reductions
Last week, Cerebras reported an adjusted loss of $6.9 million on sales of $180.1 million. This new hardware supports the growing demand for AI infrastructure. For Asian economies, the reliance on TSMC for fabrication shows Taiwan's central role in advanced chip production.
The system's easier setup and faster processing could drive new data centre investments across Asia, particularly as regional firms expand AI capabilities. This could reduce operational costs for Asian companies deploying large-scale AI models.
This article is journalism, not investment advice; consult a licensed professional before making financial decisions. Market data is indicative, may be delayed, and should be verified with your broker or exchange before use.
Comments.
Comments are moderated. We remove what is unlawful, abusive or off-topic, and and you remain responsible for what you post.
Reader comments open soon. Until then, corrections and responses go to our newsroom, and we publish what we get wrong on Corrections.