Skip to main content
Globe And Street

The world, accounted for.

Global business and finance

SectionsMenuClose

Google adds AI chip variants for secure and data-heavy workloads

The 8tx uses AMD processors alongside Google's chip technology, while the 8i Max packages two chips together for more memory and computing power.

Illustration of processors on a circuit board inside a data center
Photo: CNA

Alphabet's Google introduced the 8tx and 8i Max on November 9, expanding its tensor processing unit lineup with designs for customers handling sensitive data or information-intensive AI tasks. The launch adds to Google's offerings in a chip market dominated by Nvidia.

Google said both variants rework existing processors. The 8tx targets public agencies and regulated businesses whose information needs protection during processing. It couples AMD processors with Google's chip technology and uses hardware-level data encryption.

The company said the 8tx can run at up to twice its predecessor's speed, with virtually no performance trade-off. For the 8i Max, Google places two chips in one package, increasing memory and computing capacity for workloads such as coding assistance, mathematical reasoning and drug discovery.

Google has used TPUs mainly in its own Search, YouTube and Gemini products, while also offering them through its cloud operation to attract enterprise AI customers. Alphabet said TPU sales contributed meaningfully to September-quarter results, as Google Cloud revenue more than doubled to $30.71 billion.

Alphabet's stock gained more than 3% during early trading on November 9.

Editorial policyReport a mistake

About the author

Hao Chen

Technology Editor

Edits and reviews technology business stories: artificial intelligence, chips, big tech, software, cybersecurity incidents, start-ups and platform regulation.