Google adds chips for secure and demanding AI workloads

Google announced TPU 8tx and TPU 8i Max on Monday, expanding its processor lineup for customers handling sensitive information or demanding AI workloads. Both adapt existing tensor processing units rather than introduce an entirely new processor design. Customers are expected to gain access in 2027.

The 8i Max combines two chips in one package to increase memory capacity and computing resources for applications that process substantial amounts of information simultaneously. Google identified coding assistance, mathematical reasoning and drug discovery as suitable uses. The 8tx combines Google chip technology with AMD processors and hardware-based encryption to protect data during processing. Designed for public-sector agencies and regulated businesses, it can run as much as twice as fast as its predecessor, Google said, with almost no performance penalty from the protection.

Google’s processors reportedly cost about 30% below Nvidia’s while offering comparable performance and using less electricity for some AI workloads. Google Cloud is extending their reach beyond internal services such as Gemini, Search and YouTube. Alphabet said TPU sales contributed meaningfully to its September-quarter results, when Google Cloud revenue reached $30.71 billion, more than twice the year-earlier level.