Google expands hardware offerings with chips for secure data and large contexts

Google has announced two custom microchips aimed at helping businesses run complex artificial intelligence programs. As Google executive Amin Vahdat shared on the company's official blog, the new hardware models, named TPU 8tx and TPU 8i Max, cater to specialized cloud workloads. The TPU 8tx prioritizes hardware-based protection for confidential data, while the TPU 8i Max packs high computational power into single server setups to help AI agents process huge documents and video files.

These additions expand the tech giant's eighth generation of custom processors, bringing its total to four specialized chips released within a single year. By designing chips tailored to specific tasks instead of relying solely on general purpose graphics hardware, cloud providers can lower power consumption and accelerate response times. For ordinary consumers, these cloud infrastructure improvements generally lead to faster, more capable online services and digital assistants.

However, it remains uncertain whether custom cloud chips will convince a majority of developers to pivot away from industry standard Nvidia hardware. Furthermore, as corporate AI tools gain access to broader sets of private data, an open question is whether hardware-level security alone will be enough to prevent complex safety leaks in multi-step AI workflows.