Google has expanded its eighth-generation hardware lineup with two new configurations designed to handle massive context windows and secure inference tasks.

Key Developments and Technical Context

Google has added two new systems named TPU 8i Max and TPU 8tx to its eighth-generation hardware lineup.

The TPU 8i Max system combines two TPU 8i chips into a single package to double compute, SRAM, and high-bandwidth memory performance.

The TPU 8tx system pairs TPU 8t silicon with fifth-generation AMD EPYC host processors to target secure inference workloads.

Strategic Analysis and Industry Implications

Combining multiple dies into a single package increases compute density within existing rack footprints without requiring a complete redesign of the base silicon architecture.

Integrating specialized host processors highlights a hardware-level focus on security, helping regulated organizations manage sensitive data more safely during inference.

Outlook and Key Factors to Watch

Google has not published absolute technical specifications, benchmarks, pricing, or regional availability for either new system configuration.

The exact high-bandwidth memory capacity of the TPU 8i Max system remains undisclosed in the official announcements.