Google's Triggerfish TPU v9 generation deepens its proprietary AI silicon stack — faster RL and agent inference in-house means lower compute costs and less reliance on third-party chips as AI workloads scale.
Google's TPU v9 Triggerfish upgrades memory from HBM4 to HBM4E, making SK Hynix — the dominant HBM supplier — a direct beneficiary of the new incremental chip order.