According to Jeff Dean on X, Google introduced TPU 8t for large-scale training and inference with a pod size of 9,600 chips delivering about 121 exaflops FP4 per pod, roughly 3X the FP4 performance of Ironwood’s 42.5 exaflops per pod (as reported in Dean’s April 23, 2026 post). According to Jeff Dean, the FP4-focused uplift targets high-throughput inference and frontier model training, signaling lower cost per token and faster time-to-train for multi-trillion parameter workloads. As reported by Jeff Dean, the pod-level scaling implies denser datacenter footprints and higher utilization for Google Cloud customers building LLMs and VLMs, creating business opportunities in model serving, batch inference, and fine-tuning at scale.

According to Satya Nadella on Twitter, the new Maia 200 AI accelerator is now available on Azure, offering industry-leading inference efficiency and delivering 30% better performance per dollar compared to existing systems. With over 10 PFLOPS FP4 throughput, approximately 5 PFLOPS FP8, and 216GB HBM3e memory with 7TB/s bandwidth, Maia 200 is optimized for large-scale AI workloads. As reported by Satya Nadella, this addition expands Azure’s portfolio of CPUs, GPUs, and custom accelerators, providing customers with more options to execute advanced AI workloads faster and more cost-effectively.