Google Cloud has announced the general availability of Ironwood TPUs, its seventh-generation Tensor Processing Units, and new Axion VMs, marking a major leap in AI infrastructure performance and efficiency. Ironwood, designed for high-demand workloads such as large-scale model training and low-latency inference, delivers 10x higher peak performance than TPU v5p and 4x greater efficiency than TPU v6e, positioning it as Google’s most powerful and energy-efficient silicon to date. Early adopters like Anthropic, Lightricks, and Essential AI are leveraging Ironwood to scale advanced AI models and enhance generative capabilities.
Also Read: ServiceNow and NTT DATA Deepen Partnership to Accelerate Global AI-Led Enterprise Transformation
“Ironwood’s improvements in both inference performance and training scalability will help us scale efficiently while maintaining the speed and reliability our customers expect,” said James Bradbury, Head of Compute at Anthropic. Complementing Ironwood, Google unveiled Arm-based Axion instances—N4A, offering up to 2x better price-performance, and C4A metal, its first bare-metal Arm instance for specialized workloads. Together, Ironwood and Axion strengthen Google Cloud’s AI Hypercomputer ecosystem, combining hardware-software co-design, advanced networking, and liquid cooling to deliver unprecedented scalability, reliability, and cost efficiency for next-generation AI and general-purpose computing workloads.