The new AWS Neuron 2.31.0 release includes NKI 0.5.0 with MX FP8 scale dtype support and tensor indirection for more efficient compute operations. It also launches the Neuron UltraServer Operator for Amazon EKS in public beta, automating workload allocation for Trainium UltraServer deployments.
AWS is improving the performance and operational efficiency of its Trainium AI accelerators, making them more competitive for large-scale AI training on EKS.