Google Cloud announced on Friday the preview of A4 VMs powered by NVIDIA's new Blackwell B200 GPUs. The A4 VMs promise a significant performance boost, featuring eight interconnected GPUs and offering 2.25 times the compute power and HBM capacity of the previous A3 VMs.
The new cloud infrastructure aims to enhance AI model training, fine-tuning, and low-latency serving. The A4 VMs integrate Google’s infrastructure innovations, including enhanced networking with Titanium ML adapters, Google Kubernetes Engine support, and access through Vertex AI.
"We're looking forward to leveraging the innovations in Hypercompute Cluster to accelerate deployment of training our latest models that deliver quant-based algorithmic trading," Gerard Bernabeu Altayo, Compute Lead, Hudson River Trading.