Next-generation AI accelerator clusters open for inference workloads

Summary

Cloud vendors are pitching lower token costs as they compete for LLM inference and real-time multimodal workloads.