Post by CoreWeave

As AI models scale toward trillions of parameters, the infrastructure that powers them must be as flexible as it is powerful. Today, live from #GoogleCloudNext, we are excited to announce cross-cloud AI capabilities designed to simplify training and inference across clouds. We’re introducing three critical capabilities to simplify the AI-native lifecycle: 🔹 CoreWeave Interconnect: A private, high-performance fiber connection directly to Google Cloud with MACsec encryption. No complexity—just low-latency, secure connectivity at line rate. 🔹 SUNK Anywhere: Simplifies cross-cloud AI training by creating a unified control plane for scheduling, scaling, and running distributed workloads across clouds supporting Google Cloud, Amazon Web Services (AWS), or Microsoft Azure without changing your operational model. 🔹 LOTA™ Cross-Cloud: Achieve near-local data throughput (up to 7 GB/s per GPU) across regions and clouds. Store your data where you want, and run your workloads where they perform best – all with zero egress fees. We’ve also deepened our integration with Weights & Biases and Google Cloud adding native support for Gemini CLI, integration with Gemma, and comprehensive TPU performance metrics to give developers unprecedented visibility into their agentic AI workflows. The era of "multi-cloud by necessity" is evolving into cross-cloud by design. Read the full announcement on our blog: https://hubs.la/Q04d1zRm0 Cohere Periodic Labs #GoogleCloudNext