AWS and Azure launch multicloud interconnect preview; AWS expands NVIDIA GPU deal

AWS entered public preview with AWS Interconnect–multicloud with Microsoft Azure, enabling private, encrypted Layer-3 connectivity between the two hyperscalers across US East, US West, Sydney, and Frankfurt, with paths to 100 Gbps at general availability. The rare AWS-Azure cooperation acknowledges enterprise reality: most large customers run workloads across multiple clouds and want secure, low-latency links between them rather than routing over the public internet.

Mechanically, the interconnect provides dedicated private paths for cross-cloud traffic — critical for AI workloads that may train on one cloud and serve on another, or that split data-residency requirements across regions. Both AWS and Azure entered the preview simultaneously, signaling mutual demand.

The more strategically significant number: AWS expanded its NVIDIA partnership to deploy an additional 2 million GPUs across its infrastructure in 2027-2028, deepening collaboration on CPUs, networking, and secure 'AI factories.' In a week where compute scarcity is the dominant constraint — OpenAI rationing subscriptions, NVIDIA touting 400K new GPUs — locking in 2 million more GPUs is a major capacity commitment.

Competitively, AWS is positioning to remain the default AI-infrastructure provider even as it hosts rival models (GPT-6 Astra, Claude) on Bedrock. It was also named a Leader in the 2026 Gartner Magic Quadrant for Strategic Cloud Platform Services for the 16th consecutive year. Watch the interconnect's GA timeline and whether the 2-million-GPU commitment materializes on schedule given global supply constraints.