Rumor
Google Reportedly Accelerating Deployment of NVIDIA GB200 NVL72 Platform
Tuesday, December 2, 2025 at 02:57 AM
A supply chain watcher claims Google is ramping up the deployment of NVIDIA's new GB200 NVL72 AI infrastructure alongside its continuing investment in its in-house Tensor Processing Units (TPUs).
Context
Supply-chain chatter and official announcements indicate Google is accelerating its deployment of Nvidia’s new GB200 NVL72 platform, a strategic move to supplement its in-house TPU accelerators. In February 2025, Google Cloud announced the preview of its A4X virtual machines, becoming the first cloud provider to offer instances based on the powerful Nvidia GB200 NVL72 system, with general availability following in May 2025. This platform integrates 72 Blackwell GPUs and 36 Grace CPUs into a single liquid-cooled rack, promising up to a 30x performance increase in real-time LLM inference and a 4x boost in training compared to the previous H100 generation.
This early adoption by a major hyperscaler like Google provides significant validation for Nvidia's Blackwell architecture ahead of its broader rollout. High-volume shipments of GB200 NVL72 cabinets are scheduled to begin in Q2 2025, with analyst forecasts for the year revised to between 25,000 and 35,000 units. Google's commitment signals strong initial demand for these high-performance systems, which is a positive indicator for Nvidia and its supply chain partners specializing in components like liquid cooling and high-bandwidth memory.
Sources (26)
Nvidia Suppliers Send Mixed Signals for Delays on GB200 SystemsNVIDIA Accelerates OpenAI gpt-oss Models Delivering 1.5 M TPS ...Enabling Multi-Node NVLink on Kubernetes for NVIDIA ...GB200 NVL72 | NVIDIAUS technology stocks surged on Thursday after Nvidia ...GB200 Hardware Architecture - Component Supply Chain ...Nvidia Blackwell, Google TPUs, AWS Trainium: Comparing top AI ...Accelerating Reinforcement Learning on Google Cloud using ...
Related Companies
Nvidia
NVDA
Google
GOOGL