The first NVIDIA GB300 NVL72 systems, powered by Blackwell Ultra GPUs, began shipping in December 2025, with major cloud providers like Microsoft, CoreWeave, and others securing initial allocations. Volume production ramped in Q4 2025, prioritising hyperscalers driving AI reasoning workloads.What is the NVIDIA GB300 NVL72?The NVIDIA GB300 NVL72 is a liquid-cooled, rack-scale platform integrating 72 Blackwell Ultra GPUs and 36 Grace CPUs. It delivers 1.5 times more AI performance than the standard Blackwell GB200, with up to 1.1 exaFLOPS of FP4 compute, making it ideal for advanced AI inference and reasoning tasks.Who Are the Key Recipients of Early GB300 NVL72 Shipments?Early recipients include leading cloud and AI infrastructure providers focused on scaling generative AI:Microsoft: Secured one of the largest initial allocations for Azure AI expansions.CoreWeave: Received the first full racks via Dell, deploying GB300-powered instances for customers.Other hyperscalers: AWS, Oracle, and specialised providers like Lambda accessed units as volume shipments began in Q4 2025.Sources: NVIDIA announcements, Dell press releases, and industry reports from SemiAnalysis and Data Center Dynamics.How Will GB300 NVL72 Volume Shipments in Q4 2025 Impact Cloud Pricing?Volume shipments of the GB300 NVL72 starting in Q4 2025 are expected to increase AI compute supply, potentially stabilising or reducing cloud GPU rental prices over time. Higher efficiency (up to 50% more performance per rack) could lower operational costs for providers.Key impacts include:Improved availability: Easing shortages seen with Hopper and early Blackwell systems.Competitive pressure: Encouraging providers to pass efficiency gains to customers through lower per-hour rates.Long-term price moderation: As supply ramps, hourly costs for advanced AI inference may decline, though initial premium pricing reflects high demand.What Are the Benefits of GB300 NVL72 for Cloud AI Workloads?The platform offers significant advantages for cloud-based AI deployment:Superior reasoning performance: Up to 10x faster user responsiveness for agentic AI.Energy efficiency: 5x better throughput per watt compared to Hopper.Scalability: Full NVLink domain acts as a single massive GPU for trillion-parameter models.Higher memory capacity: Supports larger models and batch sizes.Why Is the GB300 NVL72 a Game-Changer for Cloud Availability in Sydney and NSW?With growing demand for AI infrastructure in Australia, increased GB300 NVL72 availability through global cloud providers will enhance access to cutting-edge compute for NSW businesses and researchers in Sydney. Local data centres can leverage these systems for faster AI development.Follow us on X @ElymentGroup for the latest in technology infrastructure updates, or visit elyment.com.au.Secure Your AI Edge Today