Click to Play — Analysis of the GTC Announcements
NVIDIA is transforming from a chip designer into an infrastructure orchestrator. The GTC announcements signal that the company is using its balance sheet to remove supply-side uncertainty for the entire AI stack.
The core of the strategy is a series of binding capacity agreements that extend from substrate manufacturing to advanced packaging and final testing. By providing multi-billion dollar prepayments to partners like Ibiden, Amkor, and KYEC, NVIDIA is effectively locking in production capacity years ahead of demand.
NVIDIA confirmed that the Vera Rubin platform is now in full production, delivering approximately 30x faster performance on large language model workloads compared to Blackwell. The first major cloud deployment is imminent.
NVIDIA's next-gen platform is being manufactured at scale, with supply chain partners ramping capacity.
Google Cloud will launch A5X cloud instances based on Vera Rubin NVL72, with plans to scale to up to 960,000 Rubin GPUs across multiple sites.
The real revenue test begins as Vera Rubin systems enter mass deployment across hyperscalers and enterprises.
The first named A5X customer is Ineffable Intelligence, a London-based startup founded by David Silver (creator of AlphaGo), which will use the platform for large-scale reinforcement learning research. This signals that Vera Rubin's initial demand is coming from frontier AI labs, not just general cloud workloads.
The GTC announcements have ripple effects across the semiconductor and cloud industries: