Nvidia CEO Jensen Huang used his CES 2026 keynote to announce that the company’s Vera Rubin AI platform has entered full production, with shipments to partners planned for the second half of 2026. The platform, which pairs new Rubin GPUs with Vera CPUs, represents Nvidia’s next architectural leap beyond Blackwell.

Five minutes flat

Nvidia tackled assembly bottlenecks by introducing a modular, cable-free tray design for Vera Rubin racks that slashes assembly time by roughly 95%. What previously required a skilled technician spending two hours carefully connecting components now snaps together in about five minutes.

Microsoft Azure is already among the partners testing the first Vera Rubin racks, according to the announcement.

A supply chain spanning 30 countries

Nvidia has doubled its supply chain size for Vera Rubin. The production network now spans more than 350 factories across 30 countries, with a notable concentration of 150 factories in Taiwan. The platform is built on TSMC’s advanced 3nm fabrication process.

Vera Rubin uses next-generation HBM4 memory supplied by Samsung and SK Hynix.

Performance claims and competitive positioning

Nvidia is claiming up to 3.5 times improvement in AI training performance and 5 times improvement in inference compared to the Blackwell architecture.

Nvidia isn’t just selling GPUs anymore. It’s selling entire systems, from the processor to the networking fabric to the physical enclosure.

What this means for investors

Nvidia’s ability to scale production while simultaneously improving assembly efficiency addresses the single biggest criticism of its last product cycle: that demand massively outstripped supply. If Vera Rubin ships on schedule in H2 2026 with adequate volume, it removes a key risk factor that has historically created volatility around Nvidia’s earnings.

The supply chain diversification across 30 countries provides a degree of geopolitical hedging, though the heavy concentration of 150 factories in Taiwan tempers that advantage.