NVIDIA GTC 2026: Vera Rubin Launches, $1 Trillion in Orders, and Rubin Ultra Revealed for 2027

At GTC 2026 on March 16, NVIDIA CEO Jensen Huang unveiled the full details of the Vera Rubin platform — the company's next-generation AI accelerator built on TSMC's 3nm process. The Rubin GPU features a dual-die design with two reticle-sized compute chiplets containing a combined 336 billion transistors, a 1.6× increase over Blackwell's 208 billion. With 50 petaflops of FP4 performance, Vera Rubin delivers 5× inference and 3.5× training performance compared to Blackwell. Vera Rubin is now in full production, with products reaching cloud partners — including AWS, Google Cloud, Microsoft Azure, and OCI — in the second half of 2026.

Huang projected that combined purchase orders for Blackwell and Vera Rubin will reach $1 trillion through 2027, more than doubling NVIDIA's previous $500 billion revenue opportunity projection. The Vera Rubin NVL72 rack is 100% liquid cooled with a cable-free modular tray design that reduces installation time from two hours to five minutes. Azure was the first hyperscale cloud to power up the NVL72, and NVIDIA Cloud Partners have collectively deployed more than one million NVIDIA GPUs representing over 1.7 gigawatts of AI factory capacity — double the footprint from a year ago.

Looking ahead, Huang detailed Rubin Ultra, slated for the second half of 2027. Rubin Ultra will double Vera Rubin's performance to 100 petaflops in FP4, featuring approximately 500 billion transistors, 384 GB HBM4E, and 32 TB/s bandwidth. NVLink 7 will be 6× faster than on Rubin, with 1.5 PB/s of throughput. The Kyber rack design — integrating 144 GPUs in vertically oriented compute trays — arrives with Rubin Ultra and further increases density while lowering latency.

Beyond Rubin, Huang showed early plans for the Feynman architecture, NVIDIA's next major platform expected in 2028. Feynman introduces a new CPU called NVIDIA Rosa (named for Rosalind Franklin) alongside LP40, BlueField-5, and CX10 connected through NVIDIA Kyber — advancing every pillar of the AI factory: compute, memory, storage, networking, and security. The GTC 2026 keynote cemented NVIDIA's multi-year roadmap and reinforced its dominant 92% share of the data center GPU market, with analysts projecting NVIDIA will surpass Apple as TSMC's largest customer sometime in 2026.