Nvidia confirms H2 2026 Vera Rubin rollout amid record $75B data center quarter

Nvidia confirmed the Vera Rubin AI platform is on schedule for partner availability in the second half of 2026. [2] The flagship NVL72 system packs 72 Rubin GPUs and 36 Vera CPUs into a single rack, delivering 3.6 exaflops of NVFP4 inference compute. [2] The architecture promises five times the inference performance of current Blackwell systems at the rack level. [2]

Dell is deploying this hardware, which 5,000 enterprises including Lilly, Samsung, and Honeywell are already utilizing. [1] The architecture is designed to deliver a ten times lower cost per token for AI inference compared to Blackwell. [1][2]

Nvidia reported data center revenue of $75 billion, marking a 92 percent increase from the previous year. [3][6] Networking revenue nearly tripled year over year. [6] Overall company revenue for the quarter reached $81.6 billion. [5] CEO Jensen Huang stated that demand has gone "utterly parabolic" and expects $1 trillion in Blackwell and Rubin revenue from 2025 through calendar 2027. [1][6]

Alongside these financial milestones, the company refreshed its share buyback program with an $80 billion authorization and raised its dividend to 25 cents per share. [5][6]

Some of Nvidia's top customers have been told that prices will jump by more than 15 percent on systems including Vera Rubin and Grace Blackwell starting in early 2027. [7] Customers were reportedly warned about these AI-related price hikes as the hardware transitions approach. [8]

Competitor AMD reported record data center revenue of $6.72 billion, a 107 percent increase year over year. [6] Anthropic has committed to up to two gigawatts of AMD's MI450 series GPUs in Helios racks, with the first gigawatt starting in 2027. [6]

Nvidia’s combination of massive hardware scale and significant token cost reductions paints a challenging picture for competitors, even as AMD gains traction with major clients like Anthropic. While the anticipated 15 percent hardware price increase could pressure capital expenditures, the 10x drop in token costs may ultimately offset those upfront premiums for heavy inference users. Nvidia's staggering $75 billion data center quarter highlights how firmly it controls the high-end infrastructure market compared to AMD's $6.72 billion data center segment. What the sources don't address: How the projected 3,400 percent growth in global token consumption will strain data center power grids before these next-generation exaflop racks even come online.