Jensen Huang: Nvidia Quarterly Revenue Nearing $100 Billion, Rubin Ultra on Schedule

Jensen Huang told Morgan Stanley analysts that Nvidia quarterly revenue is approaching 100 billion dollars. Growth continues to accelerate. The comments came during a roadshow presentation last week. According to Reuters, Huang stated that "our quarterly revenue run rate is fast approaching $100 billion" during the Morgan Stanley event. Morgan Stanley maintained its overweight rating with a $288 price target.

Huang also rejected rumors that the Rubin Ultra architecture is delayed. He said the next flagship remains on schedule for next year. Rack design changes are only optimizations for system architecture.

Revenue Scale Signals Continued Dominance

Nvidia projects its CPU business will reach roughly 20 billion dollars this fiscal year. The next generation Vera CPU is aimed at the general server market. These moves expand beyond the AI accelerator focus.

A frontier AI model project that once relied mainly on ASIC chips now allocates nearly half its compute to Nvidia GPUs. Market observers link the shift to Anthropic. The change highlights growing GPU preference among large training runs. Teams working on such projects see concrete benefits in practice: engineers gain immediate access to CUDA-optimized libraries and debugging tools that reduce iteration cycles by weeks, while missing those features on pure ASIC setups often means longer debugging sessions and narrower model architecture choices. Bloomberg reported similar shifts among frontier labs reallocating racks.

Morgan Stanley kept its overweight rating on the stock with a 288 dollar target price. The firm cited the revenue trajectory and architecture timeline as key supports.

Rubin Ultra Timeline Holds Despite Adjustments

Huang said the Rubin Ultra architecture will ship as planned next year. He dismissed delay reports outright. The rack adjustments address power and networking needs rather than core silicon issues.

Current systems already handle massive clusters. The next architecture adds further scaling without altering the delivery window. Engineers are refining cooling and interconnect layouts to support higher densities.

Anthropic Shift Shows GPU Preference Trend

One large training effort moved from heavy ASIC reliance toward balanced GPU usage. The change brings Nvidia share close to 50 percent. The project is widely viewed as Anthropic work. In practice, teams reallocating 30-50% of racks from ASICs to Nvidia GPUs gain the ability to run mixed workloads in a single cluster - such as simultaneously training a base model on GPUs while performing inference fine-tuning on remaining ASIC capacity - avoiding the downtime that occurs when custom silicon queues back up. This flexibility lets smaller sub-teams test new model variants without waiting for ASIC reconfigurations.

This pattern appears across other frontier labs as well. Teams balance custom silicon with standard GPU fleets for flexibility. The 50 percent figure marks a notable departure from prior ASIC-heavy plans.

CPU Expansion Targets Broader Server Demand

Nvidia expects 20 billion dollars in CPU revenue this fiscal year. Vera will extend that push into mainstream servers. The strategy reduces dependence on any single product line.

Enterprise buyers already test Grace CPUs in selected workloads. The company aims to widen those deployments through 2027. Revenue growth in this segment adds a second leg to the business.

Investors Focus on Execution After the Update

Analysts watched the roadshow comments for any sign of slippage on Rubin Ultra. None appeared. The denial was direct and tied to specific rack work rather than silicon delays. Nvidia's official blog post from the event reiterated the original Rubin timeline.

The revenue run rate near 100 billion dollars per quarter sets a high bar. Continued acceleration depends on sustained AI spending by cloud providers and labs. Morgan Stanley sees the path as intact.

Future checkpoints include the next earnings release and any fresh cluster announcements. Those updates will show whether the 50 percent GPU share trend spreads to other projects.