Jensen Huang, CEO of Nvidia, surprised attendees at CES by prematurely unveiling the new “Vera Rubin” AI server generation. This step comes in response to the accelerating pace of innovation and the urgent need for massive computing capabilities to train future models and simulate physical reality.
Article contents:
- Introduction
- New Vera Rubin systems
- Accelerated timeline
- Simulation and training power
- Market impact
- FAQs
Introduction
As global competition in artificial intelligence heats up, Nvidia has broken its usual schedule. Instead of waiting for the spring conference, Jensen Huang revealed in Las Vegas the latest “Vera Rubin” server systems, which will roll out in the second half of the year. This announcement reflects the immense pressure to provide infrastructure capable of keeping pace with the evolution of smart models.
New Vera Rubin systems
Named after the pioneering American astronomer Vera Rubin, these servers are designed to handle massive computing workloads. The company stated that developers can now train a model containing 10 trillion parameters in just one month, using a quarter of the chips required by the previous “Blackwell” generation. These systems also deliver a tenfold reduction in the cost of the “inference” process (running models for users).
Accelerated timeline
Huang explained that “the amount of compute required for AI is growing astronomically,” pushing the semiconductor industry to move faster. This acceleration aims to enable companies to reach the “next frontier” of artificial intelligence without waiting for traditional development cycles.
Simulation and training power
Nvidia emphasized that the next phase relies on “Omniverse,” where artificial intelligence learns through reality simulation rather than traditional data alone. “Rubin” systems are designed to run these complex simulations, which is vital for developing robotics and autonomous vehicles, or what is known as “physical AI.”
Market impact
Nvidia integrated networking and memory products into the new system, strengthening its position as the largest combined networking and computing hardware company. Analysts described the announcement as a “stunning generational leap,” noting that the early reveal sends a reassuring message to investors and developers about the company’s ability to meet future demand.
FAQs
What is the main advantage of the Vera Rubin chips?
Answer: The ability to train massive models (10 trillion parameters) 4 times faster and at 10 times lower operating costs.
When will these systems be available?
Answer: They are scheduled to hit the market in the second half of 2026.
Why were they named this way?
Answer: In honor of astronomer Vera Rubin, who made pioneering contributions to understanding galaxy rotation and dark matter.
