Nvidia unveils new AI chip set for March launch

Lead

Nvidia will unveil a new artificial intelligence (AI) chip during the GTC (GPU Technology Conference) in California, held from March 16 to 19. CEO Jensen Huang stated the product will be "a chip that will surprise the world." Full technical details are yet to be released, but significant gains in performance and efficiency are expected.

Development Section

The new chip is expected to target the high-performance AI market, where Nvidia leads with architectures like the Rubin CPX, launched last September with 30 petaFLOPs processing power. Rumors suggest the announcement may feature an even more advanced Rubin version aimed at professional applications, data centers, and generative model training. A less likely possibility is a teaser for the Feynman architecture, rumored to use 1.6 nm transistors and involve collaboration with Intel.

It is expected Nvidia will address common AI performance bottlenecks such as memory. The CEO also mentioned a "close" partnership with memory maker SK Hynix, hinting the new chip might adopt 6th generation HBM4 memory. Experts say this memory could ease bandwidth limits and boost AI accelerator efficiency.

The announcement comes amid rapid AI sector growth with massive infrastructure investments. Huang said there is no AI "bubble," but the start of "humanity’s largest infrastructure project." Nvidia, which has acquired key technologies from Grok and invests in AI startups, reinforces its strategy to dominate the segment. However, the CEO avoided direct comments on investments in companies like Anthropic and OpenAI.

GTC is traditionally a professional and developer-focused event, indicating the announcement will emphasize corporate applications, large-scale AI training, and data center solutions. No updates are expected for the GeForce RTX gaming GPU line.

Developments and Outlook

The new chip presentation is scheduled for the conference from March 16 to 19. The AI market awaits details on architecture, computing power, and possible integration with new HBM4 memory. If Feynman generation rumors prove true, Nvidia could make a major leap in AI accelerators, further solidifying its lead over competitors like AMD and Intel.

Nvidia’s stance underscores its commitment to AI and computing infrastructure advances, signaling strategic partnerships to overcome technical limits. The market expects the new chip to set new performance standards, especially for data centers and companies relying on generative AI.

Readers should note Nvidia is preparing a major professional AI announcement in March, with potential breakthroughs in architecture and next-gen memory, keeping it central in accelerated computing discussions.