nvidia unveils 'rubin' chip roadmap after blackwell ultra launch

2025-03-22 nvidia

Santa Clara, Saturday, 22 March 2025.
nvidia ceo jensen huang has revealed the company’s next-generation chip, ‘rubin’, immediately following the ‘blackwell ultra’ launch. the ‘rubin’ chip is set to launch in the second half of 2026. it will feature 288 gibibytes of hbm4 memory. it will deliver 50 petaflops of fp4 inference performance per chip. a full nvl144 rack configuration will provide 3.6 exaflops of fp4 compute. the announcement underscores nvidia’s dedication to innovation within the rapidly evolving semiconductor market. this rapid iteration highlights the intense competition and pace of development in ai hardware.

GTC 2025: nvidia’s ai vision

Nvidia’s GPU Technology Conference (GTC) 2025 has become a focal point for unveiling innovations in artificial intelligence and high-performance computing [7]. Jensen Huang, Nvidia’s CEO, introduced the ‘Vera Rubin’ platform, designed to power the next generation of large-scale computing [7]. This platform connects millions of GPUs, addressing the processing demands of machine learning and generative AI applications [7]. The rapid succession of chip announcements, from ‘Blackwell Ultra’ to ‘Rubin,’ demonstrates Nvidia’s commitment to maintaining its competitive edge [1].

competitive landscape

The unveiling of ‘Rubin’ occurred alongside Nvidia’s announcement of ‘Blackwell Ultra’, a more powerful version of its flagship GPU, expected in the second half of 2025 [4][6]. This highlights the intensifying competition between Nvidia and AMD in the AI chip market [1]. Huang also claimed that tariffs would not cause significant damage in the short run [5]. TSMC has pledged to invest at least $100 billion in U.S. chip facilities, reflecting a broader trend toward bolstering domestic chip production [5].

market dominance and revenue

Nvidia’s dominance in AI-powered computing is evident in its data center division’s revenue, which reached $35.6 billion in the final quarter of 2024 [7]. In comparison, Intel’s data center revenue has declined to $3.4 billion, underscoring Nvidia’s growing leadership in data infrastructure [7]. This shift highlights the increasing importance of data centers in supporting cloud services, enterprise software, and AI training [7]. Nvidia’s focus on data centers is further emphasized by the ‘Vera Rubin’ platform, built for vast data center clusters [7].

broader ai initiatives

Nvidia is expanding its AI ecosystem through various initiatives. The company is collaborating with Disney Research and Google DeepMind to develop Newton, a physics engine for simulating robotic movements [5]. Nvidia also plans to release an early, open-source version of Newton later in 2025 [5]. Furthermore, Nvidia acquired synthetic data startup Gretel, signaling its investment in AI model development [5]. These initiatives, combined with the launch of new chips, solidify Nvidia’s position as a leader in the AI revolution [6].

stock market implications

While Nvidia has reaffirmed its leadership in AI chips, this hasn’t significantly boosted its stock price [8]. Some analysts express doubts about the sustainability of Nvidia’s high growth rate [8]. However, the company’s continuous innovation and strategic partnerships position it favorably for long-term growth in the AI market [3][5]. Jensen Huang’s vision of AI as ‘the operating system of every industry’ reinforces Nvidia’s commitment to driving advancements across various sectors [5]. The announcement of new chips and platforms is expected to positively influence investor confidence [3].

Bronnen

next-gen chip rubin architecture