One company continues to maintain dominance in several crucial areas. Under the leadership of Jensen Huang, Nvidia has emerged as a major new force in the tech industry and recently became the largest company by market capitalization on the U.S. stock market, surpassing both Apple and Microsoft.

What led to this development?

Nvidia has become highly skilled at shaping trends in hardware. Its GPUs are present in the largest data centers, including XAI’s Colossus and other large-scale operations, particularly notable when Musk unpredictably increased GPU numbers by 100,000 units for that same project.

In addition, Nvidia has secured a significant portion of TSMC's foundry services, with the Taiwanese chipmaker supplying production capabilities to many top-tier clients.

Bold New Architectures

In addition to its manufacturing expertise and market presence, Nvidia has been unveiling a series of new processor models.

Not long ago, we were discussing the Nvidia Blackwell architecture, named after David Blackwell, and the Grace Blackwell superchip, which brought groundbreaking performance and attracted a global customer base.

However, during a tech event in 2024, Huang revealed the company’s plans to introduce a new microarchitecture known as “Rubin,” named in honor of physicist Vera Rubin. This will include a GPU called Rubin and a CPU named Vera, delivering an impressive combination of speed and efficiency.

Following that, an enhanced variant called Rubin Ultra will be introduced, along with a future hardware platform named Feynman (after Richard Feynman), expected in 2028.

Regarding memory bandwidth, Rubin is projected to reach around 900 GB per second, while Rubin Ultra could achieve approximately 1200 GB per second. The Feynman architecture is anticipated to operate around 1000 GB per second, optimized for specific computational tasks and precision math.

Additionally, there will be a Rubin AI platform, forming part of the company's broader ecosystem built around Rubin chips. It will utilize ultra-fast HBM4 memory and NV Link 6 switches, enabling up to 3600 GB per second in bandwidth.

All of this is tailored for data centers and designed specifically for AI applications.

Putting LLMs on Devices

Now, let’s explore how Nvidia’s Rubin will influence the market.

Analysts note that it may not immediately mean running full AI models offline—but these architectures will help accelerate edge computing strategies, leading to more powerful local systems. As a result, the AI features already present on personal devices will become smarter, more capable, and able to process complex reasoning.

We can expect to soon witness the impact of these hardware innovations in consumer technology, especially in wearables and mobile devices.

From health and education to retail, healthcare, and legal services, entire industries are being reimagined through portable, high-performance LLMs that no longer rely solely on centralized data centers.

“These devices span from high-capacity edge servers to low-resource IoT sensors, and include everyday examples such as smartphones, smart home gadgets, self-driving cars, and industrial robots,” explains Bhavishya Pandit at Datacamp, highlighting the benefits and potential transformation brought by edge computing.

Certainly, critical work will continue in those massive data centers, but increasingly, it will shift toward research and niche IT functions. Meanwhile, we’ll see growing deployment directly on end-user devices.

This trend will also affect regular users more directly—whether through greater adoption of AI-powered apps or increased appreciation for what AI can do, likely both.

How do you envision the future of AI when everyone carries advanced artificial intelligence tools in their pockets?

Let’s monitor developments through the rest of 2025 and into 2026 as Rubin begins its rollout in the hardware space.

The above is the detailed content of The Blackwell Successor: What Nvidia's Rubin Means For Your Phone. For more information, please follow other related articles on the PHP Chinese website!