Nvidia Corp. CEO Jensen Huang has revealed the company’s plan to release annual upgrades for its AI accelerators. He announced that a Blackwell Ultra chip is set to debut in 2025, followed by a next-generation platform called Rubin in 2026.
In addition to this, Nvidia introduced new tools and software models ahead of the Computex trade show in Taiwan. The company believes that the emergence of generative AI signifies a new industrial revolution and anticipates playing a significant role as the technology transitions to personal computers. Huang shared these insights during his keynote address at National Taiwan University.
Nvidia has greatly benefited from the significant increase in AI investment, which has propelled the company to become the top chipmaker globally. However, Nvidia is now aiming to expand its customer base beyond the few dominant cloud-computing giants that currently drive a significant portion of its sales. In this expansion, CEO Huang foresees a wider range of companies and government agencies adopting AI technology, including industries such as shipbuilding and pharmaceutical development. Huang reiterated his stance from a year ago, emphasizing that those without AI capabilities will be at a disadvantage and risk falling behind.
According to Huang, there is a phenomenon of computation inflation occurring due to the exponential growth of data that needs to be processed. He believes that traditional computing methods are unable to keep up with this demand, and that Nvidia’s accelerated computing style is the solution to reducing costs. Huang claims that Nvidia’s technology can achieve 98% cost savings and 97% less energy consumption, although he acknowledges that this calculation may not be entirely accurate.
Following the announcement, shares of Taiwan Semiconductor Manufacturing Co. and other suppliers experienced a rise. TSMC’s stock increased by as much as 3.9%, while Wistron Corp. gained 4%.
Huang revealed that the upcoming Rubin AI platform will utilize HBM4, the next generation of high-bandwidth memory that has become a bottleneck in AI accelerator production. He noted that SK Hynix Inc., a key player in this market, is already sold out through 2025. However, Huang did not provide detailed specifications for the upcoming products, which will be released after Blackwell.
According to Dan Newman, CEO and chief analyst at Futurum Group, the strategic move to release Rubin and Rubin Ultra by Nvidia demonstrates their commitment to a yearly refresh cycle. Newman also highlights the company’s emphasis on innovation and their relentless pursuit of pushing the boundaries of technology, encompassing software, process, packaging, and partnerships to protect and expand their market position.
Having initially started by selling gaming cards for desktop PCs, Nvidia’s gaming background is now proving advantageous as computer manufacturers seek to incorporate more AI functionalities into their machines.
At Computex, Microsoft Corp. and its hardware partners are showcasing new laptops with AI enhancements under the Copilot+ branding. These devices predominantly utilize a new type of processor, provided by Nvidia’s competitor Qualcomm Inc., which enables them to have longer battery life.
According to Nvidia, although those devices are capable of basic AI functions, their performance can be significantly enhanced and new features can be introduced to popular software such as games by adding an Nvidia graphics card. The company mentioned that PC manufacturers like Asustek Computer Inc. are already offering computers with this capability.
In order to assist software developers in bringing more advanced capabilities to the PC, Nvidia is providing tools and pretrained AI models. These resources will handle complex tasks, including determining whether to process data locally or send it to a data center via the internet.
In addition, Nvidia has unveiled a new server computer design that utilizes its chips. The MGX program is utilized by major companies like Hewlett Packard Enterprise Co. and Dell Technologies Inc. to expedite the development of products for use by corporations and government agencies. Interestingly, even competitors Advanced Micro Devices Inc. and Intel Corp. are leveraging this design by incorporating their processors alongside Nvidia chips.
Following Huang’s remarks, Lisa Su, CEO of AMD, took the stage at Computex to discuss her company’s advancements in AI chips. AMD is accelerating the release of its AI processors in order to narrow the gap with Nvidia in this rapidly expanding industry.
According to Nvidia, the previously announced Spectrum X and Nvidia Inference Microservices (NIM) are now readily available and gaining significant adoption. Nvidia refers to NIM as “AI in a box” and will be providing free access to these products. NIM consists of software and models that act as intermediaries, enabling companies to efficiently launch AI services without concerning themselves with the underlying technology. However, companies utilizing NIM will be required to pay Nvidia a usage fee.
Huang also advocated for the utilization of digital twins within Nvidia’s virtual world, known as the Ominverse. As an example of its potential, he presented Earth 2, a digital twin of our planet, showcasing its ability to facilitate advanced weather pattern modeling and other intricate tasks. Furthermore, he highlighted that companies like Hon Hai Precision Industry Co., widely recognized as Foxconn, based in Taiwan, are employing these tools to enhance their factory operations and optimize planning processes.