Nvidia CEO Jensen Huang said the company plans to upgrade its AI accelerators annually, unveiling the Blackwell Ultra chip in 2025 and its next-generation platform, Rubin, in 2026.
The company is currently best known for its artificial intelligence data center systems, but on the eve of the Computex trade fair in Taiwan, it also unveiled new tools and software models. Nvidia sees the rise of generative AI as a new industrial revolution and expects the technology to play a key role in the transition to personal computers, the CEO said in a keynote speech at National Taiwan University.
Nvidia has been the biggest beneficiary of huge investments in AI that have made it the world's most valuable chipmaker. But the company is now looking to expand its customer base beyond the few cloud-computing giants that make up the bulk of its revenue. As part of this expansion, Huang predicts a broader range of businesses and government agencies, from shipbuilders to drug developers, will adopt AI. He returned to themes he presented at the same event a year ago, including the idea that companies without AI capabilities will be left behind.
“Computing inflation is happening,” Huang said Sunday. As the amount of data to be processed grows exponentially, traditional computing methods cannot keep up, and only NVIDIA's accelerated computing methods can reduce costs, Huang said. He claimed NVIDIA's technology can reduce costs by 98% and energy by 97%, saying, “This is the CEO's calculation, it's not exact, but it's correct.”
Hwang said the upcoming Rubin AI platform will use HBM4, the next version of the essential high-bandwidth memory that has become a bottleneck in AI accelerator manufacturing. Leader SK Hynix has nearly sold out of inventory through 2025. He did not provide detailed specifications for upcoming products following Blackwell.
Nvidia got its start selling gaming cards for desktop PCs, but that background is influencing computer makers as they look to add more AI capabilities to their machines.
Microsoft and its hardware partners are showing off new AI-enhanced laptops at Computex under the brand name Copilot+, and most of the devices coming to market are based on a new type of processor from Nvidia rival Qualcomm that allows the devices to run for longer on a single charge.
The devices are good for simple AI functions, but adding an Nvidia graphics card could significantly boost performance and add new features to games and other popular software, Nvidia said. PC makers such as Asustek Computer Inc. offer such computers, the company said.
To help software makers add even more new features to PCs, Nvidia provides tools and pre-trained AI models that handle complex tasks like deciding whether to process data on the machine itself or send it over the internet to a data center.
Separately, Nvidia has unveiled new designs for server computers that use its chips. The MGX program is being used by companies including Hewlett Packard Enterprise Co. and Dell Technologies Inc. to help bring products to market faster for use by businesses and government agencies. Rivals Advanced Micro Devices Inc. and Intel Corp. are also using the designs in servers that use their processors alongside Nvidia chips.
The company said previously announced products, including Spectrum X for networking and Nvidia Inference Microservices (NIM), which Huang calls “AI in a box,” are now generally available and widely adopted. The company will also offer free access to NIM products. Microservices are a set of intermediate software and models that help companies deploy AI services faster without worrying about the underlying technology. Companies that deploy microservices must then pay a fee to Nvidia for their use.
Huang also promoted the use of digital twins in a virtual world that Nvidia calls the Ominverse. To show the scale of it, Huang introduced Earth 2, a digital twin of the Earth, and showed how it can help perform more sophisticated weather pattern modeling and other complex tasks. Huang noted that contract manufacturers such as Taiwan-based Hon Hai Precision Industry Co., aka Foxconn, are using the tool to plan and run their factories more efficiently.
