On June 2nd, NVIDIA founder Huang Renxun gave a keynote speech titled "Unleashing the New Industrial Revolution", which was another heavyweight speech he gave after the NVIDIA GTC conference. Following the previous release, Huang Renxun integrated two GPUs and one CPU chip into a GB200 motherboard. This time, he has once again "accelerated" the computing power of GPUs and announced the launch of the next generation architecture called "Rubin" in 2026.
Huang Renxun stated that Nvidia will launch new products at the rate of one new chip per year. In 2025, the company will release Blackwell Ultra, launch a new architecture Rubin in 2026, and release Rubin Ultra in 2027. This rapid pace of updates indicates that Nvidia not only maintains a leading position in technology, but also strives to drive overall industry progress by reducing inference costs.
GPU drives computing costs down, AI can evolve
In his speech, Huang Renxun reviewed in detail the successful journey of NVIDIA along the way and emphasized that it was NVIDIA that facilitated the current AI era.
Huang Renxun believes that NVIDIA's success is not accidental, but based on the company's profound understanding of future technology needs and long-term technological accumulation. Looking back at the past, Nvidia foresaw the bottleneck of CPU long ago, from deep cultivation of GPU to the launch of CUDA parallel computing architecture, and then to the launch of NIM microservices, so it chose the path of GPU and accelerated computing. This choice not only enabled Nvidia to achieve great success in the context of the surge in demand for artificial intelligence computing, but also established the company's leadership position in the field of generative AI.
In his narration, Huang Renxun specifically mentioned the importance of CUDA. He stated that CUDA was Nvidia's first intimate encounter with artificial intelligence, and since then, the company has deeply understood the essence of deep learning and consciously reinvented everything. By continuously optimizing and expanding CUDA, Nvidia has made it a powerful development platform with 5 million developers. Huang Renxun proudly pointed out that it is precisely because of CUDA that global deep learning scientists can fully utilize its potential, thereby driving the progress of the entire industry.
When you speed up data processing, demonstrations don't take long. CUDA has now reached what people call a critical point, but it's even better than that. CUDA has now reached a virtuous cycle. If you look back at history and all the computing architectures and platforms, this situation rarely happens Huang Renxun said.
Huang Renxun repeatedly mentioned Nvidia's efforts in reducing computational costs in his speech. He said that it was because Nvidia reduced the marginal cost of computing by one million levels by using specific algorithms that it was possible to use Internet data to train large models. This technological breakthrough not only promotes the development of artificial intelligence, but also makes generative AI an important feature of today's era.
We are not in the era of AI, but in the era of generative AI
Huang Renxun's speech not only reviewed Nvidia's technological development process, but also looked forward to the future of the era of generative AI. He pointed out that we are not currently in a simple AI era, but a generative AI era. In this era, almost everything can be converted into tokens and processed and optimized through generative AI.
Huang Renxun emphasized that generative AI will reshape various industries and drive the entire $3 trillion IT industry to transform into an AI factory, producing AI products for each industry. He explained that in the future, every PC equipped with an RTX graphics card will become an AIPC, capable of efficiently processing and generating various types of data. In fact, this also corresponds to Microsoft's recent announcement at a press conference that they will upgrade their computers from PCs to AIPCs.
Huang Renxun further believes that generative AI is not only a technological breakthrough, but also a new industrial revolution. By continuously improving computing performance and reducing costs, Nvidia is making this revolutionary technology accessible and utilized by every company. He said that every Internet or computer interaction in the future may run generative AI in the cloud, which will greatly change our way of life and work.
However, he also did not blindly praise the ability of AI. Regarding the controversy over whether Sora understands the laws of physics in AI since the beginning of this year, Huang Renxun stated that although the next generation of AI needs to be based on physics, most AI today do not understand the laws of physics. "It is not rooted in the physical world and cannot generate images, videos, 3D graphics, and many physical phenomena. We need AI based on physics and understanding the laws of physics. Of course, learning from videos is a source. Another way is to synthesize data, simulate data, and use computers to learn from each other."
"Through continuous innovation and technological breakthroughs, Nvidia is leading the entire industry towards a new era. In this era, generative AI will become the core force driving progress in various industries, and Nvidia is undoubtedly a key driver in this process." Huang Renxun said.
New Beijing News, Shell Finance News, Reporter Luo Yidan