Nvidia has another heavyweight news.
According to foreign media reports, on June 2nd, Nvidia founder and CEO Huang Renxun announced that Nvidia Blackwell chips have now begun production. In his speech, Huang Renxun announced that Nvidia will launch Blackwell Ultra AI chips in 2025. The next generation AI platform is named Rubin and will use HBM4 memory. The next generation Rubin platform is under development and will be released in 2026. The Rubin AI platform will use the HBM4 memory chip.
Video screenshot
It is reported that Nvidia's first Blackwell chip, named GB200, claims to be the "most powerful chip in the world" at present. At present, the supply chain has high expectations for GB200, and it is estimated that the shipment volume will have the opportunity to exceed one million units in 2025, accounting for nearly 40% to 50% of Nvidia's high-end GPU shipment volume.
Currently, Blackwell architecture GPU products are in production and will become an important revenue driver for 2024 and 2025. According to a research report by JPMorgan Chase, Nvidia is expected to maintain growth in the second quarter, mainly due to sustained spending from customers on AI/accelerated computing plans, as well as strong demand for its Hopper H100 and new H200 GPU platforms (Blackwell GB200/B200/B100). It is expected that the new products related to Blackwell will undergo initial production and shipment in the third quarter, and achieve large-scale shipment in the fourth quarter.
Video screenshot
Nvidia releases new generation AI chip Rubin
According to foreign media reports, Huang Renxun showcased the latest mass-produced version of Blackwell chips and stated that he will launch Blackwell Ultra AI chips in 2025. The next generation AI platform will be named Rubin, and Rubin Ultra will be launched in 2027. The update rhythm will be "once a year", breaking Moore's Law.
Nvidia released its new generation of artificial intelligence chips on June 2nd local time to replace the previous generation model announced only a few months ago in March. NVIDIA CEO Huang Renxun announced the new AI chip architecture, codenamed "Rubin", on the eve of the Taipei Computer Show.
Rubin's release is only a few months away from the upcoming "Blackwell" model announced in March, which is still in production and is expected to be shipped to customers in the second half of 2024.
So, what exactly is Blackwell? Its powerful performance is simply unbelievable. Please observe these data carefully. In just eight years, Nvidia's computing power, floating-point operations, and artificial intelligence floating-point operations have increased by 1000 times. This speed almost exceeds the growth of Moore's Law during the optimal period.
Video screenshot
Video screenshot
Huang Renxun announced that Rubin seems to have accelerated the company's already rapidly growing pace of AI chip research and development.
Nvidia has promised to release new AI chip models at the pace of "one update per year" as stated by Huang Renxun on the 2nd. Previously, the company had been following a slower two-year update cycle to launch chips.
The transition from Blackwell to Rubin took less than three months, highlighting the intense competition in the AI chip market and Nvidia's strong push to maintain its leading position.
Although AMD and Intel have lagged behind Nvidia in gross profit margins in the recent fiscal quarter, they are still Nvidia's main competitors and are striving to catch up. At the same time, companies such as Microsoft, Google, and Amazon are also competing for Nvidia's leading position, although they are also some of Nvidia's major customers. In addition, a large number of startups are also striving to enter this field.
Video screenshot
"Today, we are at a critical point of significant transformation in the field of computing," Huang Renxun said on the 2nd. "With our innovation in AI and accelerated computing, we are breaking through potential boundaries and driving the next wave of technological progress."
The Rubin chip platform will be equipped with a brand new GPU, which is a key graphics processing technology to help train and start AI systems. In addition, it will also be equipped with other new features, such as a central processing unit called "Vera", although the release on the 2nd did not provide further details.
"A new era of computation is beginning"
Next, Nvidia plans to release an enhanced version of Blackwell Ultra GPU (8S HBM3e 12H), expected to be launched in 2025. The next generation Rubin GPU (8S HBM4) and corresponding platform will be launched in 2026, followed by the Rubin Ultra GPU (12S HBM4) version in 2027.
Video screenshot
Nvidia has already launched the Grace CPU in its CPU products. While upgrading Grace CPU, Nvidia also confirmed the next generation CPU platform Vera CPU. In Huang Renxun's view, there have been 2-3 major technological transformations in the field of computing over the past 60 years, and now we will see it happen again. "A new era of computing is beginning," and the driving force behind this industrial transformation is precisely AI.
Video screenshot
In his speech, Huang Renxun also mentioned several new RTX laptops with generative AI capabilities. An interesting detail is that on site, when he picked up the GeForce RTX Super graphics card and stood next to the DGX data center rack, the difference in size was clear. Huang Renxun pointed to the large rack and said, "We think this is a GPU, this is the real GPU."
Video screenshot
This is also consistent with the current revenue share of Nvidia. Currently, Nvidia's data center business has accounted for nearly 90%, surpassing the revenue of PC consumer graphics cards. Of course, the definition of GPU is also constantly expanding, and it is no longer just a concept of graphical computing. It has truly become the strongest engine of AI, leading Nvidia to climb to a market value of $2.7 trillion.
In addition, Nvidia also announced updates in areas such as communication and software. For example, its Spectra-X Ethernet network platform has been widely used in the industry and will accelerate new product launch plans; 28 million developers worldwide can now download NVIDIA NIM inference microservices.
As of last Friday's close, Nvidia's stock price remained relatively stable at $1096.
The "AI War" has completely begun in Silicon Valley
Currently, a fierce "AI battle" has been completely launched in Silicon Valley.
The latest report released by service consulting firms Dealroom and Flow Partners shows that the global technology industry is entering a new innovation cycle represented by AI and automation. The scientific and technological innovation cycle is about once every 20 years. The previous two innovation cycles occurred in the PC era (the popularity of personal computers) and the Internet era (including the transition to mobile devices and cloud computing).
The report points out that the "seven sisters" (Invista, Apple, Amazon, Meta, Google, Tesla and Microsoft) in the US stock market with a total market value of $14 trillion (about 32% of the S&P 500 index) invest up to $400 billion (about 2900 billion yuan) in AI and cloud infrastructure every year. These investments cover various fields, from AI chips and large models to humanoid robots, autonomous driving, AI healthcare, and more.
Among them, the most intense competition is in the hardware and model layers of AI, with more and more technology giants competing against each other. For example, Nvidia, which dominates the AI chip market, is facing a collective siege from peers to customers, with Microsoft, Google, and Amazon all developing their own AI chips.
The report shows that since this year, the "Seven sisters" have invested 24.8 billion US dollars (about 180 billion yuan) in AI companies through venture capital activities, which exceeds the total amount of venture capital in the UK every year. The most popular AI start-ups among giants include OpenAI, Anthropic, Wave and other star AI start-ups.
Technology giants have a layout in various aspects of AI, with a focus mainly on basic technology (big models) and infrastructure.
It is worth noting that AI investment is gradually shifting towards the application layer. The report points out that there are numerous opportunities for AI applications, with an economic potential of $50 trillion in areas such as healthcare, equipment, media, software cloud, climate, education, defense, mobile, and manufacturing.
Industry insiders point out that the ultimate goal of technology giants is to be the first to achieve General Artificial Intelligence (AGI). Although it is currently unclear how long it will take for AGI to be implemented, it is foreseeable that AI will become the decisive factor in technological competition in this new round of innovation cycle.
