Focus:
Tencent Technology News on March 17, according to comprehensive reports from foreign media, on the afternoon of March 18, local time in the United States (early morning of March 19, Beijing time), Nvidia’s annual GPU Technology Conference (gtc) will It kicked off at the San Jose Convention Center in the United States. This conference will bring together various exciting seminars, and NVIDIA founder and CEO Jensen Huang will also personally deliver a keynote speech on stage. Analysts predict that Nvidia will unveil a series of eye-catching new products during the conference, while also sharing close partnerships with industry giants and the latest progress on its sovereign artificial intelligence plans.
01 releases ai chip b100 blackwell
Nvidia has called on people from all walks of life to pay attention to the upcoming conference in its statement: "Don't miss the keynote speech by Nvidia founder and CEO Jensen Huang. He will share how Nvidia's accelerated computing platform drives artificial intelligence, digital The next wave of twins, cloud technology and sustainable computing.”
In 2022, NVIDIA launched the groundbreaking hopper graphics architecture and h100 GPU, which quickly became the first choice for enterprises to train and deploy artificial intelligence models. Today, the industry generally predicts that Nvidia will launch upgraded versions of hopper and h100 at this conference and continue to lead the innovation trend of GPU technology.
It is reported that the theme of Huang Renxun's speech is "The Transformative Moment of Artificial Intelligence", which undoubtedly implies that he will reveal more details about the new architecture code-named blackwell and the GPU code-named b100. There is widespread speculation in the industry that the b100 is expected to become the most powerful artificial intelligence GPU to date, and will demonstrate excellent performance when running large models such as OpenAI's GPT-3. According to the development roadmap released by Nvidia in October 2023, the b100 chip, as an upgraded version of the h200 chip, is expected to be available sometime in 2024.
Earlier this year, Nvidia had vaguely revealed the amazing performance of the b100 in a document about its h200 GPU. Judging from the preliminary performance figures displayed by the company, the b100 will have a significant improvement in performance compared to the existing a100 and h100 GPUs. This high-profile new processor,
, is expected to be produced using TSMC's cutting-edge 3nm process. This advanced process will allow Nvidia to significantly increase the number of transistors, thereby significantly improving the performance of the b100. What’s more worth mentioning is that Nvidia has chosen a multi-chip design this time, which indicates that the number of transistors and execution units will usher in a qualitative leap. Of course, this design strategy has also brought about a significant increase in power, which is expected to exceed 1000w, a 42% increase compared to the previous generation product h100. Analysts from trendforce, an authoritative
industry organization, pointed out: “NVIDIA’s current flagship product H200 chip in the high-performance computing (HPC) market, with its hopper GPU architecture and hbm3e memory chip, has firmly established itself as the most powerful artificial intelligence chip in the industry. The throne of intelligent computing chips.”
For investors, this conference will be an excellent opportunity to obtain detailed information about new chips. It is reported that Nvidia may announce data on the pricing, performance and market demand of b100. Analysts at Bank of America predict that B100 will be 10% to 30% more expensive than H100.
In addition, Nvidia may also reveal further details about the chip hardware, including information on key components such as liquid cooling components and DRAM memory. These details are crucial to understanding the overall performance and reliability of b100, and will also help investors and the industry more comprehensively evaluate the market prospects of this new product.
Analysts at Wedbush Securities said: "We expect that NVIDIA's new blackwell architecture will bring significant performance improvements compared to hopper. There is already a precedent for this. The training performance of NVIDIA h100 is 9 times faster than a100. Inference performance is 30 times faster than the a100."
These analysts also said that they are extremely interested in the initial demand information for the b100 that Nvidia will soon announce. According to company management, demand for b100 far exceeds supply capacity.
In addition to the focus on the B100, analysts also expect Nvidia to share incremental information about future GPU and artificial intelligence products.They are eager to get a glimpse of Nvidia's strategic layout beyond Nvidia's 2025 roadmap and explore more details about the company's technological innovation and product iterations.
At the same time, NVIDIA may also show the latest progress of its CUDA software at this conference. As a powerful platform, CUDA allows programmers to make full use of the processing power of GPUs in fields such as artificial intelligence. This strategic platform not only helps consolidate Nvidia's leading position in the artificial intelligence market, but also further locks in developers who use its software, almost ensuring that they will continue to use Nvidia's products in the future.
In addition, Nvidia may also announce a big news: the company will start producing its h20 artificial intelligence chip for the Chinese market. This chip serves as a replacement for h100 and h200 chips.
At present, it is still unknown whether NVIDIA will launch new workstation or data center products. However, considering the company's business focus and product strategy, the launch of consumer-grade products seems unlikely.
In the software field, Nvidia’s bets on omniverse and enterprise artificial intelligence are wise and have received positive feedback from the market. It is particularly worth mentioning that Nvidia has cooperated with Microsoft to provide omniverse cloud services on Azure.
02 Artificial Intelligence leaders gather together
As in previous years, Huang Renxun will deliver a keynote speech at this conference. This technology feast is expected to consist of more than 900 conferences, 1,700 speakers, more than 300 exhibits, and more than 20 technical seminars, most of which will focus on the core topic of artificial intelligence.
Huang Renxun’s speech will undoubtedly become the focus of the conference. It is expected that this two-hour speech will provide an in-depth discussion of the latest progress and future development trends of artificial intelligence. In addition to Huang Renxun's speech,
's more than 900 conferences will also cover many hot topics, including augmented reality (AR) and virtual reality (VR), robots, networks, simulations, network security, etc.
Nvidia gtc conference also attracted many big names in the field of artificial intelligence to participate. Their arrival will provide investors with more overall information about the field of artificial intelligence, help investors better grasp the pulse of the market and investment opportunities, and understand the role Nvidia technology plays in it. Well-known speakers at
include:
- Brad Lightcap, COO of OpenAI
- Sébastien Bubeck, Vice President of Generative AI Business at Microsoft President
– Vincent Vanhoucke, Senior Director of DeepMind Robotics at Google
– Joelle Pineau, Vice President of Meta Artificial Intelligence Research
– Fei-Fei Li, Computer Science, Stanford University Professor and co-director of the Institute for Human-Centered Artificial Intelligence (HAI)
- Priscilla Chan, co-founder and co-CEO of the Chan Zuckerberg Initiative, and CEO of Meta Mark Officer Zucker Berg’s wife
– Stefan Goebel, Senior Vice President, Chief of Staff, Head of Product Engineering and Strategic Engineering Partners, SAP Labs
– Belinda Neal , US investment bank Goldman Sachs core engineering business chief operating officer and head of engineering cooperation
- Moises Hernandez-Fernandez (moises Hernandez-Fernandez), JP Morgan Chase Machine Learning Center of Excellence Vice President
- shan jegatheeswaran,
, Vice President and Global Head of Medical Technology Digital, Johnson & Johnson
- Rodolphe Katra, VP of Artificial Intelligence, Medtronic
- Aaron Saunders, Chief Technology Officer, Boston Dynamics
NVIDIA's partners include OpenAI, Microsoft, Amazon, Google parent company Alphabet, Meta, Micron Technology, Oracle, Super Micro Computer, Dell and Intel, etc. These partners will also send people to participate in activities and seminars during GTC.
Nvidia's leadership in artificial intelligence has long been established, and as demand for artificial intelligence technology continues to grow, companies in its ecosystem have also experienced significant growth.Analysts at Bank of America said they would be watching at the conference “the impact of increased competition in custom chips on companies such as Broadcom and Micron Technology.”
In addition, the competition between AMD and Google in the commercial chip market will also become one of the focuses of this meeting. Bank of America pointed out that these silicon chip manufacturers claim that their products have significant advantages in terms of price, performance, etc., especially in artificial intelligence inference. Compared with Nvidia, the cost of hardware is lower, as low as 30-50%. This trend will undoubtedly challenge NVIDIA's market position, but it will also promote technological progress and cost optimization in the entire industry.
03 NVIDIA’s sovereign AI plan
During the GTC conference, with the rapid advancement of artificial intelligence technology, NVIDIA is expected to share more details of its sovereign AI plan.
Bank of America analysts are looking forward to Nvidia's prospects, predicting that Nvidia will focus on expansion opportunities in addressable market size (TAM) beyond the four largest U.S. hyperscale companies (Microsoft, Google, Amazon and Meta). These analysts believe that with the widespread application and popularization of artificial intelligence technology, emerging markets such as secondary and tertiary customers (such as coreweave, lambda labs) and sovereign countries will become important growth drivers for Nvidia.
According to Nvidia, sovereign artificial intelligence refers to a country’s ability to use its own infrastructure, data, labor and business networks to enhance artificial intelligence. Nvidia has been outspoken about its sovereign AI plans, a stance that could help Nvidia stand out, especially amid concerns about potential national security risks associated with data sharing.
Before gtc, analysts at Bank of America raised their target price on Nvidia's stock price from $925 to $1,100, while analysts at Wedbush Securities raised it from $850 to $1,000. Both firms have a "buy" rating on Nvidia stock.
On March 15 (Friday), local time in the United States, Nvidia’s stock price fell slightly that day, closing at $878.4 per share. Despite the recent selloff, the stock remains at all-time highs, up more than 84% year to date and 1,783% over the past five years. The total market capitalization is approximately US$2.2 trillion. (Compiled/Golden Deer)
