NVIDIA Summit promotes Blackwell chip with high energy efficiency, stock price rises 4% to a three-month high

On Tuesday, October 8th, at the "AI Summit DC" artificial intelligence summit in the U.S. capital Washington, D.C., executives from NVIDIA, the "AI darling," announced that their next-generation Blackwell chip will be shipped to customers in the fourth quarter of this year with excellent energy efficiency.

Bob Pette, Vice President and General Manager of Enterprise Platforms at NVIDIA, stated, "The Blackwell platform is basically built with energy efficiency in mind." Developing OpenAI's GPT-4 software on Blackwell requires 3 gigawatts of power, compared to 5500 gigawatts needed for the same process ten years ago. Analysts have pointed out that the AI computing boom had previously raised concerns about power consumption.

In a keynote speech titled "The Transformative Power of Accelerated Computing and Artificial Intelligence," the aforementioned executives introduced how NVIDIA's accelerated computing platform is providing breakthrough performance in the next wave of AI, sensor processing, digital twins, network security, autonomous systems, and other fields.

They also highlighted NVIDIA's artificial intelligence software services and various software platforms. Analysts have noted that selling software is part of NVIDIA's strategy to ensure that customers not only seek the company's popular AI chip processors but also provide a recurring source of income beyond hardware sales. By showcasing software, NVIDIA hopes to prove that it is not just a chip company:

"On one hand, emphasizing that their software can attract developers to use NVIDIA's products, the more developers based on their services, the more companies will seek to use these NVIDIA software platforms.

At the same time, NVIDIA faces increasingly fierce competition in the hardware field from long-time rivals such as AMD and Intel, and its customers are also developing self-made chips to support AI software, which may weaken NVIDIA's profits in the future. Therefore, the company's ultimate goal is to retain customers by using software."

The AI summit on Tuesday boosted NVIDIA's stock price by as much as 4.1% in early trading, reaching the highest level in three months since July 11th. NVIDIA has risen for five consecutive days, with a market value exceeding $32,000 million, surpassing Microsoft on Monday to become the world's second-largest company by market value, second only to Apple.

In the past month, NVIDIA's stock price has been in an upward channel, rising more than 28% from its low rebound on September 6th, mainly due to CEO Jensen Huang and various partners creating momentum by stating that the next-generation AI chip Blackwell has "crazy demand."

On Tuesday, NVIDIA's chip foundry Foxconn confirmed that it is building the world's largest GB200 production facility in Mexico, which is seen by the market as strong evidence of long-term demand for the Blackwell chip.

Regarding software progress, according to NVIDIA executives today, the company has multiple AI software platforms such as NVIDIA NIM Agent Blueprints, NIM, and NeMo. Among them, NIM Agent Blueprints aims to help enterprises develop generative AI applications, NIM can quickly assemble chatbots and AI assistants, and NeMo provides enterprises with methods to build custom generative AI models From American telecommunications giant AT&T, international accounting firm Deloitte, artificial intelligence and machine learning software service provider Quantiphi, to research institutions such as the National Cancer Institute and the SETI Institute searching for extraterrestrial civilizations, all are using NVIDIA's software technology to help employees complete software development and network engineering tasks, and even search for signs of alien life.

For example, AT&T is collaborating with Quantiphi to create a conversational AI platform to assist employees in software development and financial service tasks. The University of Florida is using NVIDIA's NIM and NeMo platforms to enhance a learning management system for teaching assistants; Deloitte is using NVIDIA's NIM Agent Blueprint alongside its own network security products, the National Cancer Institute is using it to help reduce the time and cost of developing new drug molecules; and the SETI Institute is using NVIDIA's Holoscan software toolkit to develop software for receiving and analyzing radio waves from space.

Additionally, NVIDIA has partnered with Carnegie Mellon University and the University of Pittsburgh to establish two new AI technology community centers. The "NVIDIA Center for Robotics, Autonomy, and AI" established in collaboration with Carnegie Mellon University will provide students, faculty, and researchers with resources related to robotics and autonomy, while the "Center for AI and Intelligent Systems" in collaboration with the University of Pittsburgh will focus on AI for health sciences.

Furthermore, on Tuesday, Google announced that NVIDIA's Nvidia Nim is available on the Google Kubernetes Engine, and TSMC is using NVIDIA's computational lithography platform Culitho for production.

On the same day, mainstream investment bank Morgan Stanley also released a research report, expressing optimism about the potential boost to the company's revenue in the last few months of the year from Blackwell's next-generation chip, and hinting that as early as the beginning of next year, Blackwell's production will surpass the existing Hopper chip:

"Based on our checks of the GPU testing supply chain, Blackwell chip production in the fourth quarter should be around 250,000 to 300,000 units, contributing $5 billion to $10 billion in revenue, which still aligns with the optimistic forecast of Morgan Stanley's chief analyst Joe Moore. Blackwell chip production may reach 750,000 to 800,000 units in early next year, nearly tripling from the fourth quarter of this year.

At the same time, we expect Hopper chip sales (including H200 and H20) to be around 1.5 million units in the fourth quarter, gradually declining to 1 million in the first quarter of 2025. Considering that Blackwell's B200 chip price is about 60% to 70% higher than Hopper's H200, the revenue brought in by Blackwell should surpass Hopper in the first quarter of 2025."

Another analyst also released a bullish report on NVIDIA on Monday. Melius Research analyst Ben Reitzes stated that despite NVIDIA's stock experiencing a long rebound, the "current situation is still quite good":

"Strong spending on AI by companies such as Microsoft, Alphabet, and OpenAI, the parent company of ChatGPT, is expected to continue to drive demand for NVIDIA's core GPU products."

NVIDIA's Chief Financial Officer Colette Kress previously forecasted in August that the company's fourth quarter, ending in January next year, will ship Blackwell chips worth billions of dollars. NVIDIA has also announced recent changes to the Blackwell GPU mask to improve production yield, with increased production planned to start in the fourth quarter and continue through the 2026 fiscal year