Compared with Musk's sky-high salary, Huang Jenxun, who has just led LinkedIn Weida to a market value of 3 trillion yuan, has risen to $34 million, a 6-fold increase from the previous year, but this is not the focus of attention in the capital market and the industry, but is this round of generative AI in a bubble? Huang's answer is that robotics and sovereign artificial intelligence are new markets to be tapped and explored.
After surpassing Apple to become the world's largest market capitalization, although the stock price has fluctuated recently, Nvidia is undoubtedly still the company on the top of the mountain and is scrutinized by all parties.
On June 26, Nvidia held a shareholder meeting that lasted only 30 minutes and did not provide much new information that investors were expecting, but approved all 12 director nominees and approved the executive compensation plan.
However, after the shareholders' meeting, Nvidia's stock price fell intraday, closing up 0.25% on the day to close at $126.4 per share, with a market capitalization of $3.11 trillion.
In the compensation plan document, Huang's compensation package for fiscal year 2024 is about $34 million, a six-fold increase from 2023.
At the annual shareholder meeting, Huang reaffirmed the potential for the Blackwell platform to be "the most successful product" and expressed confidence that the company remains steadfast in the face of the AI wave. Including new opportunities for NVIDIA's development in the wave of artificial intelligence, he deliberately emphasized: especially in the field of robotics and sovereign artificial intelligence.
Nvidia did not disclose news about new products at this shareholders' meeting, but Huang answered questions from investors on topics such as how to deal with competition and how the company can diversify, which is also the focus of Nvidia's next layout:
Blackwell持续发力：
"A year ago, demand for the Hopper architecture was still strong, and the next-generation Blackwell architecture has been widely accepted by the industry. The Blackwell architecture product is probably the most successful product in our history, and in the history of computing in general. ”
How to stay ahead and how to perceive competitor threats:
"Ten years ago, we jumped at the chance to dive into deep learning and systematically invented everything about GPUs (graphics processing units) that connected systems, networks, and software. We've invested billions of dollars to push the boundaries of computing, and thousands of engineers have been working on deep learning for decades. Our competitive advantage lies in the expertise, scale, and speed to create end-to-end optimized AI computing systems. ”
"Nvidia has 'transformed' from a previous gaming company to a data center-focused company. The company also wants to create new markets for the AI industry, for example in the field of industrial robots. The company's goal is to work with every computer manufacturer and cloud computing provider to achieve this. ”
"The next wave of artificial intelligence will automate $50 trillion worth of heavy industry, and soon, robot factories will use robots to make robotic products."
About Sovereign AI:
"Sovereign AI refers to a country's ability to build AI using its own infrastructure, data, workforce, and business networks. Nvidia revealed in May that Sovereign AI expects to bring in nearly $10 billion in revenue for the company this year, compared with zero last year. ”
"The rise of 'sovereign AI' proves that the importance of developing AI that is appropriate to the language and culture of each country is universally recognized."
Not long ago, Zhu Xiaohu, the managing partner of GSR Ventures, bluntly said in a public exchange with the media, "Whether Nvidia's market value can continue depends on whether the Scaling Law (scale law) has an effect, and if the Scaling Law is invalid, its stock price will fall to a mess." ”
Zhu Xiaohu also predicts: "By the end of this year, if OpenAI's release of GPT-5.0 falls short of expectations, Nvidia will be finished." ”
Not only that, recently, a number of well-known industry commentators and media have published articles saying that there is a bubble in Nvidia's market value of nearly $3 trillion.
Does the rise of Nvidia have the shadow of the Internet bubble?
It's reminiscent of the tech boom at the turn of the century — in the first three years to March 1994, the network equipment company Cisco's stock soared 31-fold, but plunged 51 percent as fears of a recession and rival product launches caused customers to cut orders. In March 2000, Cisco overtook Microsoft on the Internet revolution, and its stock price rebounded to 73 times, occupying the position of the most valuable company at the peak of the dot-com bubble, but it plummeted by 90% in the following years, and has never come close to the highs of the dot-com bubble since.
How long will Nvidia's craze last, and will it become a company like Cisco?
Before making a correct judgment, we must first clarify the correct relationship between the technological revolution and the financial capital market.
We have always regarded the Internet as a panacea that can change the world, and with the explosion of the concept of AI, we have begun to think about whether embodied intelligence is a bridge between AGI and the physical world. Because they have too high expectations for the evolution of the technology paradigm, investors speculate on technology stocks to sky-high prices, but when the stock prices of these companies plunge and the general economic crisis appears, they mistakenly believe that market demand is weakening or failing.
But the reality is that the bubble bursts because our expectations are overly optimistic, not because the real market actually bursts. When our expectations outweigh the market growth of technology, creating cognitive biases, a "tulip mania" event is formed.
In recent years, as the demand for GPUs for artificial intelligence has grown, it is not necessary to worry about Nvidia's performance, but to manage the growth expected by the public. A bubble occurs when public expectations are greater than the actual growth rate of the market.
Huang previously shared the next supply plan in a speech in Taipei: Nvidia's new GPU architecture will be accelerated from a biennial iteration to a one-year cycle.
Regarding the current sales situation in the Chinese market, Huang still emphasized the huge demand for Nvidia's AI GPUs, saying that GPU resources are very tight right now, and tech giants and about 15,000-20,000 generative AI startups are competing.
When an analyst asked about the company's previous chips for the Chinese market, Huang said, "Our business in China is really significantly lower than it used to be." Competition is now more intense in China due to the limitations of our technology. ”
After tightening U.S. export restrictions banned Nvidia from exporting its cutting-edge A800 and H800 chips, Nvidia pinned its hopes on the H20, L20 and L2 chips to grow its market share in China.
Nvidia's earnings report has become one of the main references for investors to measure the AI boom, and in May, Nvidia released its latest quarterly report.
According to the first quarterly report, revenue reached $26 billion, of which data center revenue reached $22.6 billion, a record year-on-year increase of 427%, and it is still Nvidia's most profitable business.
Nvidia also highlighted how high the return on GPUs would be if cloud service providers bought it — for every $1 spent, there was an opportunity to earn $5 over four years, estimating that about 40% of its data center revenue had been driven by inference over the past year, and predicting that the automotive industry would be the largest enterprise vertical in the data center space this year.
Customer information in the automotive industry has also been shared - Tesla bought 35,000 H100 sheets to form an AI training cluster; Llama 3 is trained with 24,000 H100s; Xiaomi's first electric car, the SU7, will use NVIDIA's DRIVE Orin on-board computer, and BYD, Xpeng, etc. will use the next-generation DRIVE Thor based on the Blackwell architecture......
The outside world has always believed that the development of NVIDIA's intelligent driving software does not match the strong position of the chip. But in the first quarterly report, people saw new changes.
Although the automotive business is still small in Nvidia's current revenue, it was $329 million in the first quarter, but it increased 17% sequentially. This was the only business outside of Nvidia's data center business to achieve sequential growth in the first quarter.
Huang often uses the phrase "we're always in danger" to express Nvidia's challenges in the automotive business, which should account for 30 percent of revenue, according to Huang's plan.
It is foreseeable that there will be more and more Chinese car companies as the officially announced automotive business partners of NVIDIA.