On June 19th at 1:01 am, the market cap reached 3.33 trillion USD in intraday trading, surpassing Microsoft to become the world's highest market cap company. As of the US stock market close, Nvidia's total market value was 333.53 billion USD, surpassing the total market value of 331.73 billion USD by 1.80 billion USD. Nvidia's market cap growth seems "crazy," but it reflects the capital mapping of the AI storm and leading technology companies snapping up Nvidia's high-end chips. According to statistics from Wells Fargo & Co, Nvidia currently has a 98% share of the global data center AI acceleration market and is in an absolute dominant position. "Nvidia's high-end chips are in short supply, and many US tech giants are lining up to buy high-end AI chips such as H100 and B200. In the high-end chip field, Nvidia's market share is over 80%." Angel investor and AI expert Guo Tao revealed to "AI Light Year."$NVIDIA (NVDA.US)$The weather is good today The weather is good today.$Microsoft (MSFT.US)$Please use your Futubull account to access the feature.

Nvidia's market cap skyrocketing is a result of the AI boom and leading tech companies frantically snapping up Nvidia's high-end chips. According to statistics from Wells Fargo & Co, Nvidia currently has a 98% share of the global data center AI acceleration market and is in an absolute dominant position.

"Nvidia's high-end chips are in short supply, and many US tech giants are lining up to buy high-end AI chips such as H100 and B200. In the high-end chip field, Nvidia's market share is over 80%." Angel investor and AI expert Guo Tao revealed to "AI Light Year."

All technology companies around the world want to win the top spot in this round of the AI era, and purchasing AI chips to establish computing power barriers is the threshold. "Currently, the development of large models has Scaling Laws. We are still in the stage of computing power's brutal growth. The time and cost of training with 10,000 cards and 100 cards are certainly different." Lenovo Capital Partner Wang Guangxi expressed to "AI Light Year."

It took Nvidia 24 years to reach a market cap of $1 trillion from its IPO. It took Nvidia 10 months to reach a market cap of $2 trillion from $1 trillion. Nvidia's market cap reached $3 trillion in just 3 months. Nvidia's market cap skyrocketing was directly pushed up to the world's number one position by US tech giants themselves.

Money can't buy Nvidia's high-end chips.

On May 26, US time, CEO Elon Musk's AI startup xAI announced it had raised $6 billion in Series B financing. A week later, Musk stated that the company would spend at least $9 billion to purchase 300,000 Nvidia B200 AI chips.$Tesla (TSLA.US)$"Considering the speed of technological progress, putting 1,000 megawatts of power into H100 is not worth it. The xAl 100,000 H100 liquid-cooled training cluster will be launched in a few months. The next important step is to deploy about 300,000 B200 devices and equip them with the CX8 network next summer." Musk revealed on X on June 3.

Nvidia's high-end AI computing chips include the A100, H100, and B200 chips that will be launched later this year. Public information shows that the floating-point operation capability of the B200 chip is five times that of the H100 or will not be launched until the second half of this year. While waiting for the B200 chip, other technology companies are also snatching up H100.

In early 2024, Meta founder Zuckerberg stated on Instagram that "to train the next generation Llama 3 model, the company will purchase more than 350,000 Nvidia H100 GPUs by the end of 2024. In addition to Nvidia A100 and other AI chips, Meta will have nearly 600,000 H100 GPU computing capabilities by the end of the year."$Meta Platforms (META.US)$While waiting for the B200 chip, other technology companies are also snatching up the H100.

In terms of financial statements, Microsoft, Meta, and other companies' investment in purchasing AI chips mainly reflects "purchases of property and equipment". The latest financial report data shows that Microsoft's capital expenditures from Q2 2023 to Q1 2024 (calendar year) totaled 39.547 billion USD, approximately 286.2 billion RMB; Meta's capital expenditures from Q2 2023 to Q1 2024 totaled 26.824 billion USD, approximately 194.6 billion RMB.

Raymond James analyst once estimated that Nvidia's H100 price is between $25,000 and $30,000, and the price on eBay can exceed $40,000. At the unit price of $30,000, the purchase of 350,000 H100 chips by Meta will bring Nvidia about tens of billions of dollars in revenue.

From a financial report perspective, Microsoft and Meta's investment in purchasing AI chips mainly reflects "purchases of property and equipment."

The latest financial report data shows that Microsoft's capital expenditures from Q2 2023 to Q1 2024 (calendar year) totaled 39.547 billion USD, approximately 286.2 billion RMB; Meta's capital expenditures from Q2 2023 to Q1 2024 totaled 26.824 billion USD, approximately 194.6 billion RMB.

On Nvidia's road to challenging the world's number one market cap position, technology companies like Microsoft and Meta spend hundreds of billions every year on snatching up Nvidia chips.

"In addition to AI chips, Nvidia has new products and ecological launches in AIPC, robotics, and space computing. The new products have given the capital market high expectations, which is also reasonable for Nvidia's market cap to set a new high in the short term." Guo Tao analyzed for "AI Light Year."

"Nvidia has no opponents in the next three years."

How long can Nvidia maintain its position as the global market cap leader? When will the situation of Nvidia being the only one dominating the high-end AI chip market be broken? In fact, as the trend of large models and generative AI continues to heat up, more and more manufacturers have begun to make efforts in the AI chip field, and the competition in the AI chip market has become increasingly fierce.

In the AIPC race, Intel released a new architecture called Lunar Lake, which increased its comprehensive AI computing power to 120 TOPS. To compete with Nvidia, AMD Chairman and CEO, Hung Ren-xun's distant relative, Su Zi-feng, also took on the attitude of 'updating once a year' and openly competed with Nvidia for market share.

"AMD will launch Instinct MI325X this year, MI350 in 2025, and MI400 in 2026", Su Zi-feng revealed at Computex2024 exhibition.

In addition, in order to save the cost of purchasing high-end AI computing chips, tech giants such as Google and Amazon are also researching their own heterogeneous chips, such as Google's TPU, Amazon's Trainium, and Microsoft's Maia.

In China, Huawei is a competitor that Nvidia cannot ignore. Hung Ren-xun once stated in an interview, 'Although Nvidia has a dominant position in the AI chip market, the speed of the competitors' catch-up is very fast. Huawei is one of the very powerful competitors.'

Wang Tao, chief operating officer of Huawei's Ascend and Kunpeng departments, also stated that Huawei's Ascend 910B AI chip has successfully surpassed Nvidia's A100 AI GPU by 20% in terms of training performance.

However, the gap between Nvidia and other manufacturers is not limited to computing power itself, Nvidia's entire system includes a complete ecosystem, network, communication and other systems. For example, with CUDA, developers can use the strong parallel processing power of GPUs to accelerate compute-intensive tasks; Nvlink & NVSwitch can solve the problems of cluster collaboration and data transfer between multiple chips; InfiniBand can solve data transmission problems between servers within a data center.

An industry insider once told AI Light Year, 'Due to Nvidia's mature and extensive ecosystem, for data centers that have deployed Nvidia's high-end chips, replacing the chips not only needs to consider the compatibility of the ecosystem, but also the risks of retraining costs, software reconstruction, and performance decline.'

In Guo Tao's opinion, Nvidia's technology and ecological barriers established in the AI computing field are difficult for other manufacturers to break in the short term. Building an ecosystem takes time and cannot be done overnight. 'In the high-end chip field, Nvidia will not have opponents in the next three years,' Guo Tao predicted to AI Light Year.

At the end of May, Nvidia lowered its price for the 'castrated version' of the chip H20 in the domestic market, which is more than 10% cheaper than Huawei's Ascend 910B.

Will the price reduction of chips such as H20 affect Nvidia's stock price? An industry insider revealed to AI Light Year, 'Domestic AI giants have already hoarded Nvidia's high-end chips in advance, and H20 is not of great significance to domestic technology giants for training large models. At present, H20 is mainly purchased by small and medium-sized enterprises to balance computing power resources, so the price reduction of H20 has basically no effect on Nvidia's stock price.'

Edited by Jeffrey