The AI “fever” has helped Nvidia’s stock skyrocket 25% in trading on May 25, bringing the company’s market capitalization to about $950 billion. Previously, Nvidia’s market capitalization on May 24 was only $755 billion. According to CNBC, if it meets expectations, the chip giant will become the fifth US company to be worth $1,000 billion.
Nvidia is changing the way it builds computers to drive even more profit. Huang Jensen said the components used to build data centers could be a $1,000 market.
The most important component in computers and servers is the central processing unit (CPU). That market is dominated by Nvidia’s rivals Intel and AMD. But with the rise of AI applications that require powerful computing power, Nvidia now dominates the GPU market.
Nvidia Makes a Lot of Money from ChatGPT Boom
Huang said that the data center of the past was mostly CPU-based for accessing files, but in the future it will be general data. Instead of accessing data, you will access some data, but you will generate most of the data with AI. So instead of using millions of CPUs, you will need a lot fewer CPUs, but they will be connected to millions of GPUs, the Nvidia CEO added.
That’s one reason Nvidia’s data center business grew 14% in Q1 2023. Meanwhile, Intel’s data center and AI business saw revenue decline 39% to $3.7 billion, while AMD’s growth was flat.
Additionally, Nvidia GPUs tend to be much more expensive than central processors. Intel’s latest generation of Xeon CPUs can cost as much as $17,000 at list price, while an Nvidia H100 chip can be resold for as much as $40,000 on other platforms like eBay.
Nvidia’s competition will only increase as the AI market heats up. Big rivals AMD and Intel each have their own GPUs, and tech giants like Google and Amazon are also designing AI chips. But Nvidia’s high-end GPUs remain the chip of choice for AI training. Analysts say Nvidia remains ahead in AI chips thanks to its proprietary software that makes it easier to implement AI.
Mr. Huang shared that the company's software will not be easy to copy because you have to design all the software, all the libraries, all the algorithms, integrate them and optimize the frameworks and optimize it for the architecture, not just a chip but the architecture of the entire data center.
Source link
Comment (0)