How AI helps Nvidia reach the trillion-dollar club

The AI ​​race helped Nvidia reach a market capitalization of nearly a trillion dollars by selling expensive chips for artificial intelligence training systems.

Nvidia last week reported its financials for the first quarter of 2023, mentioning an "incredible increase" in chip demand for next-generation AI systems, and forecast strong growth in the next quarter. according to. Following the announcement, Nvidia's market capitalization increased 24% from $750 billion to $935 billion.

As of the end of May 27, the company's share price reached $ 390, bringing the market capitalization to nearly $ 970 billion, much higher than the figure of $ 400 billion in January.

In just 5 months, the US chip company has surpassed Tesla and Facebook ($584.7 billion and $647.6 billion respectively), and is expected to soon join the trillion-dollar club with the "big" companies. technology Apple, Google, Microsoft and Amazon, and the oil company Saudi Aramco.

Nvidia CEO Jensen Huang introduces the Volta GPU model at the event in Las Vegas 2018. Photo: Reuters

Nvidia was founded in 1993 by three engineers Jensen Huang, Chris Malachowsky and Curtis Priem - who predicted that as computers grew, they would need more complex graphics processing units (GPUs). The bet paid off when in 2000, the company won its first big deal: to make chips for Microsoft's Xbox game consoles.

Since then, Nvidia has gained more attention for its GPU products for computers. As the wave of cryptocurrency mining unfolded over the years, the American company continued to benefit from the thirst for graphics cards.

At the same time, a number of companies in the AI ​​field also purchased chips from Nvidia to train large language models (LLMs). However, the company only really became a bright star from the first half of this year when the super AI craze exploded and a series of large and small companies hunted for two chip models A100 and H100 .

"Training an AI model requires chip models capable of processing large amounts of data at a high frequency. Nvidia happens to be the only company that can meet that need," said Greg Osuri, founder. founded consulting firm Akash Networks, commented.

Most large and small companies, if they want to build generative AI models , need Nvidia chips. "When a business wants to come up with an AI-driven business model, the first thing to do is send Jensen a check," David Luan, a former Google engineer and co-founder of Adept Labs, said at the AI ​​for Startups event back. March in the US, alluding to the need to pay for chips to Nvidia CEO Jensen Huang.

According to Osuri, a set of eight of Nvidia's most advanced chips — the bare minimum for an AI training system — can run up to $300,000. But building a super AI training machine will require thousands of chips. According to Business Insider , Elon Musk is said to have recently purchased 10,000 Nvidia chips to fuel his new AI ambitions.

According to experts, Nvidia has benefited since OpenAI's ChatGPT exploded late last year. Essentially, ChatGPT is like the starting gun for the AI ​​arms race. During a meeting with investors last week, Colette Kress, Nvidia's CFO, called the appearance of ChatGPT an " iPhone moment ".

The dependence on Nvidia's GPU is the reason for analysts to believe that the US company will soon join the trillion-dollar club in the near future. In the first half of this year alone, its shares have increased more than 2.5 times.

However, the future Nvidia will face great competition because Google, Microsoft and Amazon have all developed their own chips to process AI data, in the same way that Apple has created chips for iPhones and MacBooks to not be dependent on third parties. father. Two big competitors AMD and Intel have also expanded the GPU line.

However, over a period of several years, Nvidia will still be the dominant name in the GPU market. The company's chip architecture is considered difficult to copy because it involves a series of factors such as software design, libraries, algorithms, integration and optimization into a single system.

"I think Nvidia's dominance will continue," said CJ Muse, an analyst at semiconductor research firm Evercore.

Bao Lam (according to Washington Post )

Nvidia tweaks AI chips to sell to China 23
AI chip sells for $46,000 on eBay 40
The $10,000 chip behind the global AI race sixty one