After seeing its stock price almost triple in one year, the semiconductor specialist Nvidia entered the very exclusive club of companies whose capitalization exceeds $2,000 billion at the end of February, a level reached by only three other firms so far: Apple, Microsoft and the oil company Saudi Aramco.
Driven by the colossal needs of artificial intelligence in electronic chips, the Californian company has established itself in record time as a tech giant, surpassing the behemoths Google, Amazon and Meta on the stock market.
This upward trajectory dates back to the release of ChatGPT in November 2022. The success of OpenIA's conversational robot with the general public makes artificial intelligence the new goose that lays the golden eggs for investors and pushes tech companies to develop new language models (LLM) that are ever more demanding in computing power.
Also read: When North Korean cybercriminals take drugs on ChatGPT
A boon for Nvidia which, since the early 2000s, has dominated the market for graphics cards (GPUs) popular with video game enthusiasts looking for detailed graphics and smoother gameplay.
“These Nvidia chips which, originally, were made for video rendering have been diverted from their use because they happen to be very well optimized for the algorithms used in artificial intelligence,” explains Laurent Daudet, co-founder of the start -up LightOn.
Nvidia's success is down to a double bet won by its visionary boss, Jensen Huang.
At the beginning of the 1990s, he was one of the few to believe in the democratization of video games which, at the time, was a niche market, reserved for a privileged few.
At the turn of the 2000s, the success of its GeForce series of graphics cards and that of the PlayStation 2 – which remains to this day the best-selling home console in the world – proved him right.
The second bet is called Cuda.
In 2006, Nvidia developed a new computing architecture intended to appeal to the world of finance.
But the very expensive project is scaring away investors.
This is without counting the advent of AI.
Praised by researchers, Cuda allows Nvidia to return to favor in the eyes of financiers and deliver its first supercomputer in 2016, the DGX-1.
“The managers of Nvidia have very well anticipated this wave of AI with specialized processors which know how to do what we call matrix calculations in very large dimensions. Their big advantage is also to have developed a perfectly adapted programming language. It's a 'hardware/software' combination which means that they have this almost monopolistic position today", believes Laurent Daudet.
Nvidia Almighty
Nvidia is estimated to control 80% of the market for advanced AI chips.
In this gold rush, the American company is selling "shovels and picks", summarizes Alexandre Baradez, analyst at IG France.
“All the tech giants are forced to go through Nvidia to order components for data processing, storage and artificial intelligence because they are much more powerful than those offered by competitors. Their competitive advantage is this unrivaled stadium. It’s an ogre that devours everything,” adds the expert.
Symbol of this domination, the H100 chip that all tech has been chasing since last year.
Unit price: 37,000 euros.
According to Nvidia, this power monster, made up of 80 billion transistors
,
allows AI models to be trained nine times faster than average.
Mark Zuckerberg announced in January that Meta would have 350,000 by the end of the year.
Billionaire Elon Musk reportedly bought around 10,000 himself so he could develop a language model competing with ChatGPT.
And the Californian company intends to maintain its technological lead.
During 2024, the new generation of chips dedicated to AI, the H200
,
will get to work in several supercomputers.
See alsoAnd the most powerful supercomputer in the world is...
But this quasi-monopoly position is not without posing a certain number of problems, believes Laurent Daudet.
“First there are the questions of delay. To buy such cards, you sometimes have to wait up to a year for delivery. So this slows down the progress that can be made in the field of AI. And then there are also the very high prices which form a sort of barrier to entry, really very high for new entrants,” regrets the entrepreneur.
This hegemony may not last forever, however.
Nvidia's insolent growth is starting to annoy and several tech giants have announced plans to move away from their dependence on components from the chameleon brand.
According to Bloomberg, OpenAI boss Sam Altman is trying to raise up to $7 trillion to design an empire dedicated to chip manufacturing.
Microsoft also plans to manufacture in-house chips by partnering with AMD, Nvidia's direct competitor.
Finally, Google already offers its Tensor Processing Units (TPU) to accelerate the rate of calculations linked to artificial intelligence.
"This monopoly situation is viewed with jealousy by many large manufacturers who are trying to catch up, notably AMD and its MI300X chips, which are very competitive. From our point of view, it would be healthier for the entire market if it there are several alternatives", argues Laurent Daudet.
But according to experts, Nvidia should retain its throne for several more years.
“As it stands, I think the only risk for Nvidia is to face supply problems,” considers Larry Tentarelli, of the Blue Chip research firm Daily Trend Report, interviewed by AFP.
Because unlike its rivals Intel, Micron or Texas Instruments, Nvidia, like AMD, does not manufacture its own semiconductors.
It uses subcontractors, mainly the Taiwanese TSMC (Taiwan Semiconductor Manufacturing Company).
Too early, too expensive?
Today, only Microsoft and Apple are doing better than Nvidia on Wall Street.
An insolent growth which may come as a surprise when the company is not even among the top 150 in the world based on turnover, and barely in the top 1,000 in terms of number of employees.
“Its turnover of 60 billion dollars is growing very strongly. It is true that it is little compared to behemoths like Amazon. But Nvidia's big advantage, and this is what explains its value on the stock market , is that there are absolutely titanic gross margins, currently 74%. So when you sell an item for one euro, you have 74 cents left. This is enormous profitability for a company like this." , explains Alexandre Baradez.
“If Nvidia maintains the pace of growth expected by the market, it is possible that within two years, perhaps sooner, the company will exceed Apple's capitalization
,
” anticipates Larry Tentarelli.
"On the other hand, it must be emphasized that a large part of the turnover is linked to its 'data centers' activity and is based on the renewal of hardware to increase the power of the servers. But once this transformation will have been carried out by large service providers like Amazon and Microsoft, at what pace will this growth continue?" asks Alexandre Baradez.
Also readDiscrimination, manipulation, job destruction... The biggest risks linked to AI
Finally, the other big unknown is the promises of AI itself.
Will it really increase the productivity of businesses and individuals?
Will the big players in the sector succeed in correctly monetizing their services?
If some experts have mentioned a risk of a speculative bubble like the one that affected technology stocks in 1999, the parallel is more to be found in the period of the Internet's emergence a few years earlier, says Alexandre Baradez.
“At that time, we paid too early and too dearly for something which was indeed a revolution but which took several years to become profitable,” analyzes the expert.
"Today, I find that there are characteristics that take us back a few years. The market is paying in anticipation of very positive things regarding AI. It remains to be seen whether the pace of events will be as is anticipated by investors."
The France 24 summary of the week
invites you to look back at the news that marked the week
I subscribe
Take international news everywhere with you!
Download the France 24 application