Adam, 44, decided to invest in Nvidia last month after receiving “hot tips” from a friend, but had only bought shares in one other company before.
“This is AI and there's clearly money involved,” said Adam, who works in London's hospitality industry and asked not to be named because his family are unaware of his stock trading.
“This is the future. It's from Cyberdyne Systems,” Adam said, struggling to remember the company's name, how it's pronounced (en-vid-iya), or even what exactly the artificial intelligence does, but he alluded to the world-ending AI company in the “Terminator” movies. “People are kind of freaking out.”
While Nvidia has only recently begun to capture the public's imagination, the company has been catching the attention of Wall Street: This week, the 31-year-old chipmaker briefly became the world's most valuable company, with a market capitalization of $3.3 trillion, overtaking Apple and Microsoft.
You're seeing a snapshot of an interactive graphic. This may be because you're offline or have JavaScript disabled in your browser.
The company's graphics processing unit is generally considered the best way to build large-scale AI systems by the likes of Meta and Microsoft, and explosive demand for it has seen the company's shares rise about 700% since the launch of OpenAI's hit chatbot ChatGPT in November 2022.
The unprecedented rise of a company that was until recently unknown to most outside the tech industry reflects the AI fever that has gripped Silicon Valley and Wall Street alike, but its return to third place in just a few days highlights the fierce competition in this nascent technology field.
Nvidia's meteoric rise is emblematic of the AI economy: explosive growth, investor appeal and an unpredictable future. The company's future moves will reflect, and perhaps define, the course of the AI economy.
The company finally Nvidia, a relatively unknown brand, achieved this status in March 2000, when network equipment maker Cisco overtook Microsoft at the height of the dot-com bubble.
Now, as then, companies are pouring billions of dollars into building an infrastructure that promises to revolutionize not just computing but the global economy. Like Nvidia, Cisco struck gold by selling digital picks and shovels to internet prospectors. But the company's stock price has never returned to its 2000 peak since the bubble burst later that year.
The fact that big tech companies' surge in AI capital spending is based more on revenue projections than actual revenues is stoking fears that history could repeat itself.
262%NVIDIA's most recent quarterly revenue growth over last year
“I understand the concern,” says Bernstein analyst Stacey Rasgon, but there's a crucial difference: “Cisco's concern is that they built out a ton of capacity for what they expected to be demand, and today they have fiber in the ground that's never been used.”
Rasgon added that compared with Cisco's stock price at the height of the dot-com bubble, Nvidia's stock is trading at a much lower multiple of forward earnings.
While companies like Microsoft have already seen some profits from their investments in AI chips, others like Meta have warned that it will take longer for the benefits to kick in. Rasgon added that even if an AI bubble is forming, its collapse doesn't seem imminent.
Cisco's dot-com rise and fall contrasts with Apple and Microsoft, two technology companies that have long competed for the top spot on Wall Street not just by making wildly successful products but also by building platforms that underpin vast business ecosystems. Apple says its App Store has about 2 million apps and generates hundreds of billions of dollars in developer revenue each year.
Nvidia's economics are very different from those surrounding Apple. In many ways, the popularity of one app, ChatGPT, has been a major factor in driving Nvidia's stock price up over the past few months. The chipmaker says its software ecosystem includes 40,000 companies and has 3,700 “GPU-accelerated applications.”
Nvidia became the world's most valuable company by selling relatively small numbers of expensive AI chips, mainly to a handful of companies, for data centers, instead of selling hundreds of millions of affordable electronic devices to the general public each year.
Nvidia said last month that big cloud-computing providers such as Microsoft, Amazon and Google account for nearly half of its data-center revenue. Nvidia sold 3.76 million graphics processing unit chips to data centers last year, according to chip analyst group TechInsights. Still, it has a 72% share of this specialized market, far ahead of rivals such as Intel and AMD.
Still, sales are growing fast: Nvidia's revenue rose 262% year over year to $26 billion in the most recent quarter through April, faster than Apple's growth rate in the early days of the iPhone's launch.
You're seeing a snapshot of an interactive graphic. This may be because you're offline or have JavaScript disabled in your browser.
Demand for Nvidia's products has been fueled by technology companies looking to deploy its chips to answer questions about AI capabilities.
Aiming for the next leap in machine intelligence, companies including OpenAI, Microsoft, Meta, and Elon Musk's startup xAI are racing to build data centers that would connect up to 100,000 AI chips to supercomputers, three times the size of today's largest clusters. These server farms would cost $4 billion each for the hardware alone, according to chip consultancy Semianalysis.
The appetite for more computing power for AI is palpable: Nvidia CEO Jensen Huang predicts that more than $1 trillion will be spent over the next few years to retool existing data centers and build “AI factories” as everyone from big tech companies to nation states build their own AI models.
The scale of the investment That won't continue unless Nvidia's customers figure out how to make AI profitable for themselves, and at the very moment the company hits a stock market high, more people in Silicon Valley are questioning whether AI can live up to expectations.
In a blog post this week, David Kahn, a partner at Sequoia, one of Silicon Valley's largest startup investors, warned about a “speculative fever” around AI and the “delusion” that advanced AI and stockpiles of Nvidia chips will “allow us all to get rich quick.”
Khan predicts that AI will generate huge economic value, but estimates that big tech companies will need to collectively generate hundreds of billions of dollars in new revenue per year to recoup their investments in AI infrastructure at the current accelerated pace. Companies like Microsoft, Amazon Web Services, and OpenAI typically expect generative AI to boost their revenue in the billions this year.
The days when tech executives could make grand promises about AI capabilities “are coming to an end,” said Euro Beinat, global head of AI and data science at Prosus Group, one of the world's largest technology investors. “Over the next 16 to 18 months, we'll start to have a more realistic view of what we can and can't do.”
Nvidia is unlikely to become a mass consumer company like Apple, but to continue to thrive, analysts say it needs to follow the iPhone maker's lead and build a software platform that ties business customers to its hardware.
“The case that Nvidia isn't just going to blow up and become a Cisco-like entity once the hardware hype cycle is over needs to be tied to the software platform,” said Ben Bajarin of Silicon Valley-based consulting firm Creative Strategies.
Hwang has long argued that Nvidia is not just a chipmaker. Rather, he said, the company provides all the pieces to build an “entire supercomputer.” That includes chips, networking equipment and the CUDA software that allows AI applications to “talk” to the company's chips and is seen by many as Nvidia's secret weapon.
In March, Huang unveiled Nvidia Inference Microservices (NIM), a set of off-the-shelf software tools that enable companies to more easily apply AI to specific industries or fields.
Huang said these tools can be understood as an “operating system” for running large-scale language models like the ones that underpin ChatGPT. “My guess is that we'll produce NIM at a very large scale,” he said, predicting that the company's software platform, named Nvidia AI Enterprise, “will become a very big business.”
Nvidia has offered its software for free up until now, but now it plans to charge companies a subscription to Nvidia AI Enterprise, charging $4,500 per GPU per year — a move that's crucial to winning over more corporate and government customers that don't have the in-house AI expertise of big tech companies.
The problem for Nvidia is that many of its major customers want to “own” developer relationships and build their own AI platforms. Microsoft wants developers to build on its Azure cloud platform. OpenAI has launched a GPT Store modeled after the App Store, offering a customized version of ChatGPT. Amazon and Google have their own developer tools, as do AI startups Anthropic, Mistral and others.
It's not the only way Nvidia is competing with its biggest customers: Google has developed a custom AI accelerator chip, the Tensor Processing Unit, and Amazon and Microsoft have followed suit with their own chips, though on a smaller scale. The TPU in particular shows that customers can become less reliant on Nvidia.
Nvidia, meanwhile, is nurturing potential future rivals for big tech customers as it looks to diversify its ecosystem. It supplies its chips to the likes of Lambda Labs and CoreWeave, cloud-computing startups that specialize in AI services and that rent out access to Nvidia's GPUs, as well as to local companies, such as France-based Scaleway, rather than multinational corporations.
The moves are part of a broader acceleration of Nvidia's investment activity across the burgeoning AI tech ecosystem: In the past two months alone, the company has participated in funding rounds for data-labeling company ScaleAI, which raised $1 billion, and Mistral, a Paris-based OpenAI rival that raised €600 million.
Nvidia has closed 116 such deals in the past five years, according to data from PitchBook. Investing in startups not only offers potential financial returns, but also a way to get an early look at what the next generation of AI will look like, helping shape the company's product roadmap.
“[Huang] “He knows the nitty-gritty details of AI trends and what they mean,” said Kang-jun Chiu, CEO of Imbue, an AI research lab that Nvidia invested in last year. “He's put together a large team that works directly with the lab so that he can understand what they're trying to build, even if they're not his customers.”
It's this kind of long-term thinking that has put Nvidia at the center of the current AI boom. But Nvidia's path to becoming the world's most valuable company has included several near-death experiences, Mr. Huang said, adding that no company is guaranteed survival in Silicon Valley's cutthroat market.
