When Jensen Huang spoke at Nvidia’s annual general meeting last week, he made no mention of any stock price decline.
The US chipmaker, boosted by its key role in the rise of artificial intelligence, had briefly become the world’s most valuable company on June 18 but the crown quickly slipped. Nvidia shed around $550bn (£434bn) from the peak market value of $3.4tn (£2.68tn) it had reached that week as tech investors, combining profit-taking with doubts about the sustainability of its breakneck growth, applied the brakes.
Yet Huang spoke like the CEO of a company that took 30 days this year to go from a $2 trillion valuation to $3 trillion — and sees $4 trillion just around the corner.
He described an upcoming group of powerful new chips, called Blackwell, as potentially “the most successful product in our history” and perhaps in the entire history of the computer. He added that the new wave of AI would automate $50 trillion of heavy industry and described what sounded like an endless loop of robotic factories orchestrating robots that “build products that are robotic.”
In closing, he said: “We have reinvented Nvidia, the computing industry and, quite possibly, the world.”
These are the kinds of words that underpin a $4 trillion valuation and the AI cycle. Nvidia stock is on the rise again, breaking above $3 trillion this week, because it remains the best way to buy stocks in the AI boom. Will that be enough to propel it to $4 trillion despite investor doubts?
Alvin Nguyen, a senior analyst at research firm Forrester, said “only a collapse of the genAI market” would prevent Nvidia from reaching $4 trillion at some point, but whether it would get there before its tech rivals was another question. Currently, Microsoft (another big AI player) and Apple are first and second respectively in terms of market size, with Nvidia in third place.
If OpenAI’s next big AI model, GPT-5, and other new models were surprising, the stock price would remain buoyant and could reach $4 trillion by the end of 2025, Nguyen said. But if they failed to meet expectations, the stock price could suffer, given its status as a flag carrier for the technology. A technological breakthrough could result in less computing power being needed to train models, he added, or that business and consumer interest in generative AI tools was less robust than expected.
“There are a lot of unknowns that are outside of Nvidia’s control that could impact its path to $4 trillion,” Nguyen said. “Such as disappointment with new models coming out, model improvements that reduce computational needs, and weaker-than-expected demand from businesses and consumers for genAI products.”
Private AI research labs like OpenAI and Anthropic (the entities behind chatbots ChatGPT and Claude) aren’t listed on public markets, leaving large sums of money floating around in investors’ accounts with no way to access some of the heavyweights of the generative AI frenzy.
Buying shares in multinationals like Microsoft or Google is already expensive, and only a fraction of the investment is tied to the latest new thing. There could be a huge boom in artificial intelligence, but if, say, Google’s search ad business faltered as a result, the company would not necessarily be a net winner.
Nvidia, by contrast, is selling shovels in a gold rush. Despite years of investment in capacity, it continues to sell its high-end chips faster than it can manufacture them. Huge proportions of cutting-edge AI research investments flow straight out of the labs and into Nvidia’s coffers, and companies like Meta are committing billions of dollars of spending to secure hundreds of thousands of Nvidia GPUs (graphics processing units).
That type of chip, the company’s specialty, was once marketed as allowing gamers to experience crisp, fluid graphics in 3D games, and through a monumental stroke of luck, it turned out to be exactly what cutting-edge researchers needed to build massive AI systems like GPT-4 or Claude 3.5.
GPUs are capable of performing, at high speed and in high volume, the complicated calculations that underpin the training and operation of AI tools such as chatbots. So any company that wants to build or operate a generative AI product, such as ChatGPT or Google’s Gemini, needs GPUs. The same is true for deploying freely available AI models, such as Meta’s Llama, which also require large numbers of chips as part of their training phase. For systems known as large language models (LLMs), training involves analyzing huge blocks of data. This teaches the LLM to recognize patterns in language and to assess what the next word or sentence should be in response to a query from the chatbot.
However, Nvidia has never managed to fully corner the market for AI chips. Google has always relied on its own chips, which it calls TPU (for “tensor,” a characteristic of an AI model), and others want to join in. Meta has developed its Meta Training and Inference Accelerator, Amazon offers its Trainium2 chips to companies using AWS (Amazon Web Services), and Intel has produced the Gaudi 3.
None of the big rivals compete with Nvidia, yet, at the absolute top end. But that’s not the only place where there’s competition. A report from informationA tech news site, he highlighted the rise of “batch processing,” which offers companies cheaper access to AI models if they’re happy to wait for their queries to run during periods of low demand. That, in turn, allows vendors like OpenAI to buy cheaper, more efficient chips for their data centers instead of focusing all their spending on the fastest possible hardware.
After the newsletter promotion
At the other end of the spectrum, smaller companies are starting to offer increasingly specialized products that outperform what Nvidia can offer in a head-to-head race. Groq (not to be confused with Grok AI, the similarly named Elon Musk project whose launch was recently announced) sparked an ongoing trademark dispute) makes chips that can’t be used to train AI, but run the resulting models at incredible speed. Not to be outdone, startup Etched, which just raised $120 million, is building a chip that runs only one type of AI model: a “transformer,” the T in GPT (generative pretrained transformer).
Nvidia doesn’t just need to hold its own against competition, big or small. To reach the next milestone, it needs to thrive. Market fundamentals are already outdated, but if the company were valued as a traditional low-growth company, even a $3 trillion market cap would require it to sell a trillion of its high-end GPUs a year, at a 30% profit margin, forever, one expert noted.
Even if the AI industry grows enough to justify that, Nvidia’s profit margin may be harder to defend. The company has the chip designs to maintain its lead, but the real bottlenecks in its supply chain are the same as in much of the rest of the industry: at advanced semiconductor foundries, the kind operated by Taiwan’s TSMC, America’s Intel, China’s SMIC, and a few others around the world. Notably, Nvidia isn’t on that list, since it’s a customer of TSMC. No matter how advanced Nvidia’s chipsets are, if it needs to eat through the rest of TSMC’s order book to meet demand, then profit will inevitably flow that way, too.
Neil Wilson, chief analyst at brokerage Finalto, said the bearish case against Nvidia (market lingo for a sustained drop in share price) was based on the argument that once the company processed its order book, it would return to less frenetic levels of demand.
“All of their customers have been rushing to order GPUs, but they won’t be doing that forever,” Wilson said. “Customers overorder and then start canceling. We’re in a sweet spot now, but you can’t sustain it.” He could see Nvidia reaching $4 trillion and beyond, but “maybe not at the current pace.”
Jim Reid, head of global economics and thematic research at Deutsche Bank, published a note this week asking whether Nvidia was “the fastest-growing large company of all time.” Noting that Nvidia went from $2 trillion to $3 trillion in 30 days, Reid said, by contrast, that it had taken Warren Buffett 60 years to get Berkshire Hathaway close to $1 trillion.
But in a world of low productivity (a measure of economic efficiency), declining working-age populations and rising government debt, the economic promise of AI was welcome, Reid said.
“If AI is the catalyst for a fourth industrial revolution, that would be very good news,” he wrote. “If not, markets will end up in big trouble.”
There’s more at stake than winning the race to $4 trillion.