When a CEO’s private life turns into a national-security issue, you know his company matters. That’s where Nvidia is right now: Its founder and CEO Jensen Huang has admitted to an affair with a woman who U.S. intelligence officials reportedly believe is a Chinese spy. Nvidia’s board has launched an independent investigation, and according to President Trump, the FBI is looking into it as well.
Why would anyone care about this story? Because Nvidia makes most of the chips that today’s AI runs on. Much of the computing behind ChatGPT, Gemini and similar tools happens on Nvidia hardware. That has made it the second most valuable company in the world, right after Apple.
In this article, I’ll give you more details about what Nvidia does and where it comes from.
Nvidia at a glance
A few numbers show how big Nvidia has become:
- Market value: about $4.3 to 4.5 trillion (early November 2026). Only Apple is worth more.
- Revenue: $96.2 billion in a single quarter (May to July 2026), more than twice as much as a year earlier.
- Profit: about $60 billion in that same quarter. That’s roughly what the company made in revenue in all of 2023.
- Focus: about 92 percent of its revenue now comes from AI hardware for data centers. Gaming, where Nvidia started, has become a side business.
- Market position: estimates put its share of all spending on AI chips at around three quarters or more.
- Background: founded in 1993 in California, about 42,000 employees. Jensen Huang has been CEO from day one.
What Nvidia actually sells
Nvidia’s core product is the GPU, short for graphics processing unit. It was originally designed to draw video game graphics, and it works differently from the main processor in your laptop.
Think of a normal processor as a handful of brilliant professors: they can solve any problem, but only one at a time. A GPU is more like a stadium full of students, each doing simple arithmetic, all at the same moment. AI happens to need exactly that. Under the hood, an AI model mostly multiplies huge tables of numbers, trillions of times. Each step is simple, but there are a lot of them.
AI needs this computing power twice. First, a model has to be trained: it learns from vast amounts of data over weeks or months. We’ve explained this process in another article. Then it has to answer users’ questions, which is called inference. Every prompt you type into a chatbot keeps chips busy somewhere in a data center.
Nvidia’s lead isn’t only about the chips, though. Two more things matter at least as much:
- Software: Nvidia’s CUDA platform lets developers use its GPUs for all kinds of calculations, not just graphics. Over almost 20 years, it became the standard language of AI development, long before ChatGPT arrived. The catch: it only works on Nvidia hardware, and switching means rewriting and retesting a lot of code. This is how the company has been able to keep its competitors at bay (more about that below).
- Networking: Today’s AI models are far too big for a single chip. They run on thousands of GPUs that constantly exchange data. Nvidia also supplies the connections that let all those chips work together like one giant computer.
That’s why Nvidia now sells complete server racks, the building blocks of what it calls “AI factories,” rather than just individual chips. Its biggest customers are the large tech and AI companies such as Microsoft, Meta, Google, Amazon, Oracle and OpenAI. Nvidia doesn’t name them, but its filings show that just three customers accounted for more than 40 percent of its sales in the first half of 2026.
Where Nvidia comes from
Nvidia’s story starts in 1993 in a booth at a Denny’s restaurant in San Jose, California. Jensen Huang, then 30, and two engineers from Sun Microsystems, Chris Malachowsky and Curtis Priem, planned a company there that would bring realistic 3D graphics to PC games.
The beginnings were rocky. Nvidia’s first product flopped, and by the mid-1990s the company was reportedly down to about 30 days of cash. A new graphics chip, the RIVA 128, saved it in 1997. A few years later, its GeForce cards made Nvidia the brand of choice for gamers. Still, “Our company is thirty days from going out of business” became Huang’s long-running motto.
The decisive bet came in 2006 with CUDA, the software mentioned above. Huang had it supported on every GPU Nvidia shipped, at considerable cost and with hardly any customers for years. He later said it took about ten years before Wall Street believed the investment was worth anything.
The turning point came in 2012. Researchers at the University of Toronto trained an image-recognition AI on two ordinary Nvidia gaming cards. Their system, called AlexNet, beat the competition by a wide margin and kicked off the modern AI boom. One of the researchers, Ilya Sutskever, later co-founded OpenAI. From then on, AI labs everywhere wanted Nvidia’s GPUs, and Nvidia leaned into AI years before most people had heard of it. When ChatGPT launched in November 2022, the company was ready.
The stock market followed. In 2012, Nvidia was worth less than $10 billion. In July 2025, it became the first company ever to pass $4 trillion, and in October 2025 the first to pass $5 trillion. It has slipped somewhat since, but the takeaway is clear: Nvidia’s position isn’t a lucky accident. It’s the payoff of a long bet that, for years, looked like a waste of money.
Is Nvidia’s lead safe?
While Nvidia’s position is strong, it’s no longer unchallenged. Its estimated share of AI-chip spending has slipped from around 87 percent in 2024 to roughly 75 to 81 percent this year. At the same time, its sales keep growing quickly, because the overall market is expanding even faster.
The most serious competition comes from Nvidia’s own biggest customers. Google has been building its own AI chips, called TPUs, for years and now offers them to others: Anthropic, the company behind Claude, agreed to use up to a million of them, and Meta signed a deal as well. Amazon has its Trainium chips, Microsoft and Meta are developing their own, and OpenAI is building a custom chip with Broadcom. Among traditional chipmakers, AMD is the main rival and has landed large deals with OpenAI and Meta. Still, all of these companies continue to buy Nvidia chips in large volumes.
Then there are the bigger questions. Some critics see an AI bubble. Others point to “circular deals”: Nvidia invests in customers such as OpenAI and CoreWeave, who then spend money on Nvidia hardware. Investors seem cautious: Nvidia’s stock has gone sideways in 2026, even though its revenue has doubled.
The China problem
For the U.S. government, AI computing power is no longer just a business matter. It’s a strategic resource, a bit like oil in the 20th century: whoever has the most powerful chips has the best tools to build the most capable AI, with consequences for the economy and for national security.
That’s why, since 2022, the U.S. has restricted the export of Nvidia’s most advanced AI chips to China. The rules have shifted several times since. In January 2026, the Trump administration started allowing case-by-case sales of Nvidia’s older H200 chip to China, with the U.S. government taking a 25 percent cut.
So far, that has hardly made a difference. Beijing hasn’t approved large purchases and is steering Chinese companies toward domestic chips, mainly from Huawei. Those are well behind Nvidia’s, but good enough to fill a market Nvidia has largely lost. China used to matter a lot to the company. Today, Nvidia’s business outlook assumes zero data-center revenue from there.
This is why the current headlines are taken so seriously: U.S. officials reportedly suspect that Huang’s affair was part of a Chinese intelligence effort, possibly aimed at getting Nvidia’s advanced chips into the country. China denies this.
What this means for you
Even if you never buy an Nvidia chip yourself, you rely on this company’s hardware, directly or indirectly, whenever you use ChatGPT, Claude, Gemini or Copilot for your work. How many chips Nvidia can deliver, what they cost and where they may be sold all influence how quickly these tools improve, how much they cost to use and how much computing power is available to run them.
If you want to dig deeper, Stephen Witt’s book “The Thinking Machine” (2025) has been widely praised for telling Nvidia’s story as a very readable narrative. Tae Kim’s “The Nvidia Way” (2024) focuses on the company’s history and culture.
Stay up to date
AI for content creation: the latest tools, tips and trends. Every two weeks in your inbox:
