Nvidia (NVDA +2.93%) is the most important company in artificial intelligence (AI) today, the near-monopoly supplier of the chips that train the world's AI models. But five years from now, I suspect it will look like a very different business. The forces reshaping it are already visible if you look closely.
Right now, the overwhelming majority of Nvidia's revenue comes from selling powerful GPUs to a handful of giant cloud companies building AI. Over the next five years, that mix should broaden in important ways.
The market for running AI models, known as inference, is likely to dwarf the market for training them, and it favors a wider range of customers. Nvidia is also quietly building a software business on top of its hardware, from its CUDA platform to enterprise AI tools, and recurring software revenue tends to be stickier and higher margin than one-time chip sales. A company that leans more on software and inference is a steadier one.
Image source: Getty Images.
New frontiers beyond the data center
The bigger transformation may come from entirely new markets. Nvidia's leadership talks constantly about "physical AI," meaning robots, factories, and self-driving cars that need its chips to see and think. Add sovereign AI, where entire countries build their own computing infrastructure, and you can see Nvidia's growth spreading well beyond the cloud giants that dominate its sales today. If those bets pay off, Nvidia will become less dependent on any single customer group, which lowers its risk.
NASDAQ: NVDA
Key Data Points
The competitive and cyclical reality
Now the sober side. The same customers buying Nvidia's chips are designing their own to reduce their reliance on it, and rivals are pushing hard to take share. The extraordinary profit margins Nvidia enjoys today will likely come down as competition intensifies. And the AI spending boom cannot climb straight up forever, since chips are a cyclical business. In five years, Nvidia may well be much larger yet growing more slowly, with more competitors nibbling at the edges of its empire.
Put it together, and Nvidia five years out probably looks like a bigger, more diversified compute platform spanning inference, software, robotics, and global infrastructure, rather than the training-chip juggernaut it is now. That could make it more durable but also slower-growing and less dominant.
The real question for investors is whether Nvidia can broaden its moat faster than rivals erode it. I am optimistic because its ecosystem is deeply entrenched, but I would own it knowing the company and the stock will not be the same as they are today.
