By Jack Hough

Nvidia stock is depressed, Wall Street says, even though it's up 22,000% in a decade. A teensy stretch of trading weakness this year -- a pause in rip-roaring gains, really -- has left the valuation looking ordinary, despite a profit outlook that has no equal in history. Investors, it seems, have doubts that the next leg of Nvidia's growth will meet expectations. The week ahead provides the company's best opportunity to change their minds.

It's Nvidia GTC 2026, an event so important that the name stands for something that stands for something else. GTC is GPU Technology Conference, while GPU is graphics processing unit, which is a weird name for a gathering that's all about artificial intelligence, but it will make sense in a moment. I'll run through what Wall Street hopes to hear, including future chip changes. If you've written to me recently to say you're tired of so much AI coverage, see my recent stories on beef and Iran. If you've written to ask why I'm spending time on beef and Iran instead of AI, strap in.

First, the stock. Nvidia was down 8% this year at one point in early February, but it has drifted back and was recently down just 1%. Meanwhile, a consensus of estimates has free cash flow for the company rising 85% this fiscal year through January 2027, to more than $178 billion. That growth rate would mark a sharp acceleration from last year. There's a $98 billion gap between the low and high estimates in the consensus, suggesting huge uncertainty, but the overall consensus has been rising in recent months, meaning that forecasters are sweetening their math as new evidence comes in, like Nvidia's late February quarterly earnings report.

Back in 2022, when Russia invaded Ukraine, sending the oil price soaring, Saudi Arabian Oil, better known as Saudi Aramco, generated close to $150 billion in free cash, still a record. This year, Aramco isn't likely to come close to that figure, in part because the Iran war has complicated its path to market. If Nvidia meets the consensus, it will become the most prosperous company ever.

It will owe that to, for example, Amazon.com's cash burn of perhaps $10 billion this year. For the three years ended 2023, Amazon's capital expenditures came in between $50 billion and $60 billion per year, but this year they're pegged at $190 billion, and the increase has everything to do with AI infrastructure, including Nvidia chips. A big part of the debate around Nvidia stock involves how long Amazon and others will continue spending so richly. Barclays this past week predicted that industrywide capex will peak in 2028 at about $1 trillion, and "come down modestly thereafter." It reckons that consensus 2028 estimates for hyperscalers like Amazon are too low by a combined $300 billion.

Next year Nvidia is expected to smash its own free-cash-flow record, hitting $233 billion. That doesn't include the cost of stock compensation, but it's a relatively modest effect. For investors who watch earnings rather than free cash flow, Nvidia trades at 17 times next fiscal year's projected earnings, a discount to the S&P 500 index. Hence the argument that Nvidia's stock price is too low, especially considering the growth outlook. Among 70 analysts who cover the stock, 93% say to buy, according to FactSet. The average price target of $267 and change implies 45% upside.

Nvidia GTC 2026 starts Monday with a keynote presentation from co-founder and CEO Jensen Huang, and runs through Thursday. Investors will want to hear about the supply outlook for critical components like wafers, memory, and optics, and about the effect of the Iran war on power costs and demand from sovereign customers. Mostly, they will want details on future products.

AI chip spending in recent years has focused on building and optimizing models, called training, whereas in coming years it will shift more toward putting those models to work, called inference. Training favors highly parallel processing, much like rendering videogame graphics, which helps explain why the company whose chips powered 3-D shoot-'em-ups like Unreal Tournament back in 1999 came to dominate AI profits. As recently as 2019, CPUs, or central processing units, like those Intel is best known for, made up 87% of data-center compute spending, and GPUs, along with other AI accelerators, contributed 13%, according to BofA Securities. That has flipped: Last year, AI chips accounted for 88% of spending.

In training, cost has been an afterthought, but in inference, where the money will be made, cost is key, and computing needs are mixed. When you type a question into your favorite AI chatbot, it turns it into tokens representing words, parts of words, and punctuation. It processes all of these tokens at once, a step called prefill, which favors the parallel computing of GPUs. But the answers come a token at a time, a bit like speaking, where each word builds on the last. CPUs can excel in this kind of sequential computing, but what you'd really like to have are purpose-built chips that can handle decode cheaply and efficiently, without, for example, the need for pricey off-chip memory.

A privately held company called Groq specializes in just such chips, called LPUs, or language processing units, and last year Nvidia paid about $20 billion to license its technology and acquire its talent. So look for Nvidia to discuss how Groq LPUs will help broaden and customize its future chip portfolio to address training, prefill, and decode. That could help the company hold market share with hyperscalers that can produce their own chips.

UBS, bullish on Nvidia, calls a stock breakout based on thesis-altering commentary in the coming days "hard to see." But it also calls the contrast between its rapturous Nvidia earnings estimates and the stock's discount valuation "seemingly unsustainable." I predict only that Huang will give his keynote in his signature leather biker jacket. If he goes full trench coat, like Neo in The Matrix, consider taking profits.

Write to Jack Hough at jack.hough@barrons.com. Follow him on X and subscribe to his Barron's Streetwise podcast.

This content was created by Barron's, which is operated by Dow Jones & Co. Barron's is published independently from Dow Jones Newswires and The Wall Street Journal.

(END) Dow Jones Newswires

March 13, 2026 01:00 ET (05:00 GMT)

Copyright (c) 2026 Dow Jones & Company, Inc.