Nvidia's Huang forecasts $4 trillion data-center boom reshaping chip demand

Nvidia Chief Executive Officer Jensen Huang is betting open AI models will fuel a $3 trillion to $4 trillion annual data-center buildout by decade's end, dwarfing Wall Street's consensus by roughly four times.

"Great models lead to great use, which leads to great growth," Huang said in an interview with Axios, arguing that open-weight models expand demand for Nvidia's chips rather than cannibalizing it. "Whenever there will be more use, you'll have to sell more Nvidia computers."

The projection, confirmed by Chief Financial Officer Colette Kress on Nvidia's first-quarter earnings call May 20, compares with industry expectations of about $1.03 trillion in annual hyperscale capital expenditure by 2028, according to Needham analyst Laura Martin. Nvidia alone posted $81.6 billion in revenue last quarter, up 85% from a year earlier, with data-center revenue surging 92% to $75.2 billion. The company guided to $91 billion in second-quarter revenue, roughly 95% higher than the prior year.

If Huang's forecast proves accurate and Nvidia maintains its roughly 81% share of the AI chip market, the path to a $20 trillion valuation becomes straightforward, Morgan Stanley's Joseph Moore said. The company currently trades near $4.9 trillion, neck-and-neck with Apple. Nvidia forecasts $1 trillion in combined sales for its Blackwell and Vera Rubin architectures across 2026 and 2027.

Open Models vs. Washington's Crackdown

The open letter, signed by 25 companies including Microsoft, Meta, Palantir, and Dell, argues that open-weight models preserve competition and security. "Defenders need access to models with comparable capabilities so they can detect, simulate, and respond to emerging threats," the letter states. The diffusion of AI, the letter says, will create a "shared foundation of knowledge" that generations of American engineers have built upon. Notable holdouts include OpenAI and Anthropic, whose business models depend on proprietary frontier models.

Treasury Secretary Scott Bessent has taken a different view, accusing Chinese AI labs like Moonshot AI of "IP theft" through distillation, where a weaker model trains on the outputs of a frontier model. "If we see, especially that overseas models are stealing from our great companies, we have the ability to sanction them," Bessent said on Fox Business. The tension highlights a split between the administration's stated support for open-source and its close ties to frontier labs that benefit from export controls limiting competition.

Huang's push for openness carries an irony not lost on critics. His company controls roughly 81% of the data-center AI chip market, according to IDC's May 2026 estimate, yet he is preaching an open market for everything except the sector he dominates. The other 24 signatories, worth about $10.5 trillion combined or roughly 15% of the S&P 500, appear to believe they are stronger together than apart.

The $4 Trillion Bet and Its Ripple Effects

Bank of America analyst Vivek Arya projects the total addressable market for AI data-center systems will reach about $1.7 trillion by 2030, up from a prior estimate of $1.4 trillion. The 21% revision reflects how quickly spending assumptions are shifting as hyperscalers sign three-year supply contracts, some with full upfront prepayments, according to Morgan Stanley's Moore. Nvidia accounts for approximately 85% of AI processor revenue, with AMD at less than 5% and custom-designed chips from hyperscalers at just above 10%, Moore noted.

The buildout is already producing costs that reach beyond Wall Street. Goldman Sachs analysts forecast consumer electricity inflation will run at about 6% through 2026 and 2027, with data centers accounting for roughly 40% of total electricity demand growth over the next five years. Gartner director analyst Linglan Wang projects worldwide data-center power demand will rise 27% in 2026 alone to 132 gigawatts, potentially reaching 290 gigawatts by 2030. Consumer spending growth could decline by 0.2% through 2027 as higher electricity bills reduce disposable income, the Goldman team projected, contributing to a 0.1% drag on overall economic growth.

One wild card: the potential reopening of Nvidia's Chinese market. Under Secretary of Commerce for Industry and Security Jeffrey Kessler told the House Foreign Affairs Committee on July 14 that "very few" Nvidia H200 chips have been shipped to China, indicating limited sales may resume after years of export restrictions. Huang has said China will remain a leading AI innovator despite different constraints, and if regulatory barriers continue to ease, the Chinese market would add a growth channel Nvidia has entirely excluded from its revenue forecasts.

Nvidia shares, trading near $4.9 trillion in market value, have stalled despite improving fundamentals, weighed down by doubts about the durability of growth, Moore wrote in a March note. If Huang's $4 trillion CapEx thesis holds and Nvidia maintains its roughly 85% share of AI processor revenue, the stock's current valuation would prove conservative. AMD, with less than 5% of the market, and custom-chip efforts from hyperscalers at just above 10%, have so far failed to erode Nvidia's position.

This article is for informational purposes only and does not constitute investment advice.