Why the AI rally (and the bubble talk) could continue next year

James van Geelen, a market analyst who spends a lot of time thinking about AI’s potential as a disruptive force in markets and business, gets the question a lot. And he’s quick to answer: “If we’re not, we’re going to be,” van Geelen, the founder and CEO of Citrini Research, which specializes in megatrend investing ideas, told DealBook.

“Show me an instance over the past 300 years of a truly transformative technology that didn’t result in an asset bubble — railroads, steam engines, radio, airplanes, the internet,” he said. “When capital floods into a technology because everyone realizes that it’s transformative, eventually you get a bubble.”

If OpenAI’s launch of ChatGPT in November 2022 was the watershed moment that brought AI to the masses and launched a global race to dominate the technology, then 2025 was the year that the technology transformed markets and the global economy in newly powerful ways.

Investment in AI may have accounted for as much as half of U.S. GDP growth in the first half of 2025. President Donald Trump has taken notice. From the start of his second term, the president has made AI superiority a pillar of his business agenda. And he has put state regulators on notice that they’re not to slow down the sector.

Live Events

That economic firepower (and a policy embrace from the White House) has been reflected in the stock market. In mid-December, the so-called Magnificent 7 tech giants — Alphabet, Amazon, Apple, Meta, Microsoft, Nvidia and Tesla — made up 34% of the value of the S&P 500.
In late October, Nvidia, whose semiconductors power much of the world’s biggest AI models, became the first company in history to hit $5 trillion in market valuation. (It’s a mere $4.5 trillion today.) And the Morningstar Global Next Generation Artificial Intelligence index was up roughly 40% in 2025 through mid-December, doubling the tech-heavy Nasdaq Composite index.It’s that outsize performance that has many investors fretting about a bubble.

In Deutsche Bank Research‘s annual survey of global asset managers, 57% of respondents picked waning enthusiasm for AI and a drop in tech valuations as the biggest threat to the bull market rally, easily outpacing concerns about Fed policy.

“I think it is misleading to try to encapsulate everything in the idea of one bubble, because there are at least three bubbles worth talking about in terms of valuations, investment and technology,” Adrian Cox, thematic strategist at Deutsche Bank Research, told DealBook. “And in each of those, I think there is evidence that there is some inflation going on which could at some point become a bubble, which could at some point burst. But at the moment, it still feels as though we’re in the early stages of that process.”

Van Geelen sees the technology becoming a transformative productivity tool for businesses in the near future. He singles out AI in robotics and agentic AI — autonomous systems that can plan, research and work toward goals with limited human supervision — as potential game changers.

“People were talking about this last year, but the technology just wasn’t there,” he said. “And now it really is.”

Building as Fast as They Can

But to get there requires a lot of data centers. That need has unleashed a construction boom for the ages, one that has put serious strain on local resources and has unleashed a torrent of debt-fueled spending.

Electricity consumption at U.S. data centers is projected to more than double last year’s total by 2030, according to the International Energy Agency. That thirst for power is contributing to rising costs for everyday Americas, and could become a major issue in next year’s midterm elections.

In the latest example of this dash to secure electricity supplies, Google’s parent, Alphabet, agreed on Monday to acquire the clean energy developer Intersect Power in a $4.75 billion all-cash deal, with an eye toward powering Google’s data centers.

The sheer volume of investment needed to fund this boom has spooked some on Wall Street. According to Bloomberg, tech giants including Microsoft and Meta have agreed to spend some $500 billion total to lease data centers over the next several years. Oracle alone has committed to $248 billion on such leases — news that caused its stock to tumble earlier this month. More optimistic observers argue that the Mag 7 companies funding much of the build-out have the cash flow to do so without taking on onerous debt loads.

Brian Colello, a senior analyst at Morningstar, doesn’t see any major red flags. In fact, he believes some of the AI build-out’s inherent challenges will help keep the industry from moving too fast. The sheer cost of the chips required to run sophisticated AI models, he contends, as well as the fact that they become obsolete after a few years, means that companies are less likely to overspend for computing power they don’t need. (Contrast that with the dot-com bubble, when billions flowed into laying more fiber-optic cable than the market demanded.) That, and the electricity constraints, could slow the infrastructure build-out. But, ultimately, it won’t slow down the sector.

“We would argue there is no AI bubble to date, and we think it’s unlikely there will be one in 2026 as well,” Colello told DealBook. “We’re seeing AI demand still exceed supply. We’re seeing the hyperscalers increase their capital spending plans, and we’re seeing an AI supply chain doing everything it can to meet this booming demand.”

Tech-Tonic Shifts

If any one company encapsulates the wild enthusiasm for — and deep skepticism about — AI right now, it’s OpenAI. ChatGPT has some 800 million active users. But just a small fraction are paying customers. Sam Altman, OpenAI’s CEO, recently said it was trending toward an annualized revenue run rate of $20 billion by the end of this year. But it’s also planning to spend gargantuan sums on infrastructure: It has committed $1.4 trillion to building data centers over the next eight years.

Despite that accounting disconnect, the startup continues to raise money at eye-watering valuations. In March, SoftBank led a funding round that valued the company at $300 billion. And its most recent financing round, in October, put a $500 billion value on OpenAI. Last week, it was reported that OpenAI is in talks to raise a $100 billion round that would value the company at $830 billion. There’s also buzz about an IPO in 2027.

Even as OpenAI seeks to monetize its enormous user base, the competition is getting tougher. Google’s Gemini 3, which debuted in November, was recently ranked above ChatGPT in industry benchmark performance testing.

Anthropic, the rival startup founded by former OpenAI executives, has focused on enterprise applications. The newest version of its Claude chatbot can work autonomously with little oversight for 30 hours. (When The Wall Street Journal let Claude run a vending machine in its office, however, it lost hundreds of dollars.)

Then there are open-source AI models, like that of the Chinese startup DeepSeek and Alibaba’s Qwen, which are attracting a wave of startups to build with their architecture.

Van Geelen believes there is room for multiple models. And he cautions that what may look like slow progress now in integrating AI-powered tools into everyday business operations could change rapidly.

“Technology progresses at an exponential rate,” he said, “and humans adopt technology at a linear rate.”

But humans are going to need to move faster, because van Geelen thinks that 2026 is the year when AI will begin replacing people in certain jobs, and companies outside Silicon Valley will start reaping the rewards of improved efficiency.

By this time next year, we may be pondering a new question altogether: How big can the bubble get?