Bubble Fears Emerge About Trillion-Dollar AI Bet Fueling One Of Real Estate's Hottest Sectors

Big Tech’s push to build the infrastructure for generative artificial intelligence has come with an unprecedented trillion-dollar price tag and has made data centers one of the hottest real estate sectors. 

But a growing chorus of skeptics over the last month have said this massive wave of spending is unlikely to pay off, and that the clock is ticking toward the end of the “AI bubble.”

Firms like Amazon, Microsoft, Google and Meta are locked in an AI arms race, driven by the conviction that the world is in the early innings of a technological shift that will touch every corner of the economy. These companies have bet big on the transformative potential of artificial intelligence, spending tens of billions of dollars annually developing the infrastructure to support AI technologies that they believe will drive paradigm-shifting opportunities akin to the creation of the internet itself.

Wall Street traders, private equity firms and major institutional investors have been on board with this vision, funding a data center building boom and fueling a stock market rally that has pushed major indexes to record heights.

But doubt may be starting to creep in. A series of new reports from top institutions like Goldman Sachs and MIT say hope and hype have created an AI bubble, and that it is only a matter of time until it starts to deflate. This narrative has crept into the mainstream press, with Forbes and New York Magazine this month both publishing columns suggesting the AI boom is a "bubble."

“A new AI bubble has formed, fueled by dollars in a mad rush to avoid FOMO,” James Berman, founder of investment advisory firm JBGlobal.com LLC, wrote in the Forbes column, adding that it will play out the same way as the dot-com bubble that burst in 2000. 

According to the critics of Big Tech’s infrastructure push, there is little evidence that AI applications will create enough revenue to justify the massive costs of the infrastructure needed to support them. It’s not even clear what use cases could potentially provide a return on more than a trillion dollars of spending, they say, with projections of AI’s transformative economic impact based on little more than faith.

“AI bulls seem to just trust that use cases will proliferate as the technology evolves,” said Goldman Sachs Head of Global Equity Research Jim Covello in an interview published by the firm last month.  “But 18 months after the introduction of generative AI to the world, not one truly transformative — let alone cost-effective — application has been found.”

This skepticism runs counter to the prevailing opinions driving decision-making at the world’s largest investment and asset management firms, as well as the data center providers on the receiving end of the AI spending wave. Several industry leaders who spoke with Bisnow chafed at the suggestion of an AI bubble.

They say that unlike the dot-com boom or other speculative bubbles, AI is being driven by the world's largest, most stable companies, and there is little concern that underwhelming revenue or stock market volatility would meaningfully slow Big Tech’s AI spending or carry significant risks for data center providers and other firms down the AI infrastructure value chain.  

“I think using the term bubble here is a bit misplaced,” said Anthony Wanger, a digital infrastructure investor and adviser at KKR. “When you look at something like the dot-com boom, infrastructure providers like data centers were asked to take big risks and got left holding the bag. I just don’t see that here.”

Since ChatGPT first captured the public's imagination in late 2022, the scale and speed at which a handful of tech firms have directed resources toward expanding their computing power has no historical precedent. Asset manager Blackstone projects there will be close to $2T in global AI investment in the next five years, while Goldman Sachs anticipates $1T in U.S. AI capex over the same period.  

The bulk of this spending will continue to go toward building new data centers or leasing them from third-party providers, computing equipment like GPU processors that will be housed inside those data centers, and the electricity and energy infrastructure needed to operate them.

This capex explosion supercharged what was already a rapidly expanding data center sector coming off three years of record demand. While U.S. data center capacity totaled 17 gigawatts at the end of 2022, that figure is expected to reach 35 gigawatts by 2030, according to a January report from Newmark. Global data center inventory is expected to triple within six years, according to Synergy Research — and those data centers are larger than ever before. 

The scale of this build-out is on pace to be significantly larger in real dollar terms than other recent digital infrastructure cycles, experts say — dwarfing the investment required to expand fiber infrastructure during the dot-com boom or upgrade wireless networks to support 3G, 4G and then 5G as mobile data use exploded in the years following the introduction of the iPhone.  

“People used to marvel that AT&T and Verizon were in a capex war spending $17B to $20B per year on 5G,” Wanger said. “Everyone was blown away by that, but some companies now are running that per quarter.”

Cloud and social media giants footing the bill for this infrastructure boom believe the transformative potential of AI will make the massive cost worth it. And so far, economists and investors largely agree.

A Goldman Sachs analysis suggests AI will lead to a 9% increase in U.S. productivity and a 6.1% increase in GDP in the next ten years, while PWC expects AI to contribute $15T to global GDP by 2030.

Wall Street has also bought in. The stock market’s overall climb since the start of 2023 has been fueled in large part by investor enthusiasm over the major tech companies’ investments in AI.  Amazon, Microsoft, Google and Meta, along with AI chipmaker Nvidia, have driven the stock market to record heights, making up five of the “Magnificent Seven” firms that accounted for 24% of the S&P 500’s gains in 2023.

Major institutional investors and private equity are similarly bullish on AI’s prospects. Blackstone announced this month that its accelerating investments in AI were responsible for driving up its total capital deployment by 73% year over year — a significant chunk of it going toward AI-focused cloud firm CoreWeave and the expansion of its data center footprint through subsidiary QTS Realty Trust. 

But not everyone is on board.

Critics say the cost of building AI infrastructure is so astronomically high that any returns in the form of revenue growth will fall far short of the initial investment.  Investors' enthusiasm for Big Tech's infrastructure spending spree is a bad bet, they say — a hype-driven AI bubble based on unrealistic expectations that is likely to burst if AI doesn’t produce significantly high returns soon.

As Goldman Sachs' Covello asks: Companies are spending $1T building data centers and other infrastructure for AI, but what $1T problem is AI solving?

MIT economist Daron Acemoglu, a vocal AI skeptic, predicts that AI will contribute only a half-percent increase in productivity and a 1% increase in U.S. GDP within the next 10 years — far below Goldman Sachs’ projected increases of 9% and 6.1%, respectively.

Covello says bullish predictions on the future of AI assume that truly transformative use cases for the technology will emerge — “killer applications” like the e-commerce and search applications of the early 2000’s or the applications enabling cloud computing that emerged over the past decade, both of which enabled significant value across nearly every sector of the economy. 

Not only have these transformative use cases failed to come to fruition, but it’s also not even clear what these use cases might eventually be, Covello says. With no roadmap to profitability, he anticipates that the story around AI will begin to shift within two years as investors and the public at large realize they overestimated the technology’s capabilities and potential.  

“How long investors will remain satisfied with the mantra that ‘If you build it, they will come’ remains an open question,” Covello said.  “The more time that passes without significant AI applications, the more challenging the AI story will become. And my guess is that if important use cases don’t start to become more apparent in the next 12-18 months, investor enthusiasm may begin to fade.”

While many analysts and digital infrastructure leaders vigorously disagree with this appraisal of AI’s potential, even the most bullish on AI’s future acknowledge that Wall Street could get skittish if significant revenue growth and new large-scale use cases don’t emerge quickly. Some kind of AI-related downturn in the stock market may be inevitable, particularly with so much of the market’s value tied to just a handful of firms.

Wall Street’s reaction to first-quarter earnings reports in April showed that investors are looking for tech firms to tie their infrastructure spending to short-term revenue. Cloud providers Google and Microsoft went to great lengths to point to specific areas of revenue growth enabled by AI and saw their share prices climb as a result. Meta, on the other hand, saw its value dip after being unable to show that its AI capex was already bearing fruit. 

But despite this potential stock market volatility, leaders from across the digital infrastructure landscape who spoke with Bisnow universally pushed back on the idea that the zealous investment in AI infrastructure is a bubble akin to the dot-com boom of the late 90’s. When the dot-com bubble burst in 2000, it set off a wave of high-profile bankruptcies that created collateral damage across the economy. Experts, including those that are skeptical of AI, see little risk of that here. 

Whereas the companies fueling the dot-com bubble were startup companies with few real revenues, weak balance sheets and unsustainable business models, the AI investment wave is being fueled by the largest, best-capitalized credit grade companies in the world. Even if AI adoptions falters and their AI spending ultimately looks foolish, there is little risk to the overall stability of companies like Amazon or Microsoft. 

The strength of the companies driving AI spending also limits the blast radius of potential damage across the digital infrastructure space if anticipated AI revenues fail to materialize, experts say. Hyperscalers may cut back on their AI infrastructure capex, but they’re not going to default on the hundreds of megawatts of preleased data center capacity or other contractual obligations. There’s none of the exposure to risk down the infrastructure value chain that caused significant pain for the industry in 2000.

“When the dot-com bubble burst, in the data center world all these unsustainable business models were our tenants, and data center companies went bankrupt because our tenants went bankrupt,” said John Day, chief commercial officer at wholesale data center developer CleanArc. “Now you're talking about the best, most credit-worthy tenants in the world who are the biggest drivers of this AI spend. That’s one of the biggest differences between then and now.”

Should tech giants not see dramatic revenue growth and public markets sour on AI’s potential, hyperscalers are still unlikely to substantially slow their investment in data centers and other AI infrastructure, experts say.  

They may have no choice. 

Even if AI never creates a substantial change in revenue growth, companies like Amazon, Microsoft and Google likely need to continue deploying capex towards AI to hold on to the revenues they already have. AI computing is already changing the standards and expectations for everything from search engines to enterprise cloud services, and hyperscalers who don’t keep up will end up shedding customers to their competition.  

It’s a dynamic that Partners Group senior infrastructure investment leader Fentress Boyse compares to wireless providers’ 5G arms race, which saw hundreds of billions of dollars deployed in less than a decade. This massive investment in digital infrastructure did not lead to a steep change in revenue, but a wireless firm that chose not to upgrade its network would not be viable in today’s marketplace.

Big Tech may be in a similar position, with no choice but to double down on its AI bet.  

 “The opportunity cost and the potential consequence of anybody taking their foot off the pedal is just too great. If your competition puts the clutch in and you don’t, you lose,” said Ali Fenn, president of data center developer Lancium. “I share the suspicion that there will be some version of market corrections in public equities and the market will ask questions, but the underlying investment is still going to be there.”