Building AI’s Future, Or The Next Big Bubble?

As debates rage over whether the AI industry is in a sustainable boom or a speculative bubble, Anthropic placed a staggering bet on the former.

Earlier this month, the AI startup announced plans for a $50 billion U.S. infrastructure build-out, beginning with custom data centers in Texas and New York developed with its GPU cloud partner, Fluidstack.

The move positions the firm as a major AI player potentially securing domestic compute capacity at a time when policymakers are focused on U.S. technological sovereignty. The project is expected to create thousands of jobs with the first sites slated to go live in 2026.

In a statement to CNBC, CEO Dario Amodei said the investment was essential for building “more capable AI systems” that can drive scientific breakthroughs.

The bubble debate intensifies

The scale of financial commitments is historic. Oracle has already contracted $300 billion in cloud services to OpenAI, while Meta has pledged to spend $600 billion on infrastructure over the next three years — spending that underpins the belief that AI could unlock new, multi-trillion dollar markets.

However, this belief is also currently running ahead of tangible revenue. According to a Financial Times analysis, ten leading AI startups, including Anthropic, OpenAI, and xAI, have seen their valuations soar by almost $1 trillion in the past year, despite most being deeply unprofitable.

Furthermore, Oracle is facing significant investor skepticism and a major stock decline following its $300 billion deal with OpenAI stemming from the enormous debt the firm is taking on to fund the project, with cash flow expected to remain negative for years.

The sheer volume of capital has also led prominent figures to sound the alarm. Famed “Big Short” investor Michael Burry recently shut down his hedge fund, Scion Capital, stating that his valuation models were “not now, and [have] not been for some time, in sync with the markets.”

In a post on X, Burry alleged that major tech companies were artificially inflating their earnings by using overly long depreciation schedules for their AI equipment, which has a short two to three-year life cycle. He estimates this will understate depreciation by $176 billion from 2026-2028, overstating earnings for companies like Oracle and Meta by more than 20%

Meanwhile, speaking to CNBC, former Biden CEA chairman Jared Bernstein called an AI bubble the “likely outcome,” pointing to “extreme valuations” and the vast gap between investment and “credible expectations” for future profits. He stated that OpenAI, despite securing over $1 trillion in infrastructure commitments, is projected to generate only a fraction of that in revenue.

“The problem with bubbles is it’s hard to know if it’s really what you think they are until it’s too late,” he said. “The share of the economy devoted to AI investment is nearly a third greater than the share of the economy devoted to internet-related investments back during the dot com bubble.”

In a post on Substack co-authored by Bernstein and Ryan Cummings, chief of staff for the Stanford Institute for Economic Policy Research, they outlined why they think the bubble could burst, how we got here, and what the risks could be.

“Given that the underlying economy and the things we care most about — jobs, wages, incomes, affordability — are already showing some fragilities, it’s important to carefully monitor this potential bubble,” they wrote. “We’ve got enough to worry about without adding a big, negative wealth effect to the mix.”

In August, Praetorian Capital Founder and CIO Harris Kupperman Harris Kupperman published a blog analysis with “back-of-the-envelope-math” that similarly shed light on the alarming scale.

“The first shocking revelation: the AI datacenters to be built in 2025 will suffer $40 billion of annual depreciation, while generating somewhere between $15 and $20 billion of revenue,” noted Kupperman. “The depreciation is literally twice what the revenue is.”

“By my math, you need $160 billion of revenue at that 25% gross margin, which gives you $40 billion of gross margin against $40 billion of depreciation. Now, remember, revenue today is running at $15 to $20 billion. You need revenue to grow roughly ten-fold.”

In fact, some believe that the broader market — from AI stocks and Bitcoin to gold and real estate — shows signs of a more extensive widespread bubble, and AI spending may be creating a precarious “everything boom” built on hope, rather than fundamentals.

The bull case

Not all experts see a bubble, though, with many industry leaders arguing that this spending is a necessary investment in a foundational technology.

BlackRock CEO Larry Fink told CNBC that far from being a bubble, the spending represents a crucial investment for U.S. competitiveness.

“Investing in AI does not just mean investing in GPUs and chips, it means investing in HVAC and IT, investing in power grids and power supplies,” Fink said, adding that while these massive stakes could lead to some failures, major hyperscalers like Meta, Microsoft and Alphabet are in a “really good position” to win.

Also, the AI boom is, in many ways, a data-center boom, points out The Atlantic staff writer Charlie Warzel. “They’re spending tons of money to build out these data centers, and there’s this notion that there’s never enough. We’re going to need to keep building data centers. We’re going to need to increase the amount of power. So, what you have is this really interesting infrastructure problem, on top of what we’re thinking of as a technological problem.”

The sheer physical limitations of construction, power grids, and chip supply may naturally moderate the pace of investment, possibly preventing a purely speculative frenzy.

At large, the verdict is still out.

“Of course there’s a bubble,” Hemant Taneja, CEO of venture firm General Catalyst, told the FT. “Bubbles are good. Bubbles align capital and talent in a new trend, and that creates some carnage, but it also creates enduring, new businesses that change the world.”

Anthropic’s bet is a $50 billion testament to that high-stakes belief, leaving the industry to wait and watch whether it is building the future, or merely inflating it.

Forbes