AI Bubble Myth or Reality?

In recent months, two dominant narratives have emerged that capture prevailing public sentiment toward AI. Each narrative draws on selective metrics and market signals to substantiate its claims; however, both are fundamentally misguided and lack the strategic foresight needed to guide effective public policies that safeguards the future of work in the digital era.

On one hand, doomsayers cite headlines on rising unemployment – particularly among new grads – and layoffs in the tech industry as evidence that we are at the brink of massive job displacement as companies deploy language learning models and automation to replace humans at new unprecedented scale. One headline fueling this belief, is Amazon announcement in late October 2025, that it will reduce its corporate workforce by roughly 14,000 people, or 4%, with more cuts expected next year. It explicitly linked part of the reduction to generative AI. The announcement came 2 days ahead of its third quarter financial results and just days following a major Amazon Web Service (AWS) outage. The timing of the events is telling and challenges the notion that companies such as Amazon are likely realizing significant gains using AI to support substantial cuts in their workforce – at least not yet.

Some argue that geopolitical tensions and economic uncertainty- brought forth by the tariff wars and inflation- are driving the layoffs and rising unemployment. Ai is the convenient scapegoat. They point to similarly workforce reductions announcements by major retailers to substantiate their claim. This is partly true, but incomplete explanation for what is driving the layoffs- at least for major players such as Amazon.  

In its third quarter financial results, Amazon reported revenues for its cloud computing business increased 20% from a year earlier to $33 billion. Yet this growth is overshadowed its competitors Microsoft Azure and Google Cloud, which reported revenue growth of 40% and 34 %, respectively during the quarter.  Although AWS is the dominant market player in cloud infrastructure with 32% of the market share, Microsoft Azure is second at 23%, followed by Google’s cloud unit at 10% —and both competitors are rapidly gaining ground.

Much of their momentum stems from aggressive investments in AI. In October Google announced a cloud partnership with Anthropic to run its Claude language model across up to 1 million of Google’s custom-designed Tensor Processing Units (TPUs). Microsoft, meanwhile, opened  a new datacenter in Atlanta, the second in Microsoft’s Fairwater AI superfactory.  To counter concerns that it was falling behind, Amazon unveiled an $11 billion AI data center, Project Rainier designed to host Anthropic’s Claude chatbot.

The capital expenditures for these investments are substantial. Amazon raised its projected capital expenditures from $118 billion to $125 billion for 2025 and expects further increases in 2026. The increase capital expenditure on AI is perhaps the most compelling argument driving the recent layoffs, as Amazon seeks to offset the cost by reducing its workforce. What remains clear is that there is little evidence to support the claim that companies are cutting workers because AI is already delivering massive productivity gains. Market forces are at play here and companies, such as Amazon, Microsoft, Google, and a slew of other companies such as Meta, Oracle are making major investments in AI infrastructure is a bet in its future potential.

These investments have fueled another prevailing public sentiment—particularly among AI skeptics—who fear that such capital inflows may be inflating a market bubble with the potential to spill over into the broader economy. Economic theory teaches us that bubbles form when the market price of an asset exceeds its intrinsic value—in other words, when the fundamentals of the asset do not align with the value assigned to it by investors. Nobel Prize–winning economist Robert Shiller, famously described such moments as  “irrational exuberance”: investors engage in speculative buying by ignoring the income (or lack thereof) of the asset because they believe prices will continue to rise, increase their borrowing or leverage to fuel additional purchases (thereby amplifying their risk), and when returns fail to materialize, panic sets in—leaving investors and the broader economy with substantial losses.

Skeptics point to recent reporting, including a Financial Times headlines, revealing that 10 leading AI companies—including OpenAI and Anthropic—have seen their valuations rise by nearly $1 trillion in the past 12 months, despite lacking demonstrated commercial viability. U.S. venture capitalists are on track to invest nearly $300 billion this year alone in companies that have yet to show any economic returns. There is also strong evidence to suggest that companies are making these capital expenditures not from operating profits but from debt. For example, a consortium of 20 banks financed $18 billion to Oracle to support the construction of a data center in New Mexico. While there is credible precedent for an AI market bubble potentially forming, it would be a mistake to conflate the risk of such a bubble bursting with a diminishing ability—or lack of staying power—of AI to transform the global economy.

Two historical comparisons illustrate the nuance. Let’s take the two common cited bubbles- the internet bubble of the 1990s and the subprime mortgage bubble of 2008.  In the most recent example, financial engineering - of high-risk (sub-prime) mortgages – led to the collapse of the housing market in 2008. Wall Street generated a variety of complex financial products based on pools of risky loans. Banks and investors, linked through opaque and highly leveraged instruments, financed these loans. The opaque nature of the business masked the systemic risk these products posed to the broader economy.  As housing prices peaked, mortgage refinancing and selling homes became less viable leading to defaults and substantial losses for lenders and investors. The ensuing decline in demand, lead to a market correction, with home prices falling so much that it became hard for troubled borrowers to sell their homes to fully pay off their mortgages.

The magnitude of the crisis was rooted in the fact that these assets were deeply embedded in the global financial system. Unlike the underlying asset (mortgages) at the root cause of 2008 financial crisis, AI is not a foundational asset propping up the world’s financial architecture.  That said, recent headlines of banks backing financing of AI infrastructure is a worrying trend, and warrants careful monitoring. What is required is transparency to fully gauge the true scale of investment and interdependencies to get a clear picture of the potential for losses, and in turn risk to the broader economy.

Investments in AI today more closely mirrors that of the 1990s internet stock bubbles. Then, the promise of the internet sparked massive investment as businesses and investors raced to determine how the “World Wide Web” could generate value. Capital poured in, telecommunications firms laid more than 80 million miles of fiber-optic cable in the United States alone, and valuations soared. When economic returns lagged behind expectations, the bubble burst, producing winners and losers but only modestly affecting the broader financial system: U.S. GDP contracted 1.3% in 2001, compared with a cumulative 5.1 percent during the 2008 financial crisis, and required mass bailouts and stimulus from the federal reserve.  

The internet however did eventually transform the global economy.  It became core infrastructure supporting commerce, finance, and communication worldwide. AI is poised to follow a similar trajectory—but likely at a much faster rate. As of September 2025, OpenAI boasts of 700 million weekly active users of ChatGPT, only three years since it was launched. Rapid adoption, combined with exponential advances in model capabilities, positions AI to reshape global economies more swiftly and more profoundly than previous technological waves.

Viewed in this context, both dominant public narratives—fears of immediate, massive job displacement and claims of an imminent AI market collapse—miss the larger point. Each overstates one dimension of the present while understating the significance of the long-term transformation already underway. The central issue is one of timescale.

As AI continues to advance at remarkable speed, public policy must evolve with equal thoughtfulness and urgency. Policymakers should focus on shaping the technology’s development, protecting the workforce, and reinforcing the economic foundations that support stable, resilient societies. Above all, safeguarding people—not just markets—must remain at the center of the AI transition.