Economic headlines and the dominant Silicon Valley discourse are gripped by a pervasive fear: "The AI bubble is bursting." Temporary sell-offs in tech giants’ shares, disappointing reports on unrealized dream revenues, and alarming data—such as MIT’s finding that 95% of corporate AI projects have yet to show tangible returns—have converged to create a climate of pervasive pessimism. We appear to face a stark dichotomy: if AI succeeds, a wave of mass unemployment awaits; if it fails, a devastating financial crisis looms.

Yet is the real question "Is AI a bubble?" Or should we ask: "When this bubble bursts, what civilization will stand behind the dust?"

 

Anatomy of an Illusion: The Gap Between Intrinsic Value and Market Hype

A bubble forms when collective expectations drift miles away from an asset’s intrinsic worth. Five years ago, Jack Dorsey’s first tweet sold as an NFT for $2.9 million, hailed as the "Mona Lisa of the digital age", an asset whose market value today hovers near zero. In the late 1990s, merely affixing ".com" to a company name lifted its valuation by more than 70% within ten days on average. Today, that identical dynamic plays out with the acronym "AI."

Today’s figures are staggering:

Meanwhile, circular financial arrangements. often called round-tripping, have taken root. Nvidia invests billions into AI startups; those startups contract with cloud hyperscalers; and hyperscalers channel those funds back to Nvidia for GPUs. As the IMF has documented, these circular flows have mounted steadily. The loop holds only so long as real outside demand flows in; the moment demand wavers, the dominoes fall.

Revisiting Two Collapses: Telecom in 2000 vs. Real Estate in 2008

To comprehend this moment, one must distinguish between two fundamentally different types of bubbles.

Lesson from the Dot-Com Era: The Potential of a New Technology

Many remember the dot-com crash merely as Pets.com going under, but the structural core lay underground within telecommunications conduits: the telecom bubble. Firms like Lucent extended massive vendor financing to undercapitalized startups so they could purchase Lucent gear. Simultaneously, telecom executives relied on flawed metrics, such as "Internet traffic doubles every 100 days"—to justify astronomical outlays on global fiber-optic cables.

Two bottlenecks doomed the run: first, growth rates were wildly overstated; second, transoceanic and intercontinental backbones were laid, but bringing fiber to homes, the "last-mile" dilemma, proved slow and prohibitively expensive.

Consequently, the Nasdaq collapsed in March 2000, shedding 80% of its valuation. By 2001, roughly 95% of installed global fiber lay dormant or "dark." Yet around 2005, companies like Google acquired this abandoned infrastructure at pennies on the dollar. Without that manic bubble, cheap high-speed fiber would never have existed, and platforms like YouTube and Netflix could not have emerged.

The 2008 Housing Crisis: Speculation on Traditional Assets

By contrast, the 2008 crash was built on the assumption that real estate values could only appreciate. Banks extended unbacked subprime mortgages without underwriting discipline. When the bubble burst, millions lost homes and livelihoods, plunging the global economy into recession, leaving behind no breakthrough technology or productive infrastructure for posterity.

Which Bubble Does AI Resemble?

Tech giants are betting on Artificial General Intelligence (AGI), not an inert physical asset. AI already confronts its own "last-mile" bottleneck: training foundational large language models is one thing; deeply embedding their capabilities into the everyday workflows of businesses and economies is far more complex. If this bubble bursts, what legacy remains?

Chips will not be the primary heirloom. Unlike fiber optics that endure for decades, GPUs depreciate and become obsolete within three to five years. Open-source models and institutional expertise will endure and grow cheaper. The genuine legacy lies in power and energy infrastructure. The most enduring byproduct of this frenzy is the foundation that powers it: electricity generation. Data centers could consume up to 12% of total US electricity demand by 2028. This voracious appetite has prompted tech leaders to invest in nuclear fusion (such as Sam Altman’s backing of Helion), acquire dedicated power plants, and order advanced gas turbines.

If the dot-com bubble left behind high-speed fiber optics, the AI bubble will leave robust grids, expanded generation, and abundant power. Had capital allocators moved strictly by incremental logic, hundreds of billions would never have flowed into energy innovation and grid modernization.

 

Policy Missteps and the Shadow of Crisis

The true peril lies not in price corrections or bursting bubbles, but in governments' subsequent policy interventions. Following the 2000 crash, the Federal Reserve slashed interest rates from 6.5% to 1%, unleashing cheap liquidity that flowed directly into residential real estate and laid the groundwork for 2008.

When public institutions step in to bail out private risk-takers, the cost of miscalculation shifts onto society at large. Emerging calls for state guarantees on heavy AI infrastructure debt signal that this temptation has returned.

No economist can pinpoint when a bubble will burst. Alan Greenspan warned of "irrational exuberance" in 1996, yet markets climbed for nearly four more years.

The strategic questions of this era are not "When will it burst?", but rather:

  1. What is the true nature of this collective frenzy?
  2. Who will absorb the losses?
  3. What infrastructure will remain among the ashes to propel the next civilizational leap?

History occasionally requires a measure of mania to take immense civilizational strides. Bubbles will inevitably collapse, but the improbable infrastructure forged at their height will power the world of tomorrow.