Across the long arc of technological history, each transformative wave has carried within it both genuine promise and the seeds of speculative excess — and artificial intelligence is no exception to this ancient tension. Two seasoned analysts, Azeem Azhar and Paul Kedrosky, have entered into a consequential debate about whether the billions pouring into AI infrastructure represent the rational funding of a new economic era or the familiar architecture of a bubble nearing its limit. Their disagreement is not about whether AI matters, but about whether the market's faith in its near-term returns
Could the A.I. Economy Be a Bubble? Two Experts Debate
The technology could be real and important while the market is still irrationally exuberant.
Mark
So what's actually at stake in this debate? Is it just about whether stock prices go up or down?
Mimi
It's bigger than that. If AI investment is sound, we're watching the foundation of the next economic era get built. If it's a bubble, we're about to see a lot of capital destroyed and a lot of companies that bet wrong go under.
Luke
But how do we actually measure whether the spending is justified? What's the test?
Mimi
That's the hard part. Azhar looks at real applications—things that are already working. Kedrosky looks at the gap between what's being spent and what's actually generating revenue.
Mark
And that gap is big?
Mimi
Enormous. Companies are spending billions on infrastructure for capabilities that are still being figured out.
Luke
So Kedrosky's argument is basically: we've seen this movie before, and it doesn't always end well.
Mimi
Exactly. The pattern of a bubble is that everyone agrees the technology is important, but nobody agrees on the price.
Mark
What would prove one of them right?
Luke
Time. And earnings. If AI companies start generating the returns that justify their valuations, Azhar wins. If they don't, and prices collapse, Kedrosky was right.
Mimi
The tricky part is that AI might be genuinely transformative AND overpriced right now. Both things could be true.
Mark
So we could be in a bubble and still be building the future?
Mimi
Yes. The technology could be real and important while the market is still irrationally exuberant about it.
Der Puls
- Tens of billions are flowing annually into AI chips, data centers, and model development — and the question of whether that capital is wisely placed has never felt more urgent.
- Azhar argues the spending is foundational, not speculative, pointing to real productivity gains already reshaping drug discovery, manufacturing, and automation across industries.
- Kedrosky sees the hallmarks of past crashes: fear of missing out, valuations built on distant projections, and a market that has sprinted ahead of the evidence.
- The stakes are concrete — investors, companies, and policymakers face decisions now that will look either prescient or catastrophic depending on which analyst is right.
- Neither expert claims certainty, leaving the market itself as the only arbiter — through either sustained growth that vindicates the optimists or a correction that rewards the cautious.
Across the long arc of technological history, each transformative wave has carried within it both genuine promise and the seeds of speculative excess — and artificial intelligence is no exception to this ancient tension. Two seasoned analysts, Azeem Azhar and Paul Kedrosky, have entered into a consequential debate about whether the billions pouring into AI infrastructure represent the rational funding of a new economic era or the familiar architecture of a bubble nearing its limit. Their disagreement is not about whether AI matters, but about whether the market's faith in its near-term returns has outrun the evidence. The answer, when it arrives, will have shaped the fortunes of investors, institutions, and entire industries.
A question has been quietly unsettling investors and technologists alike: is the artificial intelligence sector building the foundation of a new economy, or inflating a bubble destined to burst? Two prominent analysts recently sat down to argue both sides, bringing fundamentally different readings of the same data.
Azeem Azhar, who has spent his career studying how innovations reshape markets over time, sees the current wave of AI investment as proportional to the opportunity. In his view, AI is already solving real problems — in medicine, manufacturing, and automation — and the infrastructure spending required to support it is not speculative but foundational. The economy, he argues, is already beginning to depend on these tools.
Paul Kedrosky, a venture capitalist shaped by years of watching tech booms collapse, reads the same landscape with pattern-recognition skepticism. He does not dispute that AI has genuine value, but he sees the current valuations and capital rush as characteristic of bubbles past — markets projecting exponential returns far into the future on the basis of limited present evidence, driven more by fear of missing out than by sober analysis.
The practical weight of this disagreement is considerable. Investors who trust Azhar and stay in could capture a generational wealth creation — or, if Kedrosky is right, absorb a devastating correction. Companies deciding how heavily to bet on AI infrastructure face the same binary. Policymakers, too, must decide how to regulate a sector that could either anchor the next era of growth or require managing the fallout of its collapse.
Both men acknowledge uncertainty. The debate is ultimately about pace and valuation — whether current prices reflect a future that will arrive on schedule or one that may take far longer, or never fully come. The market will eventually render its verdict. Until then, the tension between these two perspectives offers the clearest map available for navigating one of the most consequential economic questions of the moment.
Two economists sat down recently to hash out a question that has been nagging at investors, technologists, and anyone watching the artificial intelligence sector pour billions into infrastructure and research: Is this a genuine economic transformation, or are we watching a bubble inflate before our eyes?
Azeem Azhar and Paul Kedrosky represent different poles of this argument. Azhar has built his career studying technological adoption and the way innovations reshape markets over decades. Kedrosky, a venture capitalist and financial analyst, has spent years watching tech booms and busts, learning to spot the patterns that precede a crash. When they met to discuss the current state of AI investment, they brought fundamentally different readings of the same data.
The core tension is straightforward but consequential. The AI sector has attracted enormous capital—companies are spending tens of billions annually on chips, data centers, and model development. Valuations for AI-focused firms have soared. The question is whether this spending reflects genuine productivity gains and real demand for AI capabilities, or whether it represents the kind of speculative frenzy that has preceded previous tech collapses. Are we funding the future, or funding hope?
Azhar's position rests on the observation that artificial intelligence is beginning to solve real problems in ways that older technologies could not. He points to the tangible applications emerging across industries—from drug discovery to manufacturing optimization to customer service automation. In his view, the current investment levels, while substantial, are proportional to the scale of the opportunity. When a technology can genuinely reshape how work gets done across entire sectors, the infrastructure spending required to support it is not frivolous; it is foundational. The companies building these systems are not speculating on a distant future; they are building the backbone of an economy that is already beginning to depend on these tools.
Kedrosky approaches the same landscape with skepticism born from pattern recognition. He notes that the current spending on AI infrastructure and the valuations attached to AI companies have characteristics that resemble previous bubbles. There is a rush to deploy capital, a sense that missing out on the opportunity is worse than the risk of overpaying, and a tendency to project exponential returns far into the future based on limited current evidence. He is not arguing that AI has no value or that the technology will not matter. Rather, he is suggesting that the market has gotten ahead of itself—that current prices assume a level of adoption and profitability that may take far longer to materialize than investors expect, or may never arrive at all.
The practical stakes of this disagreement are substantial. If Azhar is right, investors who sit out the current wave of AI spending will miss a genuine transformation and the wealth creation that accompanies it. If Kedrosky is right, those same investors will have dodged a significant correction that could wipe out billions in value. For companies deciding whether to bet heavily on AI infrastructure, the difference between these two scenarios determines whether they are making a strategic investment or a costly mistake. For policymakers, it shapes how they should think about regulating the sector and managing the economic disruption that might follow if the bubble does burst.
Neither expert claims certainty. Both acknowledge that artificial intelligence is a genuinely powerful technology with real applications. The disagreement is about pace, valuation, and the relationship between current spending and future returns. It is a debate that will likely continue until the market itself renders a verdict—either through sustained growth that validates current prices, or through a correction that proves the skeptics right. Until then, the two perspectives offer a useful frame for thinking through the risks and opportunities in one of the economy's most consequential sectors.
Bemerkenswerte Zitate
Azhar argues that current investment levels are proportional to the scale of the opportunity, as AI is beginning to solve real problems across industries.— Azeem Azhar
Kedrosky suggests the market has gotten ahead of itself, with current prices assuming adoption and profitability levels that may take far longer to materialize.— Paul Kedrosky
Möchten Sie die ganze Geschichte?
Das Original lesen bei The New York Times ↗
Economy & Finance