AI Bubbles Can Spur Progress Despite Investor Losses, Says Economist

The Paradox of Speculative Investment in AI

The current fervor surrounding Artificial Intelligence, often characterized by soaring valuations and intense investor interest, might appear unsustainable to many. However, Thorsten Hens, a professor of financial economics at the University of Zurich, posits a compelling argument: speculative bubbles, while painful for investors who misjudge the market, can paradoxically be a powerful engine for technological advancement. This perspective challenges the conventional view that bubbles are purely destructive, suggesting instead that they can accelerate discovery and innovation in ways that planned, rational investment might not.

Hens's thesis centers on the nature of decentralized investment. When a field captures broad speculative interest, it encourages a multitude of different approaches and ideas to be pursued simultaneously. This decentralized experimentation, he argues, is crucial for uncovering genuinely novel breakthroughs. The process is inherently messy; it involves exaggeration, numerous failures, and ultimately, the bursting of the bubble as unrealistic expectations meet reality. Yet, it is precisely this chaotic, widespread exploration that can yield discoveries that are difficult to predict or plan for through traditional venture capital models.

To illustrate this point, Hens employs a vivid analogy: building a rocket and intentionally blowing it up. While seemingly wasteful, this act of doing and experiencing the consequence can provide a shortcut to learning that mere theoretical consideration cannot achieve. The empirical data gained from such an experiment, even a failed one, contributes to a broader understanding of rocketry. Applied to AI, this suggests that the current wave of investment, even if it leads to a bubble, is seeding a vast array of experiments. Many of these will fail, and many investors will lose money. But the knowledge gained, the tools developed, and the research produced will not disappear.

This perspective raises a critical question: what happens to the AI ecosystem if and when current expectations collapse? If the speculative fervor subsides and investors face losses, the enduring legacy might not be the financial returns for early backers, but rather the tangible progress in AI capabilities. This could manifest as more efficient computational methods, novel algorithms, or even widely adopted AI tools that become integral to various industries. The crucial factor is that the learning and the technological artifacts created during the speculative phase persist, forming the foundation for future, more sustainable development.

The Role of Exaggeration in Driving Progress

Hens contends that exaggeration is not merely a byproduct of a bubble but an active contributor to its progress. Overly optimistic projections attract capital and talent that might otherwise be hesitant to engage with nascent or unproven technologies. This influx of resources allows for a broader and deeper exploration of the technology's potential. Think of it less like a carefully curated garden and more like a wildfire that, while destructive in parts, clears the way for new growth and reveals unexpected geological formations.

The AI field is currently experiencing precisely this kind of exaggerated attention. Startups are being founded and funded at an unprecedented rate, often with ambitious, future-oriented visions. While many of these ventures will undoubtedly fail to meet their lofty goals or prove commercially viable, the collective effort fuels rapid iteration and discovery. Researchers are pushing the boundaries of model architectures, training techniques, and application domains. Engineers are building and refining tools, platforms, and infrastructure to support AI development and deployment.

This widespread activity leads to a diversification of approaches. Some companies might focus on fundamental research, others on specific industry applications, and still others on the underlying infrastructure. This parallel exploration increases the probability that at least some of these diverse efforts will yield significant, transferable innovations. Even the companies that fail contribute valuable insights through their research papers, open-source contributions, or the lessons learned from their market experiments. These learnings become part of the collective knowledge base, benefiting future endeavors.

What Survives the AI Bubble?

The inevitable consequence of a speculative bubble is its eventual deflation. When the hype outpaces tangible returns and widespread adoption, the market corrects. For investors, this often means significant losses. However, Hens's argument suggests that this correction does not erase the progress made. The key distinction lies between the speculative value of a company (its stock price and market capitalization) and the intrinsic value of the technology it develops or advances.

When the bubble bursts, companies that were primarily built on hype or unsustainable business models will likely collapse. Their assets, including intellectual property, talent, and developed technologies, may be acquired by more stable entities, repurposed, or simply become part of the public domain through open-source releases or academic publications. The critical takeaway is that the useful tools, more efficient processes, and foundational research generated during the speculative phase are likely to endure. These are the durable outcomes of the investment, regardless of the financial fate of the initial investors.

Consider the dot-com bubble of the late 1990s. While many internet companies failed spectacularly, the period also saw the foundational development of critical internet infrastructure, e-commerce platforms, and search technologies that underpin today's digital economy. Similarly, the AI bubble, if it is indeed a bubble, will likely leave behind a more capable AI landscape, with more sophisticated models, better deployment tools, and a deeper understanding of AI's possibilities and limitations. The challenge for the industry and for researchers is to identify and build upon these enduring advancements, transitioning from speculative exuberance to sustainable innovation.

The question for the future is not whether a bubble will burst, but what specific AI technologies and methodologies will emerge as the most resilient and valuable from this period of intense, speculative investment. The process of identifying these survivors and integrating them into practical, scalable applications will define the next era of AI development.