This article offers a deep dive into the nature of the current AI investment frenzy by comparing it with past speculative bubbles. The crucial point is that, unlike previous bubbles characterized by denial ("This time it's different"), the AI boom is unique because investors are largely aware of and even embrace the notion of a bubble. This self-awareness is a critical starting point for investors and founders alike to understand the current landscape and formulate strategies. There's a strong debate between optimists who see it as a "productive infrastructure bubble" that will lay the foundation for future giants like Amazon and Google, and pessimists who warn it might be a "short-term financial bubble" leading to capital destruction due to the rapid obsolescence of core AI hardware.

The background for this discussion includes theories from scholars like Carlota Perez and Tyler Cowen, who argue that bubbles, though wasteful in the short term, can drive excessive investment during the early stages of technological revolutions. This overbuilt infrastructure eventually becomes the backbone of entire industries, citing examples like the British railway boom of the 1840s or the fiber-optic mania of the 1990s. However, the article challenges this analogy by highlighting the two-year obsolescence cycle of AI chips. This fundamental difference from long-lasting infrastructure investments leads to the criticism that AI infrastructure might be more akin to "single-use scaffolding" rather than enduring underground fiber, raising fundamental questions about its long-term value.

This debate has significant implications for the industry and startups. If the AI bubble proves to be a "productive infrastructure bubble," current massive investments could indeed be the bedrock for long-term technological innovation. Conversely, if it's a "financial bubble," it could lead to inefficient capital allocation, the demise of many startups, and significant losses for investors. AI-related startups, in particular, need to demonstrate clear visions and strategies for creating sustainable value, beyond merely accessing the latest chips or massive computing power. Warning signs like "circular investment structures" and "off-balance-sheet financing" also caution about the transparency and health of capital markets.

For Korean startups, the implications are clear. Firstly, AI core technology startups must focus on software optimization, efficient model architectures, or specialized value creation in specific domains, considering the rapid obsolescence of hardware. Business models solely reliant on purchasing high-performance chips are risky. Secondly, startups developing AI applications should prioritize solving real user problems and establish clear business models to prove "true value," rather than being swayed by excessive "bubble" expectations. Thirdly, when raising investment, they should deeply consider the long-term sustainability and recoverability of investments, avoiding blind optimism that "this time it's different." Ultimately, the key is to strike a balance between the innovativeness of AI technology itself and its practical utility in the market.