Alice wanted to get through the door into the garden. When she shrank, she could not reach the key on the table; when she grew, she could not fit through the door. In Lewis Carroll's "Alice's Adventures in Wonderland," each change in size brings a different obstacle. The market's view of Elice has shifted much the same way in the space of three days. Elice, an artificial intelligence infrastructure company, drew more than 10 trillion won ($7 billion) in subscription orders, then plunged 32% below its offering price on its first trading day. The next day it jumped by the daily limit, and on the third day it extended the rally. If the crash was a warning about an AI bubble, how should the surge be read? The company did not change in three days. Its valuation swung on growth potential and the burden of investment.
There is room to question the starting price. In setting its offering price, Elice used as benchmarks the valuations of eight overseas cloud, data center and power equipment companies with different business structures and scales. Even so, a company that posted 10.3 billion won in operating profit in the first half cannot be dismissed as selling nothing but dreams. Nor can the AI bubble debate be settled by asking whether there is substance behind it. When an AI company borrows to buy chips, that becomes revenue for the chipmaker. The boom on the selling side and the strain on the buying side are two sides of the same money. That is why debt risk at CoreWeave, the U.S. AI infrastructure company, rattled semiconductor stocks as well. A market that cheered news of brisk chip sales has begun looking into the books of the companies doing the buying.
Data centers and chips are weapons in the competition, but they are also assets whose costs must be recovered. When companies invest to avoid falling behind, the combined spending can turn into overcapacity. Even if AI usage grows, competition that drives down service prices will leave earnings short of expectations. And the newer the technology that arrives, the faster equipment bought earlier may become obsolete. The pace of innovation does not guarantee a return on investment. It is premature to judge the future of AI by Elice's plunge, or to conclude that its surge has erased concerns about profitability. The growth potential and the investment burden both remain. What is strange may be less Elice's share price than a market that rushes toward expectations and then flees from the costs.