The 30% plunge of South Korean data centre company Elice on its trading debut this week is more than a single disappointing IPO — it’s a signal that the market is beginning to question the valuations attached to the AI buildout. Elice raised $151 million and still saw its stock tumble, a rare event in a country where tech listings have historically been greeted with enthusiasm. That South Korea’s KOSPI index fell nearly 4% on the same day, while Brent crude surged past $111, suggests a broader recalibration: rising energy costs and geopolitical uncertainty are forcing investors to ask whether the promised returns from AI infrastructure justify the capital being poured in.
There is no question that AI requires enormous computing power, and that power requires data centres, chips, and electricity. But the assumption that demand will grow linearly — and that every data centre builder will capture that growth — is being stress-tested. Elice’s weak debut compounds jitters already visible in the US and Europe, where some AI-focused stocks have pulled back this quarter. The danger is not that AI is overhyped, but that the infrastructure buildout has been priced for perfect conditions: cheap energy, stable supply chains, and limitless demand. Rising oil prices, driven by Middle Eastern tensions, raise the cost of running those data centres. They also feed inflation, which could slow the broader economy and reduce corporate willingness to invest in unproven AI applications.
The lesson here is not to dismiss the AI transformation, but to recognise that sustainable progress requires grounding in real economics. Building more data centres without clear value creation or energy efficiency risks replicating the telecom bubble of the early 2000s, where fibre was laid far ahead of demand. The companies that will survive this phase are those that can demonstrate tangible returns — not just capacity. For developers and investors alike, the current market jitters are a reminder that the most durable technology bets are those that solve real problems for real users, rather than chasing the narrative alone.