KB Securities’ Lee Says Rising Rates Are the Key Signal for an AI Bubble Burst

A stock-market bubble driven by optimism over artificial intelligence is not yet near collapse, KB Securities said, because interest rates remain below levels that would meaningfully restrain capital spending by hyperscalers, or operators of massive data centers.

The clearest sign that the bubble is approaching its end would be an “irreversible rise in interest rates” that begins to choke off hyperscalers’ capital expenditure, according to the firm.

Lee Eun-taek, a director at KB Securities, made the remarks at a briefing at the Korea Exchange on Aug. 18. “The parties that will ultimately stop AI capital spending are the capital suppliers,” he said, referring to creditors that lend money to finance hyperscalers’ investment in AI infrastructure.

Big Tech companies are highly unlikely to stop investing on their own, Lee said, because they have succeeded before by continuing to invest despite losses and eventually dominating their markets. Investment will likely stop only when capital providers such as banks, pension funds and venture capital firms — whose top priorities are recovering principal and collecting interest — detect risk and tighten funding.

If creditors start to sense danger through worsening free cash flow or rising credit-default swap premiums, the supply of capital could dry up and bring the bubble phase to an end.

Lee cited one common factor behind the three biggest bubble collapses of the past 130 years: the Great Depression of the 1930s, stagflation in the 1970s and the dot-com bust in the 2000s. In each case, the shared trigger was a sustained rise in interest rates. By contrast, most short-term spikes in rates merely created buying opportunities. For a bubble to actually burst, he said, two conditions must be met.

“First, it has to be an irreversible situation — a ‘No Way Back’ moment — where policymakers want to cut rates but cannot because inflation has surged,” Lee said. “Second, rates must break to new highs at levels not seen in 10 to 20 years.”

More specifically, he said the tipping point for a broad shift into safe assets — and a possible bubble collapse — would come if the yield on the US 10-year Treasury rose decisively above the 5.0% to 5.5% range. That would mark a breakout above peaks seen in 2002 and 2007.

Lee described the overnight rise in the 30-year US Treasury yield above 5.3%, surpassing 2007 levels, as a development worth watching. Still, he said it mattered less than moves in the benchmark 10-year yield.

He also said it was too early to worry about the recent market debate over slowing AI demand and profitability. Concerns over token costs helped trigger a stock-price correction, but those worries are easing.

Lee said the correction from late June through July was driven by concern that demand for frontier AI models from OpenAI and Anthropic could weaken because of token-cost issues. Chinese AI models also drew attention as a possible alternative for users seeking lower-cost options.

Since then, operators of US frontier AI models have moved to lower costs. Companies using those models have also begun optimizing spending by cutting usage of top-tier models and shifting simpler tasks to cheaper alternatives.

“Until this spring, the dominant approach was ‘token maxing,’ which meant concentrating resources on high-performance models,” Lee said. “More recently, the paradigm has shifted to ‘token optimization,’ mixing high-performance frontier models with cheaper open-source models to control costs. Short-term concerns tied to intensifying price competition — including the rise of Chinese AI models and OpenAI’s price cuts — have already been largely priced into stocks.”

Han Kyung-woo, Hankyung.com reporter case@hankyung.com