Peak-Semiconductor Fears vs. the Scale Theory: Rep. Lee Hae-min's Case for Korea's AI Megaproject Amid a Wobbling KOSPI

Amazon is reportedly preparing to issue at least $25 billion in dollar-denominated bonds, with most proceeds expected to fund AI infrastructure. Deutsche Bank data cited on the show shows hyperscaler capex increasingly outstripping operating cash flow, prompting market doubts about how long the spending can continue.

The hosts pushed back, arguing capex should be viewed as investment, not expense. Borrowing to keep investing when cash runs short is a rational choice to avoid falling behind in the AI infrastructure race — and the moment a company stops investing, everything spent before instantly becomes sunk cost, and it drops out of the race. They suggested the eventual winner may be decided between companies like Apple, which is holding back and watching, and companies like Amazon, willing to take on debt to keep expanding.

An internal U.S. Treasury report reportedly flagged AI bubble warning signs, noting that financial markets are now far more entangled with the real economy around AI than they were during the dot-com era. That, the hosts said, is precisely why Washington can't afford to let AI investment stall — a systemic risk that explains continued government-linked moves such as the OpenAI investment and the Intel stake.

Morgan Stanley said the narrow semiconductor-led growth phase is ending and market leadership is broadening, recommending investors trim semiconductor exposure and add hyperscaler exposure in the near term. Foreign selling of Samsung Electronics and SK hynix continued after the report.

China's DeepSeek has reportedly been developing its own inference AI chip for about a year to reduce reliance on Nvidia and Huawei chips. Other Chinese firms, including Zhipu AI, are also said to be pursuing custom chips, adding further downward pressure on the sector.

Separately, China's government is reportedly discussing restricting overseas access to domestic AI models with major tech firms such as Alibaba and Zhipu. This follows the US Commerce Department's export controls on Anthropic's Claude models, underscoring a broader trend toward bloc-based fragmentation in the AI industry.

Within the US, however, the prevailing view was that Samsung's decline reflected profit-taking rather than weak results. Analysts noted that, excluding one-off charges, Samsung's operating profit topped market expectations, while forecasts pointed to TSMC and ASML earnings having greater sway over chip stock direction going forward.