AI 버블론, 반도체 고점론은 사실일까

Recent claims have surfaced about the bond market of hyper-scale companies, major companies operating large data centers. Leading this discussion is Don Johnson, president and CEO of GLJ Research. With over 15 years of experience, Johnson primarily analyzes industries such as steel, mining, electric vehicles, and solar. He is often compared to famous investor Michael Burry, known for his bearish views, especially on Tesla.

Johnson frequently issues bearish investment opinions, with over 63% of his recommendations being sell-side. This is likely causing market turmoil. The reason for this turmoil is Johnson's concern over the high costs of investing in artificial intelligence (AI). This concern has led to a significant increase in credit default swap (CDS) spreads, which are essentially insurance policies against default.

In the United States, the CDS basket of the top five hyper-scale companies (Amazon, Meta, Microsoft, Google, and Oracle) saw its yield spread rise from 115bp to 162bp, a 50% increase, over the past few months. This spread represents the likelihood of a company defaulting over the next five years. If even one of these companies defaults, it could have a significant impact on the global market.

While the reasons for the CDS spread increase are clear, the lack of transparency in the repayable investment makes this situation noisy. This uncertainty is reflected in the CDS spreads. However, it's important to note that the CDS spread is not a perfect indicator of credit risk. It's more sensitive to liquidity and volatility than other measures, potentially overstating risk.

Therefore, it's beneficial to consider other indicators to get a more complete picture of the credit situation. The issue here is that these investments are highly uncertain, with no clear timeline for when and how much the companies will recoup their investments. This uncertainty is causing investor anxiety, which is reflected in the rising CDS spreads.

Furthermore, these hyper-scale companies are the largest in the United States. A credit crisis in these companies could have a far-reaching impact on the global economy, like a megaton nuclear bomb. While concerns about increasing debt issuance are valid, the CDS spreads are not an ideal measure of overall credit conditions. They are less stable, more sensitive, and tend to overstate risk.

In the case of Oracle, the CDS spread has skyrocketed, but this is mainly due to a high debt-to-equity ratio and a relatively low credit rating (BBB-). In fact, Oracle's CDS spread has increased from below 50bp in mid-2025 to 200bp today. This trend is mirrored in the credit spreads, with Oracle's 5-year CDS spread of 6.23% significantly higher than the 5-year US Treasury yield by about 180bp.

The high debt-to-equity ratio of Oracle (400%) is a major concern, but the credit rating of the other companies - Microsoft (S&P AAA, Moody's Aaa), Meta (36%), Google (18%), and Amazon (51%) - is relatively sound. The real risk lies with Oracle, whose credit risk is high. This is due to its massive debt, with a debt-to-equity ratio of 400%, putting it just above junk bond status.

The concerns from investors are justified, as Oracle's AI infrastructure investments are at risk. The fear is that these investments will not be repaid, and Oracle's credit rating may drop to junk status. However, it's crucial to maintain a critical perspective during turbulent times. Some of these theories may be purely speculative, often propagated by investment banks and hedge funds.

Many of these theories come from Wall Street, such as Morgan Stanley's warnings about semiconductor prices falling in 2021 and 2024. Jamie Covel of Goldman Sachs has also criticized the massive investment in AI infrastructure as "like building a nuclear power plant to turn on a toaster." They argue that the astronomical costs of building data centers and semiconductor infrastructure will never be recouped.

The reality is that many of these theories are noise, not long-term paradigm shifts. The key is to separate short-term noise from long-term structural trends. Most of the pessimism is likely just noise. While some of these concerns are valid, such as the high debt-to-equity ratio of Oracle, most of the pessimism is likely just noise, not a long-term structural shift.