U.S. semiconductor prices fall for 6th day... Interest rate shock spreads anxiety over Coreweave's AI investment

Interest rate burden due to 3.7% rise in US CPI... Semiconductor index plummets around 4%, Coreweave bond yields around 15%, Gartner warns of $3.5 trillion in asset damage risk

U.S. semiconductor stocks, which led the artificial intelligence (AI) investment craze, fell for six consecutive trading days. Concerns about interest rate hikes have grown due to higher-than-expected inflation rates, and anxiety over the financing of AI infrastructure companies has spread. In particular, as the corporate bond yield of AI cloud company Coreweave soared to the 15% range, the sustainability of AI investment methods that have relied on massive borrowings was put on the chopping block.

On the 14th (local time) in the New York Stock Exchange, the Philadelphia Semiconductor Index fell about 4% from the previous trading day. It has been falling for 6 consecutive trading days, the longest downward trend since May of last year. The Nasdaq index, centered on technology stocks, also fell 1.88%.

The sell-off spread beyond semiconductor companies to AI data centers and server-related companies. Coreweave's stock price plummeted by about 15%, and Nevius and Oracle also fell by about 10% each. AI server manufacturer Dell Technologies also recorded a double-digit decline.

Reuters calculated that the market capitalization of major AI-related companies decreased by about $80 billion per day. Market interest is also shifting from expanding AI investment to whether the enormous investment costs can be recovered.

#### Concerns about interest rates reignite due to inflation at 3.7%... Double burden on AI companies

The September Consumer Price Index (CPI) announced by the U.S. Department of Labor rose 3.7% compared to the same month last year. It exceeded market expectations of 3.6% and was higher than the August increase of 3.4%.

As inflationary pressures appeared stronger than expected, the possibility of additional interest rate hikes by the U.S. Federal Reserve (Fed) was highlighted. This is a direct burden for AI companies that raise external funds for large-scale facility investment.

To build an AI data center, it is necessary to secure a large number of Nvidia's high-performance graphics processing units (GPUs) and be equipped with power, cooling, and network facilities. While the initial investment is enormous, profits are recovered through long-term contracts.

As interest rates rise, the costs of new investments and refinancing existing debt also increase. Even if demand for AI computing increases, financial costs may put pressure on profitability.

This sharp drop also reflected caution about the stock prices of companies that have soared amid the AI ​​investment craze. The fact that Barclays lowered its investment opinion on enterprise hardware companies such as Dell also increased the selling trend.

#### Coreweave bond interest rate around 15%... Gartner: $3.5 trillion in AI investment at risk of damage

Rather than the plunge in Coreweave's stock price, what clearly revealed the market's concerns was the decline in corporate bond prices.

According to Market Watch, an American financial media outlet, Coreweave's $2.75 billion worth of corporate bonds with a coupon rate of 9.75% issued in April were traded at about 81 cents per dollar of face value. The secondary market rate of return exceeded 15%.

The price of corporate bonds with a coupon rate of 9.625% due in 2032 also fell from $85 in early October to $79.50 on the 14th. The rate of return increased from 13.42% to 15.06% over the same period.

If a decline in stock prices shows a change in investment sentiment, a decline in corporate bond prices reveals concerns about financing. Although the interest rate on existing loans does not immediately increase, higher interest rates may be requested during new financing or refinancing.

CoreWeave secures a large number of NVIDIA GPUs and provides AI computing services to corporate customers. Since it is a business method that first builds a data center and recovers the investment through a long-term contract, the initial funding amount is large.

According to the London Stock Exchange Group (LSEG), Coreweave's bond issuance amount is $22 billion and loan financing amount is $26 billion. This is the amount of funds raised and is distinguished from the current remaining debt balance.

Coreweave refuted the market's concerns. In an internal memo sent to employees, CEO Mike Intrater stated that the market may misunderstand the company's business situation or react irrationally.

As of the end of the second quarter, the company's order backlog was $104.2 billion. In the third quarter, we signed new customer contracts worth more than $25 billion. CEO Intrater emphasized that the contract size secured is sufficient to support debt obligations.

However, the fact that there is a large order backlog and securing the cash needed immediately are two different things. This is because there is a difference between the time when sales generated from long-term contracts are actually received in cash and the funds that need to be invested in data center expansion.

There were also warnings about the profitability of overall AI infrastructure investment. In a report released on the 14th, global market research firm Gartner predicted that AI companies will invest approximately $7 trillion in infrastructure construction from 2026 to 2030. Cumulative sales of $8.5 trillion are needed to recover investment and financial costs and achieve a 12% return on invested capital, but sales attributable to AI data centers over the same period are expected to not exceed $3.6 trillion.

Gartner warned that if the current investment pace continues, more than $3.5 trillion worth of AI infrastructure investments could be at risk of damage by the end of 2030. It is predicted that more than 80% of AI-specialized cloud operators (NeoCloud) such as Coreweave will cease independent business operations by 2030.

This does not mean that the actual loss has been confirmed, but is a forecast based on the current investment plan and expected profits. Coreweave's decline in corporate bond prices and Gartner's warning reveal the burden of financing and profitability due to expanded AI investment.

#### Spreading anxiety over AI investment... Will even demand for semiconductors be shaken?

The financing uncertainty of U.S. AI infrastructure companies is also a burden on the domestic semiconductor industry. Samsung Electronics and SK Hynix are major beneficiaries of the growing demand for high-bandwidth memory (HBM) and high-capacity DRAM for AI servers.

If AI cloud companies delay the construction of new data centers or GPU introduction schedules due to the burden of borrowing, it may also affect server memory orders.

However, the decline in Coreweave's corporate bond price does not immediately mean a decrease in HBM or DRAM demand. What has now been confirmed is that the financial market's risk assessment of AI infrastructure companies has increased. It must be distinguished from a decrease in actual semiconductor orders or a change in existing supply contracts.

There are also signs that demand for semiconductor facility investment remains strong. ASML, a Dutch semiconductor equipment company, recorded sales of 11.7 billion euros and net profit of 4.96 billion euros in its third quarter results announced on the 14th. This year's sales forecast was also raised to 44 billion to 45 billion euros.

ASML CEO Christophe Fouquet predicted that customer demand for equipment will exceed existing production capacity plans for 2027 and 2028. Some customers are already showing interest in ordering equipment for 2029 and 2030. ASML is also considering ways to further expand production capacity in 2028.

Even as the financial cost burden of U.S. AI infrastructure companies increases, demand for long-term facility investment from semiconductor manufacturers is maintained.

The impact on Samsung Electronics and SK Hynix will depend on whether the financing anxiety of U.S. AI companies actually leads to a reduction in GPU orders or changes in memory supply schedules. For now, it is necessary to wait and see how much of an impact financial market instability will have on actual demand for semiconductors.