U.S. Data Center Boom Shows Cracks

In the United States, the world's largest AI infrastructure market, at least 47 large-scale data center projects have encountered setbacks this year. This has raised doubts about the resilience of the $10 trillion AI infrastructure boom, as reported by the Financial Times. Factors contributing to these issues include deteriorating corporate financial conditions, power shortages, and opposition from nearby residents.

On October 18, the FT cited data from RegLayer, indicating that at least 47 U.S. data center development projects have been delayed, abandoned, or halted this year.

Investment bank Bernstein estimates that only 35% of the planned data center construction projects in the U.S. are currently viable. The primary cause of these disruptions is cited as funding shortages, with CoreWeave, an AI infrastructure company, experiencing a significant rise in its corporate bond yield to 15%. CoreWeave, which leases AI chips to tech companies, has also faced challenges as investor expectations for its IPO have become much lower than initially anticipated.

The overheating competition among companies in the AI industry has created a "Wild West" scenario, with interest rate hikes and a slump in the bond market further threatening the ambitions of AI infrastructure firms. Power shortages and local opposition, who view data centers as undesirable facilities, have also hindered progress. Oracle, a big tech company, recently notified the developer of the 2.45GW 'Project Jupiter' in New Mexico of force majeure, preparing to delay rent payments if power is not secured as planned.

PJM, the largest power grid operator in the US, halted a planned special auction for 15 years of power for data centers due to the AI-driven power shortage. Concerns over AI infrastructure companies are not limited to the U.S., as Nscale, a British neocloud company, is facing difficulties in pursuing an IPO due to lower-than-expected investor expectations. Similarly, Australian AI infrastructure company Firmus withdrew its IPO plan, valued at about $5 billion.

These crises for AI infrastructure companies may ultimately impact the earnings of semiconductor manufacturers. Delays or reductions in AI infrastructure construction projects could lead to postponed or diminished orders for semiconductor chips used in data centers. Recent weakness in the stock prices of domestic memory companies, such as Samsung Electronics and SK hynix, is attributed to deteriorating investor sentiment due to concerns over U.S. interest rate hikes and the uncertain long-term profitability of AI technology.