The Democratic Party's control of both chambers of Congress following the U.S. midterm elections has raised concerns about potential regulatory changes affecting the artificial intelligence (AI) industry, which could impact South Korea's semiconductor sector. Fears are growing that the construction of data centers in the U.S. may face delays, leading to a significant drop in stock prices for major South Korean semiconductor companies like Samsung Electronics and SK Hynix. However, analysts note that there have been no confirmed cancellations of AI semiconductor orders or reductions in supply contracts, leaving uncertainty about whether political instability will translate into a slowdown in industrial demand.
According to financial investment sources, the Democratic Party secured control of both the House and Senate in the midterm elections held on November 3. This has led to predictions of intensified regulatory scrutiny over the AI industry and data centers. Key issues include energy consumption, water usage, and the financial burden on local residents related to data center construction.
Some Democratic lawmakers have proposed eliminating the 100% bonus depreciation benefit for data center investments. This provision allows companies to fully deduct eligible capital investment costs in the year they are incurred, thereby reducing initial tax burdens. Taxation measures related to AI usage and profits from AI companies have also been discussed. If these policies are enacted, data center operators may face increased investment costs, complicating decisions about new facility construction.
However, the Democratic Party's control of Congress does not automatically signal a shift in federal AI policy. The Bank of Korea's New York office has analyzed that the midterm election results are unlikely to significantly alter the overall policy direction established during the Trump administration. Even with control of both chambers, the Democrats do not possess enough seats to override a presidential veto. Therefore, any large-scale federal regulations on AI investments are more likely to emerge through congressional oversight and discussions, as well as state and local permitting regulations that could affect the investment environment.
At the state level, regulatory movements are already becoming apparent. Texas, a major hub for data centers in the U.S., has seen its newly elected Democratic governor, Gina Inohosa, advocate for a halt on new data center developments until protections for residents are established. This stance is based on concerns that large data centers could strain the power grid and increase electricity costs for local residents, as well as exacerbate water shortages.
According to the British energy market research firm Modu Energy, if newly elected governors fulfill their promises to halt new data center construction, approximately 11 gigawatts (GW) of new data center projects in the U.S. could be affected, with about 10 GW concentrated in Texas.
These concerns are resonating within South Korea's semiconductor industry. While data center regulations do not directly impact semiconductor manufacturing plants, a slowdown in AI infrastructure investment could indirectly affect major customers and the overall supply chain for domestic companies. High-performance memory semiconductors, such as graphics processing units (GPUs), high-bandwidth memory (HBM), and server DRAM, are essential for AI data centers. Delays in data center construction could postpone the establishment of AI servers and the introduction of GPUs, ultimately affecting the pace of demand growth for memory semiconductors.
SK Hynix, a key supplier of HBM to AI accelerator supply chains, is closely tied to the investment plans of global tech giants in AI infrastructure. Samsung Electronics is also expanding its HBM and server memory business, operating semiconductor production facilities in Austin, Texas, and constructing a foundry in Taylor.
The financial markets have reacted sensitively to the uncertainty surrounding these policies. The KOSPI index fell 5.28% on November 5 and dropped an additional 6.02% to close at 5,059.69 on November 6. During trading on November 6, the decline reached as much as 8.8%, triggering a circuit breaker. Samsung Electronics saw a 6.4% drop, while SK Hynix fell by 10.4%, with foreign investors selling approximately 2.8 trillion won worth of domestic stocks on that day.
However, some argue that it is premature to interpret the stock market decline as a direct indication of reduced semiconductor demand. According to the Seoul Economic Daily, customers of Samsung Electronics and SK Hynix have not shown signs of canceling existing semiconductor supply contracts or reducing order volumes due to the regulatory movements concerning data centers in the U.S. Production of semiconductors needed for data centers currently under construction is ongoing, and long-term supply contracts remain intact. The industry believes that current AI memory demand continues to exceed supply capacity, suggesting that even if some new projects are delayed, it will not immediately lead to a decrease in orders.
The key factors moving forward will be the actual implementation scope of regulations and changes in data center investment timelines. Even if permitting for new projects becomes more stringent, companies may choose to relocate investments to regions with more favorable power supply and permitting conditions rather than canceling projects altogether. Modu Energy estimates that if projects in regulated areas are shifted to other regions, the overall reduction in data center capacity across the U.S. could be limited to about 3 GW by 2030. This indicates that regional regulatory tightening does not necessarily lead to an immediate reduction in overall AI investment in the U.S.
Nonetheless, uncertainties remain in the medium to long term. If delays in data center construction persist and the profitability of investments by big tech companies falls short of expectations, the scale and pace of new AI infrastructure investments may be adjusted. Coupled with high interest rates and substantial capital investment costs, data center operators may find it challenging to secure funding, leading to potential delays in investment timelines. In such a scenario, demand for high-performance memory, including HBM, and semiconductor equipment could also be affected over time.
Ultimately, the future performance and stock trends of South Korea's semiconductor industry will likely be influenced more by the actual progress of data center construction and permitting, capital expenditure plans of big tech companies, and changes in HBM supply contracts and order volumes than by the political landscape in the U.S. While there is insufficient evidence to conclude that AI semiconductor demand has already peaked, the potential for increased regulatory scrutiny to lead to investment delays and cost increases could impact the growth trajectory of domestic companies, suggesting that market caution is likely to persist for the foreseeable future.
* This article has been translated by AI.
Copyright ⓒ Aju Press All rights reserved.