[Edaily Reporter Kim Hyun-ah ] Shares of AI data center-related companies plummeted after the Democrats secured control of both the Senate and the House of Representatives in the U.S. midterm elections. On November 5 (local time), the Nasdaq fell 5.5%, and CoreWeave dropped by about 10%. Shares of NeoCloud companies such as Nevius and iRen also fell sharply. The Nasdaq 100 lost approximately $2.5 trillion in market capitalization in a single day. This was the result of concerns over data center regulations compounded by doubts about companies’ debt burdens and the profitability of their investments.

The market is focusing on the possibility that a Democratic victory could lead to stricter data center regulations. If conflicts over electricity and water usage, as well as the burden on local communities, intensify, the costs of obtaining permits and constructing new facilities could rise.

However, it is difficult to say that the AI industry’s growth momentum has been halted. The technological hegemony competition between the U.S. and China continues, and as AI models become more sophisticated, the demand for computational resources and electricity to support them will inevitably grow.

What is changing is the investment approach. The industry is moving away from aggressive expansion—where companies first secured GPUs and then sought customers—and entering a phase where actual demand and profitability are being evaluated. The process of separating the wheat from the chaff in the AI data center market has begun.

CoreWeave, SK, and Naver Have Different Investment Structures

A key point to note regarding this week’s stock decline is that data center operators have different investment structures and financial conditions.

Specialized neo-cloud companies like CoreWeave secure GPUs on a large scale and lease them to AI companies. While this is advantageous during periods of market growth, it becomes a significant burden if customer acquisition is delayed or funding costs rise. This is why their debt, totaling approximately $50 billion, is fueling concerns.

In contrast, SK(034730)and Naver (NAVER(035420)) are expanding their AI infrastructure based on existing businesses such as semiconductors, telecommunications, search, advertising, and e-commerce. This approach shares the investment burden by combining revenue and cash flow from existing businesses with funding from global investors.

SK has established a business foundation ranging from semiconductors to telecommunications networks and data centers. SK hynix’s competitiveness in high-bandwidth memory (HBM) is also a key differentiator. As the number of AI data centers increases, demand will grow not only for GPUs but also for HBM, which can quickly process large volumes of data.

SK’s data center business is divided into the operation and expansion of existing facilities and new development. SK Horizon is responsible for securing and managing a total of 318 MW of infrastructure—comprising existing operational facilities and data centers currently under construction—as well as overseeing their operation and expansion. SKTelecom(017670)secured approximately 3.08 trillion won in funding from the KKR-IMM consortium. Once the transaction is finalized, SKTelecom will hold a 51% stake, while KKR and IMM will hold 29% and 20%, respectively.

SK Hyper, a wholly-owned subsidiary of SKTelecom, will handle new development projects. The goal is to build AI data centers with a total capacity of up to 15 GW by 2035. The first phase, totaling 5 GW, is scheduled to begin operations in 2029.

In an interview with Bloomberg TV, Jeong Seok-geun, CEO of SK Hyper, estimated the cost of building the first 5 GW—including servers and equipment—at approximately $350 billion (about 464 trillion won). This figure represents the cost for the first phase, not the entire 15 GW project. SKTelecom plans to invest up to 750 billion won by 2030, with the remainder to be raised through global big tech companies and financial investors. The plan is for tenant companies and investors to share the costs and adjust the pace of investment according to actual demand.

Naver is also sharing the investment burden. Brookfield will invest up to $9 billion to secure AI infrastructure, such as GPUs and data centers, and Naver Cloud will utilize this infrastructure through long-term contracts. Naver has decided to link usage fees to business revenue to reduce the initial investment burden. However, to prepare for the possibility that Naver Cloud may be unable to pay infrastructure usage fees, Naver will guarantee payment for up to eight years.

What the two companies have in common is that they are not building large-scale infrastructure solely with their own funds. Their strategy is to collaborate with global financial capital to share investment risks and scale their businesses in line with actual demand.

SpaceX Also Expanding Its Business Through AI Infrastructure

NeoCloud is not the only company entering the AI infrastructure market. There are also companies that are leveraging capital and technology secured from their existing businesses to cultivate the AI computing market as a new growth engine.

SpaceX, which has grown through rocket launches and the Starlink satellite communications service, is a prime example. According to its third-quarter earnings report, revenue from the Superintelligence business segment totaled approximately $7.6 billion, accounting for 52% of total revenue. This segment includes revenue from cloud service contracts, the acquisition of Cursor, and sales of Grok models and software. SpaceX’s business scope, which had been concentrated on space and satellite operations, is expanding into AI infrastructure and software.

SpaceX has signed a cloud services contract with Anthropic that allows for the use of approximately 325,000 NVIDIA GPUs. As AI companies find it increasingly difficult to secure the necessary computing resources through their own data centers alone, the role of external providers in supplying GPUs is growing.

However, securing GPUs alone does not guarantee business success. If revenue is heavily dependent on a specific large customer, contract reductions or shifts in demand can directly impact facility utilization rates and profitability. Securing long-term contracts with customers who will use these resources consistently is just as important as the GPUs themselves.

Business Competitiveness Is More Important Than the AI Bubble

Some view the AI infrastructure market as comparable to the

dot-com

bubble. Although many companies disappeared due to excessive investment in the early days of the internet industry, the industry itself continued to grow. The same may hold true for the AI market. Companies that rely on funding to acquire an excessive number of GPUs or fail to secure customers and power supplies may be weeded out, but demand for AI infrastructure itself will not disappear.

The results of the U.S. midterm elections have presented the data center industry with challenges in the form of regulation and community acceptance. Even as the AI industry grows, data centers cannot expand indefinitely. Securing power, obtaining permits, and consulting with local communities have all become part of a company’s competitive edge.

The criteria for determining winners in the AI data center market are also changing. Rather than how many GPUs a company has secured, it has become more important to have a stable customer base, ensure a timely supply of the necessary power, and possess a revenue model capable of recouping investment.

This does not mean that investment in AI data centers is coming to an end. A market that once prioritized the scale of investment has now begun to focus on actual demand and business competitiveness. This week’s stock decline does not signal the end of AI infrastructure but rather the beginning of a process to separate the wheat from the chaff.