A moment to bet on national fortune? South Korea bets trillions of dollars on semiconductors, while the Bank for International Settlements warns of an AI bubble.

TL;DR

· South Korean President Lee Jae-myung will preside over three “mega projects,” elevating semiconductors, AI data centers, and physical AI to national-level investment priorities.

· The BIS warns that the top five cloud giants’ AI capital expenditures have already exceeded their earnings and free cash flow, with demand-side financing pressure beginning to factor into pricing.

· Related names: Samsung Electronics, SK hynix, NVIDIA, Microsoft, Google, Amazon, Meta, Oracle, power equipment and copper supply chain.

According to a Reuters report on June 29, South Korean President Lee Jae-myung will host an industry briefing to unveil three major projects covering semiconductors, AI data centers, and physical AI. South Korean media estimates that related long-term investments by Samsung and SK could potentially exceed 1,000 trillion won.

This news pushes the AI hardware cycle to a new level. On the supply side, national coordination of land, electricity, water resources, and regional layout is required, while on the demand side, macro institutions are questioning the sustainability of cash flows.

At nearly the same time, the BIS, in its 2026 Annual Economic Report, cautioned that the top five hyperscalers’ AI-related capital expenditures from 2025 to 2026 will exceed $1 trillion, with related commitments already surpassing earnings and free cash flow, prompting some companies to raise funds through debt issuance.

South Korea sees a window to seize the hardware gateway for AI, while the BIS sees U.S. cloud giants continuing to leverage up for their AI platform positions. What investors need to assess is not whether AI demand exists, but whether demand can continue to be supported by revenue, profits, and cash flows.

South Korea’s Aim: The Hardware Gateway

By elevating these projects to the presidential level, South Korea is essentially competing for the most irreplaceable hardware gateway in the AI supply chain.

For average investors, HBM (High Bandwidth Memory) can be understood as a high-speed memory stack adjacent to AI chips. AI training and inference require not only NVIDIA GPUs but also memory to continuously feed data. SK hynix leads in HBM, while Samsung is catching up.

The projects also cover AI data centers and physical AI. Physical AI refers to embedding models into robots, equipment, and factory systems, enabling AI to perform operations in the real world. Demand may come from manufacturing, logistics, robotics, and edge devices.

According to reports, South Korean policy advisor Kim Yong-beom believes AI demand is growing faster than expected, and South Korea cannot simply let companies expand individually—it must plan electricity, land, water resources, and regional layout together. Some discussions also involve areas outside the capital region, such as Gwangju and Jeolla, reflecting that infrastructure constraints have become a bottleneck for expansion.

The market implication of this round of South Korean investment is not simply new semiconductor spending, but an attempt to turn the corporate expansions of Samsung and SK into South Korea’s long-term position in the AI hardware supply chain.

BIS Focuses on Demand-Side Cash Flow

The BIS’s warning does not directly label AI as a bubble; it points to the gap between capital expenditure and cash flow.

Capital expenditure is money a company spends in advance for future growth, such as building data centers, buying GPUs, or expanding power systems. Free cash flow is the money a company has left after covering necessary expenses. The BIS notes that the top five cloud giants’ AI capital expenditures have already exceeded their earnings and free cash flow, and expansion is increasingly reliant on financing arrangements.

This aligns perfectly with South Korea’s supply-side bet. South Korea expands HBM, advanced packaging, data centers, and robotics hardware, while U.S. cloud giants buy AI servers, chips, power, and data center space. The former’s investment confidence comes from the latter’s order expectations.

The BIS also presents a more cautious scenario: under competitive pressure, companies may overcommit to investment to avoid losing their AI platform positions, even if the industry’s net economic surplus eventually turns negative. Similar processes occurred during the railroad and internet bubbles—the technology direction was valid long-term, but capital allocation overheated at a certain stage.

This is also where the market needs to differentiate, especially after related reports caused Samsung and SK hynix stock prices to dip briefly that day. Public reports attributed the decline more to a global tech stock consolidation after a rally, rather than a single policy factor. But the stock price still raises a question: when massive capacity expansion meets return validation, supply chain leaders will face reassessment of depreciation, utilization rates, and order visibility.

Nationalizing Supply Amplifies Demand Validation

South Korea’s move strengthens the AI hardware cycle and also magnifies its dependence on U.S. cloud giants’ capital expenditures.

In an optimistic scenario, U.S. cloud giants continue to raise AI spending, and enterprise AI services, cloud revenue, and inference demand gradually cover chip, depreciation, power, and interest costs. South Korea, by pre-positioning in HBM, advanced processes, packaging, and data centers, will build a national industry moat.

In a more cautious scenario, problems will propagate backward along the same chain. If cloud giants find that AI revenue growth cannot cover total costs, they may delay data center construction, cut server orders, or even renegotiate supply chain prices. The capacity, industrial parks, and power investments South Korea pre-built would then face utilization and return pressure.

What is most easily misinterpreted here is the investment amount. In South Korean media, there are different figures for Samsung and SK’s related long-term investments, such as “over 1,000 trillion won” or “a total of about 2,000 trillion won.” A safer understanding is that these are high-level plans or estimates on a decade-long scale, not fully finalized government commitments, nor necessarily smoothly absorbed by orders every year.

The South Korean government needs to solve the national industry position; U.S. cloud giants need to solve AI commercial returns. The former can be advanced through policy, land, and infrastructure, while the latter must ultimately be validated by revenue, profit, and cash flow.

Power and Utilization Determine Valuation Flexibility

What will first be validated in this round of the AI hardware cycle may not be grand productivity narratives, but a few harder variables: whether power is in place, whether factories are built on schedule, whether HBM and server orders remain sustained, and whether cloud giants can still support capital expenditures with operating cash flows.

Power is especially critical. AI data centers have ongoing demand for power grids, transformers, copper, and cooling systems. If South Korea’s planned regional data centers and semiconductor clusters advance simultaneously, power equipment and commodities will benefit. If grid expansion lags behind project pace, investment commitments will first turn into cost pressure.

Another variable is utilization rates after 2027. The current market assigns a high valuation to the AI supply chain based on the assumption that training and inference demand will continue to expand. Once cloud giants’ capital expenditures shift from acceleration to deceleration, what changes first may not be technical judgment, but order visibility, inventory cycles, and depreciation pressure.

South Korea’s national bet this time acts more like an amplifier. If AI returns materialize, it will amplify Samsung and SK hynix’s positions in the global supply chain. If returns fall short, it will also amplify the speed of capital expenditure backlash. For investors, the most concrete validation point going forward is whether U.S. cloud giants’ cash flows can continue to catch up with the AI spending they have already committed to.

Click to learn about open positions at BlockBeats

Welcome to join the official BlockBeats community:

Telegram Subscription Group: https://t.me/theblockbeats

Telegram Discussion Group: https://t.me/BlockBeats_App

Twitter Official Account: https://twitter.com/BlockBeatsAsia