South Korea’s storage production expansion, Meta rents computing power — Nomura discusses “two major bearish factors for storage”

Nomura believes that sentiment-driven misjudgments have obscured structural opportunities. The two so-called “bearish” factors—South Korea’s plans to expand storage capacity and Meta’s leasing out computing power—are, in fact, false propositions.

Recently, the global storage chip market has been shrouded by two major “bearish” rumor waves: first, South Korea’s storage giant has put forward an enormous capacity expansion plan, triggering deep market panic over potential future oversupply; second, Meta announced that it will lease out idle computing power to outsiders, which some funds interpret as a dangerous signal that AI hardware demand may have peaked.

However, according to information from Zhuifeng Trading Desk, Nomura Securities said plainly in a research report dated July 2 that market concerns have been seriously exaggerated. As for the current situation, the truth is:

• South Korea’s investment plan of as much as 48 quadrillion won will take at least 5 to 10 years to be converted into actual production capacity—distant water cannot quench immediate thirst. Moreover, the squeeze on general storage production capacity caused by high-profit HBM (high-bandwidth memory) is leading the market to face a severe supply shortage;

• At the same time, Meta’s leasing of computing power will not weaken hardware demand. Instead, by lowering Token costs, it will trigger the “Jevons Paradox,” thereby stimulating an even larger incremental demand for AI.

Overall, Nomura believes that the core contradiction in the global storage industry today is still a severe supply shortage. AI-driven structural demand growth has not yet peaked. Investors’ worries about oversupply are understandable, but clearly excessive—the market’s overreaction may provide a window for the storage sector to re-examine valuations.

South Korea’s Massive Capacity Expansion Plan: Distant Water Cannot Quench Immediate Thirst, and Oversupply Fears Are Greatly Exaggerated

Recently, South Korean storage companies and their affiliates, together with the government, announced a large-scale medium-to-long-term investment plan with no clear timetable. The total amount is as high as 4.8 quadrillion won (of which 3700 trillion won is directly related to storage). This huge figure has quickly intensified investors’ concerns about storage chip oversupply.

But Nomura pointed out that the so-called conspiracy theory that “global storage companies collude on prices by controlling supply,” as well as fears of oversupply capacity, is baseless.

• First, what the market is facing is an extremely severe supply shortage, not oversupply.

With unprecedentedly strong demand in the AI industry, storage companies have no choice but to prioritize producing high-margin HBM chips. This shift in capacity allocation directly leads to slower growth in production of general storage chips.

Since the second half of 2025, strong growth in general storage demand has already triggered severe supply shortages. Although storage companies are expanding capacity with an even more aggressive posture than expected, they still cannot meet the massive demand in the market.

• Second, the cycle for semiconductor investment to be converted into actual production capacity is extremely long.

The government’s involvement in South Korea is mainly because the existing production clusters of companies are about to exhaust their capacity to carry demand (land, electricity, water resources). The government needs to support the construction of new medium-to-long-term clusters after 2035.

For example, the super project “Yongin Semiconductor Cluster” launched 9 years ago is expected to complete its first cleanroom only by February 2027, and small-scale production can be achieved only by the end of the year—meaning it actually takes more than 10 years from investment to production.

Nomura expects that the newly announced investment plan will have a material impact on the market no sooner than 5 to 10 years later.

• Finally, the industry’s risk-resilience mechanisms have undergone structural changes.

In the past, cyclical fluctuations in the storage industry often stemmed from insufficient investment during downturns or record-level investment when demand surged. But now, companies not only have long-term agreements (LTAs) as a hedging tool, but also benefit from AI-driven structural growth expectations.

In addition, a profit-linked employee bonus mechanism also creates a new kind of buffer against the risks of managing overcapacity and profit decline. Companies will never make unnecessary blind investments merely because of government requirements.

Meta Leasing Idle Computing Power: Not a Sign of Demand Peaking, but an Imitation of AWS to Improve Capital Return

The market’s second concern stems from Meta’s decision to sell its excess computing capacity to external customers, which some view as a precursor to weakening AI memory and hardware demand.

Nomura explicitly rejected this, arguing that it is merely a natural evolution as the business model matures. Its logic is essentially the same as when Amazon created AWS cloud services to monetize idle data centers.

• First, leasing computing power is a necessary choice to address “peak redundancy.”

Building data centers inherently requires matching demand for “peak computing power,” which means that during non-peak hours and seasons, a large amount of computing power will remain idle. Meta’s core business (social networks and advertising) sees utilization rates fluctuate greatly across different periods.

Among cloud service providers (CSPs) that use data centers for both internal and external purposes, Meta is the only company that has not yet entered the cloud business. As economies of scale develop, following xAI’s approach of selling surplus computing power externally is an extremely natural decision to improve Meta’s return on invested capital (ROIC). If it does not do so, expanding capacity would result in a huge waste of resources.

• Second, the released computing power will nourish a larger AI ecosystem.

The computing power Meta sells to the outside world will become an important resource for companies such as Anthropic and OpenAI—firms that lack their own data centers but are extremely eager for computing power to provide enterprise-level AI services.

• Third, it triggers the “Jevons Paradox,” creating more incremental demand.

Nomura emphasized that Meta’s decision is by no means a turning point indicating a reduction in AI-related hardware demand. Instead, because the current shortage of computing power supply is pushing single Token prices upward, Meta’s entry of computing power is expected to help Token prices stabilize and move downward.

According to the “Jevons Paradox” (technological progress reduces the cost of using a resource, which instead increases the total consumption of that resource), the reduction in usage costs will create a larger scale of new AI demand. Over the long run, this will further consolidate underlying demand for memory and computing hardware.

Risk Warning and Disclaimer

        There are risks in the market; investments require caution. This article does not constitute personal investment advice, and it does not consider the specific investment objectives, financial situation, or needs of any individual user. Users should consider whether any opinions, viewpoints, or conclusions in this article are suitable for their specific circumstances. You bear all responsibility for any investment made based on this.