Nomura Securities has raised its target price for SK hynix to 4 million won and for Samsung Electronics to 590,000 won. The move reflects a reassessment of memory semiconductors not as cyclical stocks, but as structural growth stocks.
There is no single view in the brokerage community. Optimism based on surging data center investment and caution over weakening cash flow at big tech firms are emerging at the same time.
A target price of 4 million won is a number built on the premise that SK hynix is no longer a cyclical stock. If that premise wavers, the valuation changes as well.
AI demand is also spreading beyond memory into robots, power, and components. As an alternative to concentration in a single semiconductor stock, adjacent industries are coming into view.
◆ A 4 million won target price, a number created by a shift in valuation
Nomura Securities has raised its target price for SK hynix to 4 million won and for Samsung Electronics to 590,000 won. (Photo = Nomura)
KOSPI, which briefly broke through the 8,000 mark during trading last week, then turned sharply lower. As market volatility increased, investors focused their attention on semiconductor stocks that had driven the index higher.
According to FnGuide, a financial data provider, Samsung Electronics and SK hynix ranked first and second among the stocks most searched by domestic investors from the 12th to the 18th of this month. The keyword “semiconductor” also ranked first among search terms.
The most searched report during that period was a report by Japan’s Nomura Securities. On the 15th, Nomura raised SK hynix’s target price from 2.34 million won to 4 million won, and Samsung Electronics’ from 340,000 won to 590,000 won. It was the first time a target price in the 4 million won range had been set for SK hynix by the brokerage community.
What stands out is not the target price itself, but the method used to derive it. Nomura cited a “shift in valuation methodology” as the key reason for the sharp increase. Memory semiconductors have traditionally been valued using the price-to-book ratio (PBR), which is based on asset value, because of the industry’s boom-and-bust cycle.
Nomura changed that premise. It judged that, with the spread of artificial intelligence, memory demand has moved from a temporary cycle to a structural growth area.
When the premise changes, the valuation yardstick changes as well. Nomura believes SK hynix should be valued using the price-to-earnings ratio (PER), which looks at stock price relative to earnings, because it is a company that generates stable profits.
Its current forward 12-month PER is about 6 times, but Nomura argued it deserves the same treatment as TSMC, the world’s top foundry company, which trades at around 20 times PER.
In the end, 4 million won is a number derived only if the premise that “SK hynix is no longer a cyclical stock” holds true. That is why, when interpreting a target price, the figure and the assumptions behind it must be separated.
◆ A divided brokerage community, with optimism and caution pulling in different directions
Another feature of this phase is that brokerage houses are not converging on a single view.
The basis for optimism is expanding data center investment. Nomura estimated that global data center capital expenditure will rise more than fivefold, from $1.16 trillion this year to over $5 trillion in 2030. Its analysis is that memory demand for AI operations will outpace supply.
Domestic brokerages are making similar assessments. On the 13th, Noh Geun-chang, a researcher at Hyundai Motor Securities, issued a report titled “The Chasm We Fear May Not Come,” and presented a target price of 2.65 million won for SK hynix. Noh said that if demand for AI semiconductors broadens across cloud AI, edge AI, and physical AI, the depth and duration of the chasm could weaken.
There was also a contrasting view. Lee Min-hee, a researcher at BNK Investment & Securities, noted in a report titled “Risks Grow Even as Everyone is Optimistic” that actual capital expenditures by U.S. cloud providers in the first quarter of this year came in at only 98% of market expectations. Although annual investment plans have been raised, the increase was led by Meta and Microsoft, whose free cash flows are deteriorating rapidly, which lowers credibility, Lee said.
The two views focus on different time horizons. Optimism weighs the long-term trend of structurally rising demand, while caution focuses on the possibility that the investment capacity of big tech companies supporting that demand may weaken in the short term. Changes in quarterly capital expenditure plans at big tech firms and memory price trends are expected to become the dividing line between the two scenarios.
◆ AI demand spreading beyond memory, reshaping the investment landscape
As semiconductor stocks rose sharply in a short period, volatility also increased. As “30 million-won Samsung” and “2 million-won Nix” came into view, profit-taking sell orders flooded in, and as of the 18th, the shares had fallen back to the 280,000-won range and 1.8 million-won range, respectively.
LG Twin Tower in Yeouido (Photo = LG Electronics)
Amid this, the market is detecting a shift in attention from semiconductors to adjacent fields. During the same period, a report on LG Electronics ranked second among the most searched reports.
At its shareholders’ meeting in March, LG Electronics announced plans to establish a mass-production system for key components for humanoid robots within the year and to verify AI home robot technology.
This trend shows that the demand created by AI does not remain confined to memory alone. As AI computing expands, memory is needed to hold that computing, power and cooling facilities are required to cool the memory, and demand then extends to robots and components that will implement AI in the physical world. Demand spreads along the industrial chain.
From an investing perspective, this provides a clue for diversification of risk. Rather than concentrating in one stock, investors may spread risk along the path of AI demand. Even in a phase where memory prices are correcting, power, cooling, and robot component sectors may move at different speeds.
However, adjacent sectors are not free from risk. Robot businesses are still driven more by expectations than by earnings, and stocks whose prices already reflect expectations can also experience sharp pullbacks. Even among stocks grouped under the label of “AI beneficiaries,” business maturity and earnings visibility vary widely.
The key point in this phase is not the target price number itself, but the assumptions supporting it. Four million won is not a confirmed future, but a carefully constructed outlook.
Big tech companies’ investment capacity, memory price trends, and whether the chasm arrives will determine the direction of those assumptions. In a volatile market, the question the market is asking is not “Is this the peak?” but “What signals will cause those assumptions to break down?”