“Sell in May” Is the Wrong Frame for Korea Right Now

KOSPI investors enter May 2026 with a familiar anxiety: the seasonal “Sell in May and go away” refrain is getting louder. After a strong April rally across Korean AI infrastructure plays, the instinct to de-risk is understandable. But the right move in May is not a broad retreat—it’s a surgical rotation from overcrowded first-order winners into second-order proxies that haven’t fully priced in the same thesis.

This distinction matters more than the calendar.

Why the Seasonal Risk Is Real—But Narrower Than It Looks

April’s gains were unusually concentrated. Korean AI physical infrastructure stocks—power equipment, advanced PCBs, electronic materials, and select semiconductor names—ran hard. That kind of performance creates its own gravitational pull in May: profit-taking, institutional rebalancing, and theme rotation are all natural consequences.

Three specific pressure vectors are worth watching:

1. Brent and rates remain elevated. High oil prices combined with sticky interest rates compress growth multiples. For duration-sensitive AI infrastructure capex trades, this is a direct headwind. Foreign investor flows into Korean equities—already sensitive to USD/KRW movements—can turn quickly if the macro backdrop deteriorates.

2. Big-tech capex dependency is now a single point of failure. As BlackRock has noted, U.S. growth and equity markets are increasingly tethered to hyperscaler AI capex cycles. A modest downward revision to capex guidance from any of the major cloud providers could simultaneously pressure Korean semiconductors, power equipment, PCBs, data center materials, and AI accelerator supply chains. The correlation is no longer theoretical.

3. First-order winners are priced for perfection. Names like Hanwha Engine (082740.KS), Samsung Electro-Mechanics (009150.KS), and Doosan (000150.KS) have already absorbed much of the AI infrastructure re-rating. If “Sell in May” materializes, it’s most likely to show up as multiple compression in exactly these names—not a broad market selloff.

Why a Full De-Risk Would Be a Mistake

The case against blanket selling is equally concrete.

KB Securities’ May strategy note maintains an overweight on semiconductors and AI power infrastructure—institutional consensus has not flipped to risk-off. More importantly, South Korea’s April provisional export data, released by the Ministry of Trade, Industry and Energy, delivered a blunt signal that the AI hardware cycle is still running:

• SSD exports: +715% year-over-year
• Computers and consumer electronics: +242% year-over-year

These are not soft sentiment indicators. This is physical shipment data. Cutting core positions on seasonality alone while underlying demand prints numbers like these is a category error.

The internal rotation dynamic within the KOSPI also remains intact. Capital has been cycling between semiconductor names and AI power plays throughout Q1. In a breadth-supported, sector-rotating market, holding excessive cash is its own risk—the opportunity cost of missing the next leg of the rotation.

The Rotation Playbook: What to Trim, What to Add

The practical implication is a two-sided move: reduce exposure to overheated first-order winners while initiating small positions in event-dense second-order proxies.

Names to Slow Down on (Not Exit)

None of these are sells on thesis. They’re candidates for trimming size when the position has run beyond what the near-term catalyst calendar can support.

Second-Order Proxies Worth Watching

Two names deserve attention as May event catalysts approach—both with limited market recognition relative to the first-order plays:

Pamicell (005690.KS), a South Korean biotech and advanced materials company, has emerged as an indirect proxy for the AI substrate supply chain through its materials exposure to Doosan Electro-BG’s CCL (Copper Clad Laminate) production ecosystem. The stock has not repriced to the same degree as the upstream names. Key near-term triggers to monitor: Q1 2026 operating profit confirmation above ₩14 billion, a follow-on contract with Doosan, and a sustained break above the ₩20,000 level on meaningful volume.

Daeduck Electronics (008060.KS), one of South Korea’s established PCB manufacturers, represents a more direct second-order play on AI server infrastructure buildout. As hyperscaler demand for high-layer-count boards accelerates, Daeduck’s positioning in the advanced PCB supply chain becomes increasingly relevant—but at a valuation that still reflects older earnings expectations.

Both names should be treated as small, conditional entries pending event confirmation. Initiating full-size positions before Q1 results and contract newsflow is premature.

How to Think About Each Scenario

Markets rarely follow a single script. The May setup has four credible paths:

Normal rotation (base case): Semiconductors and AI power trade off leadership. KOSPI breadth holds. Core positions stay intact; small entries in second-order proxies proceed on schedule.

First-order profit-taking: Hanwha Engine, Samsung Electro-Mechanics, and power stocks pull back on crowding, while the underlying thesis (exports, capex) remains intact. This is the buy-the-dip scenario for second-order names.

AI capex shock: A hyperscaler cuts capex guidance. Semiconductors, power, PCBs, and data center materials sell off together. In this scenario: halt new buying, raise cash above 7%, trim unconfirmed positions including the second-order proxies.

Oil/rates risk-off: Brent stays elevated, U.S. 10-year yields spike, USD/KRW surges, foreign investors sell Korean equities. Raise cash, reduce overheated winners, hold only core names like Samsung Electronics (005930.KS) until the macro stabilizes.

The Bottom Line

May 2026 is not a month to fold. South Korea’s export data argues that the AI hardware cycle has real legs. The right question isn’t whether to own Korean AI infrastructure—it’s which part of the supply chain is still attractively priced.

First-order winners have earned their gains. Second-order proxies like Pamicell and Daeduck Electronics are where the next pricing opportunity may sit, if the thesis continues to propagate down the supply chain. The playbook for May: hold the core, harvest the overheated fringes, and enter the second-order names small—then size up only after the Q1 earnings and contract catalysts confirm.

Sell in May only if you’re selling the right things.

This analysis reflects Korean market conditions as of May 1, 2026, and references provisional export data from South Korea’s Ministry of Trade, Industry and Energy. Company fundamentals should be verified against the latest DART filings (dart.fss.or.kr) before making investment decisions.