SK Hynix, Samsung, MU, TSM, NVDA lead AI memory run

Published on: May 28, 2026
Author: Brandon Kwan

Memory is the new oil and the wells are in Korea and Idaho. SK Hynix cracked the trillion-dollar barrier after a 9 percent session pop, joining Samsung and Micron as AI servers bid up every gigabyte of high-bandwidth memory. With new leveraged ETFs juicing flows and the KOSPI briefly flipping the sidecar, semiconductors were the most active tape in the past eight hours, and memory names owned the podium.

1) SK Hynix (000660.KS) – fresh trillion-dollar club entrant

Driver: A 9.3 percent close after an intraday sprint of nearly 15 percent launched the stock to 1,680 trillion won in market value, riding an AI-driven HBM supply squeeze. Prices for memory doubled in Q1 and are still climbing as data centers hoard capacity for Nvidia-class GPUs. Trading profile: Closed at 2.243 million won; now a trillion-plus heavyweight that, along with Samsung, makes up roughly half of KOSPI’s market cap. Target raised by Mirae Asset Securities to 3.8 million won on the view that memory demand outstrips supply into 2028. ETF flows added fuel as Korea’s new single-stock leveraged products kicked off with double-digit gains. Key takeaway: Hynix is the HBM bellwether and the market is paying up for a multiyear shortage. The risk is simple: any wobble in AI server orders or a faster-than-expected capacity catch-up will hit a stock priced for tightness.

2) Samsung Electronics (005930.KS) – the other half of Korea’s HBM duopoly

Driver: Momentum from the trillion-dollar milestone earlier this month plus relief as unionized workers approved a wage deal, removing a near-term production risk. Shares swung as much as 8 percent intraday and locked a record close up 2.7 percent to 307,000 won. Trading profile: A 149 percent year-to-date rocket that now dominates the KOSPI alongside SK Hynix. Ramping advanced HBM for AI datacenters with next-gen parts on the roadmap, while newly launched single-stock leveraged ETFs tied to Samsung amplified demand and futures positioning. Key takeaway: Samsung is the liquidity magnet and policy hedge. With ETFs pulling in retail cash and strike risk sidelined, execution on HBM yield and volume is the only conversation that matters. If HBM yield lags or pricing momentum fades, the downside gets real because the valuation already assumes near-perfect ramp.

3) Micron Technology (MU) – America’s memory comeback, supercharged

Driver: Broke into the trillion-dollar club as the stock ripped 19 percent in a single day, marking the 28th record close of the year. UBS more than tripled its target, arguing AI has structurally reset memory economics. Trading profile: A high-beta U.S. leader that has climbed 245 percent year to date on HBM and DDR5 tailwinds, with options and flows reflecting nonstop buy-the-dip muscle memory. Benefiting from the same scarcity dynamic pushing Korean peers, and a U.S. investor base now conditioned to treat every pullback as an allocation window. Key takeaway: Micron is the pure-play U.S. way to own the HBM shortage. The cycle is doing the heavy lifting, but at a trillion cap the market now expects Micron to convert tight HBM and DRAM pricing into durable free cash flow, not just a one-quarter beat. Miss that pivot and the stock’s volatility will remind you it is still a memory name.

4) TSMC (TSM) – the quiet trillion that prints AI silicon

Driver: Not the day’s biggest mover, but the foundry remains the fulcrum of the AI buildout and one of the few Asian names already in the trillion-dollar club. News attention stayed elevated as investors triangulated memory supply tightness with upstream logic demand for AI accelerators. Trading profile: The premier advanced node foundry with a dominant share in cutting-edge logic and advanced packaging, supplying the ecosystem that consumes every HBM stack SK Hynix and Samsung can ship. Liquidity is deep, and the ADR is the cleanest way for global capital to own the AI capacity ramp. Key takeaway: If memory is the choke point, TSMC is the gatekeeper. Its backlog and pricing power tie directly to how fast the AI train runs and how constrained HBM stays. The name works as long as AI capex doesn’t blink.

5) Nvidia (NVDA) – the demand engine for every HBM wafer

Driver: Even when it is not the biggest print of the day, Nvidia remains the headline magnet because every AI server sold needs HBM-rich accelerators. The article of faith on the desk: GPU shipments are gated by memory availability, not desire to buy. Trading profile: Mega-cap with world-beating liquidity and a pipeline that keeps slipping into record territory as new accelerators roll. The company’s bill of materials is now a macro indicator for HBM capacity and pricing leverage at its suppliers. Key takeaway: If you want a simple read-through from HBM scarcity, it is that Nvidia’s unit growth rides whatever memory the duopoly can deliver. Watch memory supply announcements and yields to time digestion phases; a surprise loosening in HBM would pull some heat out of NVDA’s multiple, while persistent tightness keeps the flywheel spinning.

Investor Lens

AI has turned memory into a pricing power story again, and the market is voting with both hands. The setup is straightforward: if HBM remains structurally scarce, SK Hynix, Samsung, and Micron keep compounding pricing and mix, while TSMC and Nvidia monetize the upstream and downstream of that same bottleneck. The risk runs through exuberant flows and leverage. Korea’s debut of single-stock leveraged ETFs juiced futures and spot, triggered a sidecar halt, and left foreign investors net selling into retail enthusiasm. That feedback loop can cut the other way just as fast. The bull case needs two things to keep working: AI capex that does not pause and memory supply that stays tight closer to 2028 than to next year. Keep your eyes on HBM capacity adds, yield milestones, and any signal of a demand air pocket in AI servers. In this tape, those three lines will tell you more than any victory lap about joining the trillion-dollar club.

AI Lithium