저자: 미상 · 외국인
요약 보기
SK하이닉스는 AI 수요로 HBM 시장을 선점해 2025년 사상 최대 영업이익(47조원)과 2026년 1분기 영업이익률 72%를 기록했다. HBM 시장 점유율 56.4%와 엔비디아와의 장기 협력이 강점으로 제시된다. 다만 주가가 12개월간 500% 이상 올라 호재가 이미 상당 부분 반영됐고, 삼성·마이크론의 추격, 공격적 증설에 따른 공급 과잉과 메모리 사이클 재진입 위험이 제기된다. 결론은 '성장 기업이면서 여전히 경기순환 기업'이라는 양면적 평가다.
다듬은 글
SK hynix is a South Korean memory-chip company founded in 1983. For decades, it made the kind of technology almost everyone used and almost nobody noticed. Its chips sat inside phones, laptops and servers, usually hidden behind the name on the finished product. But memory has always been a tough business.
When demand is strong, prices rise and manufacturers expand production. Then supply catches up, prices fall… and profits can too. That’s exactly what happened in 2023, when SK hynix recorded an operating loss of nearly 8 trillion won. But then AI came along and created enormous demand for a specialised kind of memory: high-bandwidth memory, or HBM.
AI systems don’t just need powerful processors. They need memory fast enough to keep those processors fed with data. And SK hynix had already spent years building a lead in it. By 2025, SK hynix had swung to a record 47 trillion won in operating profit. In the first quarter of 2026, the company reported an operating margin of 72 percent.
And by mid-July 2026, its shares had risen more than 500 percent in twelve months. After a rise like that, you might assume the opportunity had already passed. But investors are still paying attention because SK hynix’s early lead in HBM has turned into market dominance, and they’re betting the market could get much bigger.
So, are they right? To understand that, it helps to know what SK hynix actually sells. Its two main businesses are DRAM and NAND. Here’s the simple version: DRAM is short-term working memory. It holds the data a device needs while it’s running. NAND is long-term storage. That’s what keeps your files saved when the power is off.
Then there’s HBM, a specialised form of DRAM. Normally, memory sits off to the side of the processor, with data travelling down a relatively narrow path. HBM stacks the chips vertically and places them right next to the processor. The result is that far more data can move at once, over a much shorter distance, while using less power.
And that’s exactly what AI needs. Now, SK hynix isn’t the only company making HBM. Samsung and Micron are competing for the same customers, but SK hynix has one major advantage: a head start. The company helped develop the first generation of HBM back in 2013, long before the AI boom. And it spent years improving the technology while the market was still small.
HBM is notoriously difficult to manufacture. So when demand took off, SK hynix already had more than a decade of experience developing it and learning how to produce it at scale. That experience helped it capture 56.4 percent of the global HBM market by revenue in the first quarter of 2026. Its relationship with NVIDIA makes that position even stronger.
In June 2026, the two companies announced a multi-year partnership to develop memory for NVIDIA’s future AI systems. That means SK hynix isn’t just providing what AI needs today. It’s positioning itself to be one of the key suppliers for the AI hardware of tomorrow. And that’s the big investment case.
SK hynix expects the HBM market to grow by roughly 30 percent a year until 2030 as AI data centres continue to expand. Each new generation of AI chip also tends to need more memory. So even if chip shipment growth slows, overall demand for memory can still rise. And as AI moves from training models to running them for millions of users, data centres may need more conventional memory and storage as well.
That could spread the benefit across SK hynix’s wider DRAM and NAND businesses. The bull case is simple: SK hynix leads the HBM market, demand is growing, and it’s helping develop memory for NVIDIA’s future AI systems. But after a 500 percent rise in its stock price, a lot of good news is already priced in.
SK hynix is no longer an overlooked stock. That doesn’t mean the shares can’t go higher, but it does mean there’s a lot less room to disappoint. What could those disappointments look like? For one, competitors like Samsung and Micron are investing heavily to catch up. And even though NVIDIA is a key partner, it’s still working with all three suppliers.
SK hynix is also spending heavily to expand production. New factories take years to build, so the company has to make decisions now based on what it thinks demand will look like much later. As long as AI spending keeps rising, that extra capacity could be hugely valuable. But if it slows just as new supply hits the market, those record margins could fall fast.
And that could drag SK hynix back toward the same problem it faced just a few years ago: the boom-and-bust memory cycle that caused its 8-trillion-won operating loss in 2023. So is SK hynix an AI company, or still a cyclical memory company? Right now, it’s a bit of both. And that’s what makes the stock so interesting.
If it can hold its lead in HBM while the market keeps growing, AI could turn its head start into a lasting advantage. But if competitors catch up and prices fall, investors may be reminded that memory is still one of the most cyclical businesses in tech. What do you think? Is SK hynix’s HBM dominance here to stay, or is the market pricing in profits that may not last?
That's all for today's video. Let us know your thoughts in the comments and what you want us to cover next.
Comments 0