SK Hynix’s Dominance Test: Record Margins, a Strike Threat, and the Ghost of Cisco

SK Hynix hits record 22% KOSPI weighting and 72% operating margin, topping Nvidia, but Hana Securities warns of dot-com era parallels as supply squeeze persists.

SK Hynix has become a study in contradictions. Its operating margin just topped Nvidia’s, while its weighting in South Korea’s benchmark index is stirring uncomfortable memories of the dot-com era. The memory giant now accounts for roughly 22% of the KOSPI’s total market capitalisation — a record. The previous high was 13%, set by SK Telecom at the peak of the internet bubble in May 2000.

The rally has been staggering. Shares have jumped 160% year-to-date, accelerating to a 63% gain in the past 30 days. At current levels around 1.84 million won, the stock sits about 11% below its early-May record of 1.976 million won. Yet the real story is what lies beneath that price action: a structural supply squeeze that is reshaping how analysts value the company.

Hana Securities is sounding a cautious note. Analyst Lee Jae-man warns that the bull market may end the moment SK Hynix overtakes Samsung Electronics in market capitalisation — it already trades at roughly 85% of Samsung’s value. He draws a direct parallel to Cisco Systems in 2000, which briefly surpassed Microsoft and General Electric in market cap despite generating only 20% of GE’s net profit. Today, Samsung and SK Hynix together control nearly 48% of the KOSPI index, but their combined share of projected net profit is a far healthier 72%.

That earnings backing is what sets this cycle apart from the dot-com frenzy, at least for now. Samsung is forecast to post net profit of 280 trillion won in 2026, versus SK Hynix’s 208 trillion won. The valuation gap between the two is narrowing faster than the earnings gap — a trend that Hana Securities watches with unease.

The engine behind the revaluation is a global shortage of high-bandwidth memory. SK Hynix has publicly stated that its DRAM, NAND and HBM output are completely sold out. Microsoft, Google and Amazon are said to have made investment proposals to secure future capacity. The company’s first-quarter 2026 results underscore the pricing power: revenue hit 52.58 trillion won, operating profit reached 37.61 trillion won, and the operating margin of 72% exceeded even Nvidia’s 65% — an extraordinary feat for a maker of physical chips.

New capacity cannot come soon enough. The M15X fab in Cheongju, built at a cost of more than 20 trillion won, is already in pilot operation, with mass production scheduled for November 2026. A first plant in the Yongin cluster is due to be completed by May 2027, eventually adding around 350,000 wafers per month and lifting total capacity to 900,000. Goldman Sachs has raised its projection for the DRAM supply gap in 2026 from 3.3% to 4.9%, and market researchers do not expect meaningful price relief until after 2027.

A wild card is the labour dispute at Samsung. Some 50,000 workers plan a 18-day walkout starting 21 May. Samsung was the first to ship sixth-generation HBM in February, supplying Nvidia, AMD and Google. A prolonged strike could shift additional orders to SK Hynix, compounding an already tight market. Samsung’s yield on HBM4 is below 60%, while SK Hynix has already reached 80% on its latest DRAM process. Over the longer term, Samsung’s risk is that a successful HBM4 ramp in the second half of 2026 could erode SK Hynix’s market share to 50–60%.

Brokerages are embedding a structural shift into their valuations. Nomura doubled its price target to 4 million won, arguing SK Hynix has outgrown its cyclical roots. KB Securities and SK Securities both set 3 million won targets, the latter noting a forward price-to-earnings ratio of roughly 5.2 that defies conventional cycle-based metrics. UBS raised its target to 1.7 million won and lifted its 2026 and 2027 earnings forecasts by 22% and 29% respectively, calling it a memory super-cycle not seen in nearly three decades.

The next generation is already in the pipeline. SK Hynix plans to sample its more powerful HBM4E chips in the second half of 2026, with mass production starting in 2027. The core chip uses the industry’s first sixth-generation DRAM process, the 1c-nanometre node, which the company says is 11% faster and 9% more power-efficient than its predecessor. With a 30-day annualised volatility of almost 78%, there is little room for disappointment. The real test will come with second-quarter earnings in July — and how much longer the HBM shortage can sustain this radical reshaping of the KOSPI.