Reading time: 10 min
The South Korean memory chip maker is sitting at roughly $900 billion in market capitalisation today. Its Q1 2026 operating margin just came in at 72 percent, beating Nvidia. Its entire production for the year is already sold out. And Microsoft, Google and Amazon are reportedly offering to fund its factory expansion in exchange for guaranteed future supply. This is not a hype story. The numbers are real and they are extraordinary.
Let’s Start With What Just Happened
On May 11, SK Hynix shares surged more than 15 percent in a single session, hitting an intraday all-time high of 1.949 million Korean won. That move pushed the company’s market capitalisation above $900 billion, per TradingKey data, placing it on the threshold of a milestone that would have seemed laughable eighteen months ago. Just over a year ago, SK Hynix was valued at under $100 billion. The stock is now up more than 200 percent in 2026 alone, following a 274 percent gain in 2025. If it crosses the trillion-dollar line, South Korea becomes the first country outside the United States to host more than one trillion-dollar company simultaneously, with Samsung having crossed that threshold earlier this month.
To put the speed of this re-rating in perspective: SK Hynix has added roughly $800 billion in market value in a little over twelve months. That is faster than almost any comparable move in the history of large-cap equities. It is not being driven by sentiment alone. The earnings are backing it up.
The Q1 Numbers Are Not Normal
SK Hynix reported Q1 2026 results on April 23. Revenue came in at 52.58 trillion won, approximately $35.5 billion, nearly triple the figure from the same quarter a year earlier, per CNBC’s reporting on the release. Operating profit reached 37.61 trillion won, producing an operating margin of 72 percent. That margin figure is the one that stopped analysts in their tracks. Nvidia, the company that has arguably driven more of the current AI investment cycle than any other single business, reported a 65 percent operating margin in its most recent comparable quarter. SK Hynix just beat it. Counterpoint Research analyst MS Hwang told CNBC that the results show “a lot more memory is needed for AI inference than expected, with companies rushing to secure supply.”
The full-year picture for 2025 already set records across every metric: revenue of 97.1 trillion won, operating profit of 47.2 trillion won at a 49 percent margin, net profit of 42.9 trillion won. All three were historical highs. The 2026 trajectory is running well ahead of that pace. Goldman Sachs has raised its 2026 DRAM supply-demand gap forecast from 3.3 percent to 4.9 percent, calling it the most severe shortage in fifteen years, per TradingKey analysis published in May. Multiple brokerages are projecting SK Hynix full-year 2026 operating profit at between $140 billion and $150 billion on an annualised basis.
Why HBM Changes Everything
The product driving all of this is High Bandwidth Memory, or HBM, the specialised chip architecture that sits inside AI accelerators and allows the massive parallel data transfers that large language models and inference workloads require. Standard DRAM cannot do what HBM does at AI scale. The bandwidth requirements of modern AI systems have created a structural demand for a product that only a small number of companies in the world can manufacture at the required quality and volume.
SK Hynix is the dominant player. The company holds between 59 and 62 percent of the global HBM market depending on the quarter and metric, per Counterpoint Research data running through Q2-Q3 2025. Nvidia accounts for approximately 90 percent of SK Hynix’s HBM supply, making the two companies deeply interdependent in a way that gives SK Hynix pricing power that traditional memory manufacturers have almost never enjoyed. The entire HBM production capacity across SK Hynix, Samsung and Micron is sold out for 2026. SK Hynix has stated publicly that DRAM, NAND and HBM are completely sold out, making it impossible to satisfy all customer orders. The company is not turning away demand because it lacks the technology. It is turning away demand because it cannot build the product fast enough.
SK Group Chairman Chey Tae-won stated in March 2026 that the global chip wafer shortage is likely to persist until 2030, as demand for HBM continues to outpace supply and strain manufacturing capacity, and that expanding wafer capacity could take at least four to five years, with a projected shortfall exceeding 20 percent. Goldman Sachs is slightly more conservative, flagging undersupply lasting at least until the first half of 2027. Either timeline means the current supply dynamic is structural, not cyclical. The parallel with copper, where a multi-year supply constraint driven by structural underinvestment in capacity has kept prices elevated regardless of short-term demand swings, is worth keeping in mind. When the supply side cannot respond to demand within the typical cycle timeframe, the pricing environment is fundamentally different.
Microsoft, Google and Amazon Are Offering to Pay for the Factory
This is the detail that tells you how serious the supply constraint actually is. In early May 2026, multiple media outlets reported that Alphabet, Meta and Microsoft are among the companies planning to propose investment plans to SK Hynix, offering to fund capacity expansion in exchange for guaranteed HBM capacity allocations in coming years. The hyperscalers are not waiting for the market to sort this out. Alphabet, Meta and Microsoft are moving to secure supply by effectively pre-financing the infrastructure that will produce it. That is an unusual arrangement in the semiconductor industry and it reflects the degree to which AI infrastructure investment has become an existential priority for the largest technology companies in the world.
The Big Four tech giants, including Microsoft and Meta, are expected to see combined cloud computing capital expenditures exceed $725 billion in 2026, up approximately 40 percent year-on-year, per data cited by TradingKey. Bank of America has described 2026 as a “supercycle similar to the boom of the 1990s,” forecasting global DRAM revenue to surge 51 percent and NAND revenue to rise 45 percent year-on-year, with average selling prices rising 33 percent and 26 percent respectively. BofA has named SK Hynix as the memory industry’s top pick and a primary beneficiary of the supercycle. UBS predicts SK Hynix will achieve approximately a 70 percent share of the HBM4 market for Nvidia’s next-generation Rubin platform in 2026, suggesting its current technology leadership is carrying over to the next product generation.
On April 22, SK Hynix announced it would invest 19 trillion won, approximately $12.85 billion, to build a new advanced packaging facility in South Korea focused on HBM production. Construction is set to begin within weeks. The company also announced in April that mass production of its SOCAMM2 memory modules, designed specifically for Nvidia’s next-generation compute platform, had begun ahead of schedule.
The Valuation Question Nobody Agrees On
Here is where it gets genuinely interesting from an investment standpoint. SK Hynix currently trades at roughly 8.4 times its annualised Q1 2026 operating profit, per TradingKey’s analysis. Using full-year 2026 broker consensus estimates, the forward P/E ratio is approximately 6.5 times, placing it at the midpoint of the historical semiconductor cycle rather than at any kind of extreme. The question the market is actively debating is whether to apply traditional cyclical semiconductor valuation multiples, which would suggest limited upside from here, or AI growth stock multiples, which would imply a target market cap of $1.05 trillion to $1.58 trillion depending on the P/E applied.
The bear case is real and worth stating clearly. Samsung is accelerating its own HBM4 development, targeting mass production in Q4 2026. If Samsung closes the quality gap on next-generation HBM, SK Hynix’s pricing power narrows. Memory has historically been one of the most cyclical industries in technology, with boom periods followed by severe price corrections. And the current shortage is partly a function of AI infrastructure spending being pulled forward by hyperscalers trying to establish compute advantages. If that spending moderates, the demand picture changes faster than the supply side can adjust.
The broader Asian equity rally, which has already taken the Nikkei to record territory and lifted the KOSPI by an eye-watering 88 percent year-to-date, provides the macro backdrop for SK Hynix’s move, but it also raises the question of how much of the re-rating is fundamental and how much is momentum-driven capital flowing into anything with an AI label in the region. The KOSPI’s 88 percent gain in 2026 is extraordinary by any historical standard and warrants scrutiny rather than extrapolation.
The Samsung Strike Risk Sitting Underneath All of This
There is a risk that is not getting enough attention in the coverage of South Korea’s semiconductor rally. More than 41,000 Samsung workers are threatening an 18-day strike beginning May 21 if their demands are not met, per CNBC’s reporting from the May 14 Asia session. South Korea’s finance minister Koo Yun-cheol warned that a potential strike could pose a major threat to the country’s economic growth, exports and financial markets. Samsung is simultaneously SK Hynix’s primary competitor in HBM and the anchor of the KOSPI’s broader 2026 rally. A prolonged strike at Samsung does not directly affect SK Hynix’s production, but it would affect the competitive timeline for HBM4 and potentially benefit SK Hynix’s market share position if Samsung’s ramp is delayed. It would also hit Samsung’s own earnings trajectory at a moment when the stock is priced for continued AI execution.
The China dimension adds another layer to the Korea semiconductor story. Trump’s state visit to Beijing, which dominated Asia market headlines on May 14 with Xi telling Trump that trade talks were “making progress” while warning that disagreement over Taiwan could send relations “down a dangerous path,” keeps the export control question alive. Goldman Sachs analysts said they expected the summit to focus narrowly on tariffs and export controls, including restrictions on rare earths and semiconductors. Any escalation in semiconductor export controls targeting South Korean chipmakers or their US customers would be a material negative for SK Hynix’s growth trajectory, given the depth of its relationship with Nvidia and the US hyperscalers.
What the Trillion-Dollar Threshold Actually Means
If SK Hynix crosses $1 trillion in market capitalisation, the headline is about South Korea joining a club previously reserved for US companies. But the more substantive point is what the valuation says about where value is being created in the current technology cycle. Institutional capital is moving fast and with conviction into the infrastructure layer of AI, not just the application layer. SK Hynix is infrastructure. The chips it makes are not optional components. They are the physical constraint on how fast AI systems can run, how many tokens can be processed per second, and how much compute the hyperscalers can deploy per dollar of capital expenditure.
The market is pricing in a world where that constraint persists for years and where the company sitting on the dominant position in the most constrained part of the supply chain is worth something closer to a software-style multiple than a commodity semiconductor multiple. Whether that re-rating holds through the Samsung HBM4 ramp, the eventual supply response, and whatever the Trump-Xi summit produces on export controls is the question that will define the SK Hynix trade over the next twelve months. The Q1 numbers give the bulls everything they need to make the case. The competitive and geopolitical risks give the bears enough to keep the argument alive.
Either way, going from $100 billion to $1 trillion in twelve months is a statement about the moment we are in, not just about a single company.
