Please Make More: How a Factory Fire, a Wafer Autograph, and a Capacity U-Turn Propelled SK Hynix to $1 Trillion

The moment that captured the sheer desperation of the AI supply chain came not in a boardroom or an earnings call, but on a trade show floor in Taipei. Nvidia CEO Jensen Huang walked up to the SK Hynix booth at COMPUTEX 2026, grabbed a marker, and scrawled “Please Make More” across an exposed HBM4E wafer. It was a public plea from the world’s most powerful chip designer to its most critical memory partner — and a vivid symbol of a market where demand has utterly overwhelmed supply.

That plea came just a day after a very different kind of incident at SK Hynix’s Cheongju campus. On June 1, a fire broke out in a gas room on the sixth floor of Campus 4, an area linking two of the company’s most vital DRAM and NAND fabs, M15 and M15X. Automatic sprinklers extinguished the flames within minutes, but the water triggered a release of hydrogen fluoride gas. All 3,600 workers in the affected facilities were evacuated. Seven people were taken to the on-site clinic: five reported eye irritation, two were examined as a precaution. Production continued without interruption, and analysts saw no risk of supply disruption. The cause of the fire remains under investigation.

If investors had any doubts about the durability of SK Hynix’s momentum, the stock price erased them. The shares closed Tuesday at 2,360,000 won, within a hair of their 52-week high. Year to date, the stock has surged roughly 249 percent, trading about 68 percent above its 50-day moving average. Last week, the company’s market capitalisation crossed the $1 trillion mark for the first time. It was a milestone underpinned by a record first quarter: revenue of 52.58 trillion won — the first time quarterly sales have topped 50 trillion — and an operating margin of 72 percent, another all-time high.

Chairman Chey Tae-won used his COMPUTEX appearance to announce a dramatic strategic reversal. Just three months earlier, at Nvidia’s GTC in March, he had said SK Hynix had no plans to expand overall wafer capacity. Now he declared the company would double its total wafer production capacity over the next five years. “We will double the entire capacity in five years,” Chey told the audience. “There are many obstacles, but we will overcome them.” He did not disclose a specific investment figure, but indicated that 2026 capital expenditure would significantly exceed the roughly $20 billion spent in 2025.

Should investors sell immediately? Or is it worth buying SK Hynix?

The reason for the about-face is starkly simple: the company’s entire HBM production for 2026 is already sold out. Customers, according to people familiar with the matter, have offered to buy EUV lithography scanners and pre-finance new production lines — because there is essentially no available capacity left. Chey expects the structural shortage to persist until 2030. Industry-wide, supply of high-bandwidth memory is running more than 20 percent below demand. Gartner analyst Shrish Pant sees no meaningful price correction before the end of 2027. Greyhound Research has described the situation not as a cyclical swing but as a permanent redistribution of the memory market.

Goldman Sachs recently raised its forecast for SK Hynix’s 2028 operating profit by 24 percent, to the equivalent of roughly $300 billion. The bank’s confidence reflects a competitive landscape where SK Hynix holds a commanding 58 percent share of the HBM market, according to Counterpoint Research’s first-quarter data. Samsung and Micron trail at 21 percent each.

But the Korean memory giant is not content to remain merely the volume leader in HBM. At COMPUTEX, the company unveiled a “Full-Stack AI Memory” strategy aimed at moving deeper into the architecture of AI systems. The vision includes custom HBM solutions (cHBM) tailored to specific AI accelerators, as well as emerging technologies such as HBF — a NAND-based high-speed memory — and 3D Stacked DRAM on Logic, which vertically stacks DRAM atop logic chips to reduce latency, power consumption, and footprint. The company’s booth also featured an “AI Factory” concept, with live demonstrations pairing HBM3E with Nvidia’s GB300, SOCAMM2, and HBM4 alongside a Vera Rubin 200 model. A structural prototype of HBM4E, the seventh-generation HBM generation, was also on display.

SK Hynix at a turning point? This analysis reveals what investors need to know now.

Chey made clear his ambition to secure the same central role for Nvidia’s upcoming Vera Rubin system that SK Hynix holds for the current Blackwell generation. The final decision, he acknowledged, rests with the customer. To improve its chances, the company will need to deepen its partnerships in Taiwan — not only with TSMC but with other local players.

Execution, however, is the watchword. At a $1 trillion valuation and a 249 percent year-to-date gain, expectations have been set sky-high. Samsung used the same COMPUTEX window to make its own HBM announcements, and every major memory maker is charging into the AI infrastructure business. SK Hynix has raised its own bar with a roadmap that promises HBM4E samples in the second half of 2026 and mass production starting in 2027. The next quarterly reports will reveal whether customer qualifications and delivery visibility match the ambition — or whether the market has already priced in more than the factory floors can deliver.

Ad

SK Hynix Stock: New Analysis - 3 June

Fresh SK Hynix information released. What's the impact for investors? Our latest independent report examines recent figures and market trends.

Read our updated SK Hynix analysis...

en | KR7000660001 | PLEASE | boerse | 69475509 |