SK Hynix forecast to post record operating profit of over W64tr in Q2, topping W100tr for first half

SK Hynix is expected to set another all-time record for operating profit in the second quarter, driven by surging memory chip prices. The chipmaker is forecast to post operating profit exceeding 64 trillion won ($43.6 billion), with an operating margin topping 75 percent.

SK Hynix returned from an AI Summit held recently in San Francisco having confirmed strong demand from major US technology companies. The company signed a comprehensive AI infrastructure cooperation agreement with Nvidia worth more than $500 billion, and also reached a long-term memory supply partnership with Microsoft.

According to industry sources, SK Hynix is scheduled to release its second-quarter earnings on Wednesday. The results are expected to extend the company's streak of record-breaking quarterly sales and operating profit.

According to FnGuide, second-quarter sales are projected at 84.17 trillion won and operating profit at 64.22 trillion won. That would represent increases of roughly 60 percent in sales and 70 percent in operating profit quarter-on-quarter, compared with the previous records set in the first quarter — 52.58 trillion won in sales and 37.61 trillion won in operating profit.

If results come in line with the securities industry consensus, SK Hynix would surpass 100 trillion won in operating profit in just the first half of the year — more than double the 47 trillion won it earned for all of last year. The company's quarterly operating profit stood at around 7 trillion won in the first quarter of last year, crossed 10 trillion won to reach 11 trillion won in the third quarter, then surged into the tens of trillions of won in this year's first quarter. The operating margin, which ran at 40 to 50 percent on a quarterly basis last year, has exceeded 70 percent since the first quarter of this year.

Behind the record-breaking results is, by far, the memory demand from US big-tech companies. Their appetite for memory is driving earnings gains at the world's two leading memory chipmakers. Samsung Electronics, which released its second-quarter preliminary results earlier this month, also set a new record, reporting provisional operating profit of 89.4 trillion won — up 56 percent from 57.23 trillion won in the first quarter.

SK Hynix's dominance in the HBM market, where the chips play a central role in AI accelerators, has also underpinned the strong results. According to Counterpoint Research, SK Hynix held a 58 percent revenue share in the first quarter, maintaining its No. 1 position and far outpacing Samsung Electronics at 21 percent.

Standard DRAM prices are also expected to have risen sharply in the second quarter, sustained by continued demand centered on AI data centers. Market research firm TrendForce estimated that contract prices for standard DRAM climbed 58 to 63 percent quarter-on-quarter in the second quarter.

Some in the securities industry project operating profit could fall slightly short of the consensus, landing in the low 60 trillion won range. Even so, analysts broadly agree that long-term stability is not in question. The growing share of long-term supply agreements, or LTAs, with cloud service providers has strengthened the earnings base.

"As the number of customers seeking five-year long-term contracts increases, the average selling price for cloud service providers is likely to come in below market research estimates," said Noh Geun-chang, head of the research center at Hyundai Motor Securities. "The long-term contracts currently being signed are priced above current market levels, and the strengthening of performance-related clauses also improves earnings visibility."

Remarks by SK Group Chairman Chey Tae-won at the AI Summit held in San Francisco on Friday (local time) lend further weight to that market analysis. Demand coming from big-tech CEOs, he said, has far exceeded even his own expectations.

Nvidia is the prime example. SK Group reached a comprehensive agreement with Nvidia at the AI Summit to build AI infrastructure worth more than $500 billion over five years. "I got a rough sense of the contract size Nvidia wants over five years," Chey said. "I'm afraid that the next time we meet, Jensen Huang will tell me, 'That number was wrong — we want more chips.'"

Other big-tech executives Chey met with also pressed him for more semiconductor supply. "I had dinner with Broadcom CEO Hock Tan two weeks ago," he said. "Broadcom also asked for a volume of memory that would shock you if you heard it." He added that OpenAI CEO Sam Altman told him at the meeting, "Don't give the memory to anyone else — give it directly to us," calling it something he had never heard before.

SK Hynix also agreed to pursue a long-term memory supply partnership with Microsoft, with plans to supply server memory optimized for AI workloads to Microsoft's data centers over the medium to long term.

"With the growing share of LTAs, the proportion of sales directed at big-tech and AI data centers is expected to expand to 70 percent next year," said Kim Dong-won, head of research at KB Securities. "The B2B sales share, which stood at 30 percent in 2017, is expected to reach 70 percent within a decade."