By NUPIAO
Samsung Electronics just dropped its Q2 2026 earnings guidance and the numbers are nothing short of staggering. The company expects sales to hit roughly 171 trillion Korean won (about $112.34 billion), a massive 129% leap from 74.57 trillion won a year ago. Its operating profit? A jaw-dropping 89.4 trillion won (around $58.8 billion), up from a measly 4.68 trillion won last year—that’s an 18-fold explosion.
Samsung’s forecast even blew past the already rosy consensus. The SmartEstimate model from LSEG Data & Analytics, which aggregates predictions from 30 analysts, had pegged operating profit at 86 trillion won, which would have been the third consecutive record-breaking quarter. Nope, Samsung went even higher.
Just days earlier, on July 3, a top executive in Samsung’s semiconductor business strategy had dropped a bold hint, saying “this year’s operating profit will exceed the cumulative profit of the entire 40-year history of our chip business.”
But here’s the plot twist: That blowout guidance triggered a classic “buy the rumor, sell the fact” meltdown. When the Korean market opened, Samsung’s stock nosedived, at one point crashing 10%.
It wasn’t just Samsung. SK Hynix, its archrival and the leader in high-bandwidth memory (HBM), also got slammed, plunging as much as 8% and briefly dropping over 10%.
All that carnage dragged the KOSPI index down by 8%, triggering a circuit breaker that halted trading for 20 minutes.
Japan’s chip sector felt the pain too. Kioxia, the major memory maker, tanked 10.86%, while testing giant Advantest and Tokyo Electron slipped 0.64% and 1.85%, respectively, adding pressure to the Nikkei.
Analysts pointed to overly bullish profit expectations. Even after setting aside a massive bonus pool for semiconductor workers tied to a May wage deal that links pay to operating profit, Samsung still delivered a monster bottom line. Some analysts believe that without those one-time provisions, operating profit could have easily topped 100 trillion won.
But Albert Yong, managing partner at Petra Capital Management, captured the real fear: “Samsung’s blockbuster earnings were already baked into the stock price after a long rally. Now investors are looking further out. What really scares them is whether this AI frenzy can last, and whether U.S. tech giants will start to pull back on their AI infrastructure spending.”
Right now, the memory chip market is still super tight, but the big question is sustainability. The share of cloud capex pouring into AI memory chips is expected to hit 52% this year—and could top 70% next year. That’s extreme. If AI services don’t start generating serious cash fast enough, the hardware-driven profit party could face a rude awakening.
We are in the golden age of memory chips. The insatiable AI demand for HBM is swallowing up wafer capacity and squeezing out traditional DRAM and NAND production. Chip giants are prioritizing high-margin data center orders, and that’s triggering a global shortage of standard memory chips.
Citigroup’s research shows that in Q2, average selling prices for DRAM and NAND jumped 44% and 53% quarter-over-quarter, respectively. And according to Counterpoint, this “shortage premium” handed Samsung, SK Hynix, and Micron crazy pricing power, pushing their average operating margins to an almost unbelievable 75% to 80%.
But there’s always a flip side. The upstream memory chip bonanza is squeezing downstream consumer electronics.
Samsung, the world’s biggest smartphone maker, is feeling the heat from its own chip price hikes. Though both Samsung and Apple have already raised some hardware prices, component costs are rising much faster. Analysts expect Samsung could be forced to slap another round of price increases on its phones later this year.
It’s worth noting that Samsung and SK Hynix recently pledged to invest a staggering 3,200 trillion won (about $2.07 trillion) to massively expand chip production in South Korea. Samsung’s portion is set to roll out between 2026 and 2040.
Memory chips are famously cyclical. When prices are soaring, companies rush to build fabs—but new plants take 2.5 to 3 years to reach full capacity. The wave of capex being unleashed right now will create a flood of new supply around 2028 to 2029. If AI cloud capex growth slows even a little, the supply-demand balance could flip fast, and the pressure on future prices can’t be ignored.
Still, some remain bullish. Bank of America analysts led by Simon Woo wrote in their latest global memory weekly that the Korean fab buildout won’t kill the memory super-cycle. In their view, it will take at least 8 to 10 years before the new capacity can meaningfully affect global supply.
