On July 7, 2026 (Beijing time), Samsung Electronics released its preliminary results for the second quarter of 2026. This early report—pending final audit—sent shockwaves through the entire semiconductor industry and global capital markets.
According to Samsung’s regulatory filings, preliminary operating profit for Q2 reached KRW 89.4 trillion (approximately $58.4 billion), marking a year-on-year increase of about 1,810%. Preliminary consolidated revenue came in at KRW 171 trillion (around $111.8 billion), up 129.3% year-on-year. These initial figures significantly exceeded the average analyst estimate of KRW 84.2 trillion in operating profit. Quarter-over-quarter, operating profit surged 56.2% from Q1’s KRW 57.2 trillion, while revenue jumped 27.7% from KRW 133.9 trillion. Both sales and operating profit have set new all-time highs for three consecutive quarters since Q4 2025.
Samsung clarified that these preliminary results are calculated under Korean International Financial Reporting Standards (K-IFRS), and the final audited results will be announced at the end of July 2026.
Of particular note, these preliminary results already account for funds set aside to fulfill labor agreements and performance bonuses. Samsung previously reached an agreement with its labor union to allocate 10.5% of operating profit as a bonus pool, in addition to existing performance compensation. Because negotiations dragged on, this provision was not included in Q1 2026, so the total bonus allocation for the first half of 2026 is estimated at KRW 19 trillion to KRW 25 trillion. Most industry observers believe the reserve is close to KRW 20 trillion. Factoring this in, Samsung’s actual operating profit generation capacity may have already surpassed KRW 100 trillion.
In US dollar terms, Samsung’s preliminary operating profit this quarter has already exceeded NVIDIA’s $53.54 billion (KRW 82 trillion) for Q1 of fiscal 2027, making Samsung the world’s most profitable company on a quarterly basis. Samsung’s Q2 2026 preliminary operating profit also surpassed its full-year 2025 profit of KRW 43.6 trillion, as well as the combined profits from 2023 to 2025.
However, on the day the preliminary results were released, Samsung’s share price plunged by about 6% to 8% during trading, closing down roughly 5.35% at KRW 301,000. The disconnect between earnings and share price reflects deep market anxiety over the sustainability of the AI semiconductor boom—how much longer can this AI-driven memory chip supercycle last?
HBM Demand Surge: The Core Driver Behind Preliminary Results
Breaking down Samsung’s KRW 89.4 trillion in quarterly preliminary profit, the Device Solutions (DS) semiconductor division stands out as the main engine. Analysts estimate that the DS division generated around KRW 80 trillion in operating profit this quarter, accounting for the vast majority of group earnings.
The explosive profit growth is primarily driven by a sharp increase in memory chip prices. According to market research data, Samsung led the global DRAM market in Q1 2026 with a 38.5% revenue share. DRAM and NAND prices continued to climb in Q2, with DRAM contract prices rising over 40% quarter-on-quarter and NAND prices up more than 50%. In Q1 2026, Samsung’s DRAM revenue soared 93.4% sequentially to $37.32 billion.
Behind the price surge is a structural supply-demand imbalance. Soaring demand for HBM (High Bandwidth Memory) from AI data centers has prompted major memory manufacturers like Samsung and SK Hynix to prioritize their 12-inch wafer capacity for HBM production. This shift toward high-end products has led to severe shortages in general-purpose memory chips such as DDR4 and DDR5, driving up prices across the board. Analysts expect this shortage to persist at least through 2027.
Samsung is also accelerating its push in the HBM segment. In February 2026, Samsung became the first in the world to mass-produce sixth-generation HBM, or HBM4. In just four months, related sales surpassed $1 billion (about KRW 1.5 trillion). Bernstein analysts estimate Samsung’s Q2 HBM revenue at around $4.7 billion, up 32% quarter-on-quarter. They note that Samsung has allocated about half of its monthly HBM DRAM wafer output—roughly 150,000 wafers—to HBM4, with the remainder assigned to 12-layer HBM3E. HBM4 certification is advancing rapidly, with mass production expected to begin after completion in Q2. This strategic capacity shift underscores Samsung’s ongoing bet on high-end HBM demand.
Profit Bonanza vs. Stock Plunge: What’s Worrying the Market?
On the day of the earnings release, Samsung’s share price dropped as much as 6% to 8%, dragging South Korea’s KOSPI Index down over 4% and triggering a circuit breaker. SK Hynix shares also fell 2.6% to 5.5% the same day.
This "sell the news" reaction reveals the market’s complex view of the AI semiconductor cycle. With a historic wave of AI infrastructure buildout underway globally, investors had already set high expectations for chipmakers like Samsung and SK Hynix. Samsung’s stellar earnings were no secret, and its share price had already rallied significantly. When the results came in, it became a window for profit-taking.
Deeper concerns center on the cycle’s timing. Just last week, Meta signaled caution in its future capital spending plans, sparking fears that tech giants may cut back. High-beta momentum stocks suffered their largest two-day drop since the pandemic. Brian Cho, portfolio manager at global asset manager Causeway Capital Management, noted that the market now cares more about whether companies can deliver sustainable, step-change growth in annual free cash flow, with shareholder return policies also in focus.
Additionally, Samsung’s stock performance has lagged its domestic rival SK Hynix. SK Hynix’s business is more focused on high-end memory for AI workloads, and its shares have surged since 2026. Competition-wise, Samsung also faces emerging rivals like ChangXin Memory Technologies, which are putting pressure on market share.
Three-Pronged Analysis: How Long Can the AI Semiconductor Boom Last?
To answer the key question—how much longer can the AI semiconductor upcycle last?—we must examine demand, supply, and capacity expansion.
Demand Side: Capex Still Expanding, Underlying Demand Keeps Growing
A Nomura Securities report dated July 1, 2026, warned that the AI semiconductor cycle is far from peaking, with a potential "epic" supply chain mismatch looming in H2 2026. The firm tracked around 280 new global data center projects as of late March 2026, up from about 240, with "gigawatt-scale" projects rising from over 40 to about 50. This suggests global new data center deployment capacity will rise from 26 GW in 2026 to 32 GW in 2027, translating to annual demand for roughly 4 million to 6 million AI chips in 2027–2028.
Hyperscale cloud providers—Microsoft, Google, Meta, Amazon—are still expanding their global data center footprints. Meanwhile, AI companies like OpenAI, xAI, and CoreWeave, along with "new cloud" operators, are becoming major sources of incremental demand. Meta and CoreWeave, for example, signed a long-term AI cloud capacity deal worth $21 billion, extending through 2032.
J.P. Morgan’s July 2026 report further noted that the chip upcycle will likely last at least through 2028. Multiple memory manufacturers have sold out their HBM supply through 2026, and new wafer capacity isn’t expected to come online in a meaningful way until after 2028.
Supply Side: Real Bottlenecks in Capacity Expansion
While demand keeps expanding, the supply side faces real constraints. Building a new wafer fab typically takes around two years. After cloud providers sharply raised AI chip and server procurement forecasts in late 2025, the supply chain only then began accelerating capacity expansion—but much of this new capacity won’t be fully available before 2027.
TSMC’s next major capacity ramp-up for advanced nodes won’t be significant until after 2028. More critically, supply bottlenecks are spreading from TSMC-controlled CoWoS advanced packaging to other components—IC substrates, PCBs, copper-clad laminates, high-end capacitors, power management chips, optical components, and thermal/power delivery systems—all could become new chokepoints for AI server shipments.
Most analysts believe memory market shortages caused by cleanroom constraints will persist through the end of 2027. According to Samsung Securities researcher Lee Jong-wook, the memory market is still in the "early phase" of the cycle. Samsung Securities forecasts that tight DRAM and NAND supply will last through 2026, with quarterly operating profit potentially exceeding KRW 100 trillion for over a year.
Capacity Expansion: Balancing Long-Term Confidence and Short-Term Risks
These supply bottlenecks explain why leading manufacturers are investing heavily in capacity. Samsung has announced over $70 billion in 2026 spending plans for capacity expansion and R&D. Samsung Group and SK Group plan to each build two new chip fabs in southwest Korea, with a combined investment of KRW 800 trillion, aiming to double Korea’s memory chip capacity within five years. Samsung Securities describes Samsung’s aggressive memory investment and early ramp-up at the Pyeongtaek P5 plant as a show of confidence in long-term demand.
TSMC is expected to raise its 2026 capex to $52–56 billion, with 5–10% price hikes for advanced process nodes. TSMC projects nearly 30% revenue growth in 2026 (in USD), fueled by surging compute demand as AI technology evolves from generative to agentic models.
However, large-scale expansion also brings cyclical risks. Market research firm Counterpoint warns that such high profit margins could be seen as "excessive," potentially inviting regulatory scrutiny if they persist. UBS Asset Management advised selective AI investments on July 6, 2026, to navigate volatility, noting that the market remains highly volatile as it weighs strong AI adoption against slowing capex growth.
In summary, the current AI semiconductor cycle is underpinned by robust, expanding demand, faces real and persistent supply bottlenecks, and must balance long-term capacity growth with short-term risks. The core tension—demand growth outpacing supply expansion—is unlikely to fundamentally reverse before 2027.
Conclusion
Samsung Electronics’ preliminary Q2 2026 operating profit of KRW 89.4 trillion is a clear reflection of surging AI-driven memory demand and direct evidence of the current semiconductor supercycle’s strength. It’s important to note that these results are preliminary estimates; the final audited report will be released at the end of July 2026. A single quarter’s preliminary profit exceeding the past three years combined—and surpassing NVIDIA to become the world’s most profitable company in a quarter—provides compelling evidence against the notion of an "AI investment bubble." The preliminary results also include around KRW 20 trillion in performance bonus reserves; excluding this, Samsung’s true profit-generating power is even more impressive.
However, the market’s response—as seen in the share price—warrants caution. When a company posts quarterly profits of KRW 89.4 trillion, investors naturally ask: how long can this growth trajectory last? Institutions like Nomura and J.P. Morgan see the cycle’s peak extending into 2027–2028, based on ongoing demand expansion and persistent supply bottlenecks. But as UBS and others warn, high volatility and the risk of overcapacity from large-scale expansion remain key variables.
TrendForce’s latest forecast shows that the overall DRAM market will remain extremely tight in Q3 2026, though contract price increases are moderating, with quarterly gains expected at 13% to 18%. NAND Flash demand continues to be driven by AI inference and large-scale data center buildouts, with contract prices expected to rise 10% to 15% quarter-on-quarter. The uptrend continues, but at a slower pace—a hallmark of the transition from the "boom phase" to the "validation phase" of the cycle.
For investors, the core message from Samsung’s preliminary results is this: the AI semiconductor upcycle is far from over, but the market has shifted from "expectation" to "validation" in its pricing. The key question going forward is no longer "Is there AI demand?" but rather "Can AI demand growth continue to outpace supply expansion?"—and the answer will become clear over the next 12 to 24 months.
FAQ
Q: Is Samsung Electronics’ nearly 19-fold profit increase in Q2 2026 a final, confirmed figure?
A: No. This figure is a preliminary estimate announced by Samsung Electronics on July 7, 2026, calculated under Korean International Financial Reporting Standards (K-IFRS). The final audited results will be released at the end of July 2026. The currently announced KRW 89.4 trillion operating profit and KRW 171 trillion revenue are both preliminary.
Q: What does it mean that Samsung’s quarterly profit surpassed NVIDIA’s?
A: Samsung’s preliminary operating profit for the quarter was KRW 89.4 trillion (about $58.4 billion), exceeding NVIDIA’s $53.54 billion last quarter. This highlights the growing profit share of memory manufacturers within the AI value chain, as the benefits of AI infrastructure buildout extend from compute chips to the entire semiconductor supply chain.
Q: How much longer is the AI semiconductor upcycle expected to last?
A: Institutions like Nomura and J.P. Morgan expect the upcycle to continue at least through 2027–2028. Key factors include the ongoing increase in global data center projects, upward revisions in cloud provider capex, tight HBM supply, and the fact that new wafer capacity won’t come online in a meaningful way until after 2028. Samsung Securities forecasts tight memory supply through 2026.
Q: With such strong results, why did Samsung’s share price fall sharply?
A: The market had already priced in strong earnings, leading to a classic "sell the news" scenario. Additionally, concerns about the sustainability of AI investment, potential overcapacity, possible capex cuts by tech giants, and Samsung’s catch-up pressure in the HBM segment all weighed on the stock.
Q: Has the memory chip price uptrend peaked?
A: Not yet, but the rate of increase is slowing. TrendForce data shows that DRAM contract prices are expected to rise 13% to 18% quarter-on-quarter in Q3 2026, with NAND up 10% to 15%. While the pace of increases has narrowed compared to previous quarters, the uptrend continues.
