Reports of an average bonus of 6.1 million RMB per employee at a South Korean chipmaker have recently circulated on social media. The company is SK Hynix.
The figure of 6.1 million RMB originates from a calculation by Macquarie Securities, an international investment bank, based on a 2027 profit forecast. The projection assumes SK Hynix achieves an operating profit of 447 trillion won in 2027. According to a labor-management agreement stipulating that 10% of operating profit is allocated for employee bonuses, this would translate to approximately 12.9 billion won per person, or nearly 6.1 million RMB. SK Hynix has responded cautiously, stating that "performance is not yet confirmed, and the bonus scale cannot be predicted."
However, the already distributed figures are also staggering. At the beginning of 2026, SK Hynix paid out performance bonuses for 2025, totaling 4.72 trillion won, averaging about 650,000 RMB per employee, benefiting all regular staff.
Following this news, a humorous saying emerged on South Korean social media: SK Hynix employee uniforms have become a "dating charm."
In reality, wearing the same uniform three years ago would likely have had the opposite effect. The same company was experiencing its largest historical loss just three years prior.
In 2023, SK Hynix recorded a loss of approximately 7.7 trillion won, its largest annual loss ever.
In 2025, operating profit reached 47.2 trillion won, doubling year-over-year and setting a new record, with an operating profit margin of 49%.
In the first quarter of 2026, single-quarter operating profit surged to 37.6 trillion won, a year-over-year increase of 405%, with an operating profit margin of 72%. This performance surpassed TSMC and Micron, making SK Hynix one of the semiconductor industry's most profitable companies.
The same company, the same factories, the same engineers—transformed from massive losses to record profits in just three years.
This exemplifies the semiconductor industry cycle.
The semiconductor industry's cyclical volatility stems from a fundamental structural mismatch.
Demand is short-cycle and elastic—needs for smartphones, computers, cars, and servers change rapidly with economic conditions and technological advancements.
Supply is long-cycle and rigid—building a modern semiconductor fabrication plant (fab) typically takes over three years from decision to mass production.
TSMC's new fab in Arizona, announced in 2021, is not expected to reach large-scale production until 2028 at the earliest.
The lengthy construction and commissioning process for fabs means capacity cannot respond quickly to demand shifts. This mismatch creates a recurring script: market boom → all manufacturers expand simultaneously → concentrated capacity release three years later → price collapse → losses → capital expenditure cuts → shortage returns several years later.
The industry has experienced this cycle over 15 times.
The volatility is exacerbated by the "bullwhip effect" in the supply chain. When chips are scarce, downstream customers over-order and hoard inventory. Once demand slows and they stop purchasing to digest stockpiles, chipmakers experience a demand drop far more severe than the actual decline in the end market. DRAM (Dynamic Random-Access Memory), due to its high commoditization and lack of differentiation, exhibits the most extreme cycles among all semiconductor types.
The 2017-2018 Super Cycle
The period from 2017 to 2018 is widely recognized in the industry as a semiconductor "Super Cycle." The core driver was explosive growth in memory chips (DRAM and NAND Flash). SK Hynix, as the world's second-largest DRAM manufacturer, was one of the primary beneficiaries.
Starting in 2017, demand for memory chips surged. Flagship smartphone DRAM capacities jumped from 4GB to 6GB and 8GB, while NAND storage expanded to 128GB and even 512GB. Concurrently, massive data center expansions by Amazon AWS, Google, Microsoft, and domestic giants Alibaba and Tencent drove exponential growth in demand for high-performance server memory. Furthermore, the 2017 Bitcoin price surge fueled demand for mining rigs, significantly consuming graphics memory (GDDR) capacity.
During this demand upswing, supply was constrained. The industry was undergoing a painful technological transition from 2D NAND to 3D NAND architecture, with slow yield improvements leading to prolonged market shortages and price increases for six consecutive quarters.
Consequently, SK Hynix's performance during 2017-2018 was described as "terrifying." Given its high exposure to memory chips, price increases directly translated into profits. 2018 revenue reached 40 trillion won, a 135% increase from 17 trillion won in 2016. As the profit margin rose from 19% in 2016 to 52% in 2018, total profit for 2018 was 21 trillion won, a sevenfold increase from 3 trillion won in 2016.
The margin expansion was driven by a continuous increase in the average selling price (ASP) of chips for six quarters starting Q4 2016. DRAM prices rose more sharply in this cycle than historically due to new factors.
First, SK Hynix, Samsung, and Micron formed a triopoly. During 2017-2018, these three companies showed relative "restraint" in capacity expansion. This seller's market pricing power generated significant cash flow for SK Hynix's 20nm and 1Xnm DRAM products.
Second, while SK Hynix held a solid position in DRAM, it had lagged in NAND. However, during this cycle, the company successfully developed and mass-produced 72-layer 3D NAND, narrowing the technological gap with Samsung and successfully entering the high-margin enterprise SSD market.
Third, in 2018, a consortium led by Bain Capital, which included SK Hynix, completed the acquisition of Toshiba Memory (now Kioxia). Although SK Hynix acquired only about a 15% stake, this move prevented a competitor from gaining Toshiba's technology and laid important groundwork for SK Hynix's future memory strategy.
Also in 2018, SK Hynix formulated its HBM (High Bandwidth Memory) product development strategy. This product line began to shine brightly from 2025 onwards, helping SK Hynix achieve historically unprecedented high-profit margins.
In simple terms, during past cycles, SK Hynix attempted to find alpha within the industry's beta. It sought stable profits through product R&D and collaborative market structures.
However, alpha is still encompassed within beta. The overarching industry cycle remains the most significant profit driver. In Q4 2018, the boom ended abruptly. Inventory piled up, smartphone sales growth slowed, and data center customer inventory levels were excessively high. As 3D NAND yields stabilized and new fabs from major players came online, the market rapidly shifted from shortage to oversupply. Consequently, memory prices plummeted starting in late 2018, and SK Hynix's performance entered a trough in 2019.
It was from 2018 that SK Hynix began developing HBM. Despite facing short-term setbacks in 2022-2023, the company persisted, and HBM finally achieved remarkable success in 2025, helping the company attain its highest-ever profit margins. HBM is not a standard DRAM commodity. It has extremely high technical barriers; only SK Hynix, Samsung, and Micron can mass-produce it globally. Furthermore, it is deeply integrated with NVIDIA GPUs, and its pricing mechanism resembles long-term contracts more than the spot market. SK Hynix currently holds over a 50% market share in HBM and has already entered the HBM4 mass production stage, demonstrating significant technological leadership.
A tsunami of AI infrastructure investment has stimulated SK Hynix's HBM orders, with production slots reportedly booked through 2027.
The 2026 Cycle Position
Indicator 1: DRAM Price Increase vs. Historical Peak During the previous super cycle (August 2016 to August 2018), the average DRAM price increased by approximately 165% cumulatively. In the current cycle, year-over-year spot price increases for DRAM have generally exceeded 200%, surpassing the涨幅 levels seen at the same stage in the previous cycle. In this涨价 cycle, contract prices have been relatively moderate, while spot prices surged significantly due to inventory and other factors. However, spot prices are now experiencing high volatility, beginning to digest the irrational panic buying from earlier.
Indicator 2: SK Hynix Inventory Levels and Customer Stockpiling Status This is the most direct real-time indicator reflecting current supply-demand tightness. Currently, SK Hynix's HBM capacity for 2026 is fully sold out. Some DDR5 products are being delivered from a "zero inventory" state. Memory demand growth for 2026 is still projected to exceed 20%. The current inventory-to-revenue ratio is below 10%. Clearly, the cycle remains in a prolonged period of peak prosperity.
Indicator 3: SK Hynix Profit Margins vs. Historical Peak SK Hynix's Q1 2026 profit margin data has comprehensively surpassed the levels seen at the peak of the 2018 cycle. Profit margins exceeding historical peaks indicate that the current profitability has entered an unprecedented range. Based on the experience of the past two cycles, when SK Hynix's operating margin begins to decline from its peak, the speed often far exceeds market expectations—from a peak of 51% in 2018 to an average of 12.9% in 2019/2020, a significant correction completed within a year. "Mean reversion" is not inevitable, but it warrants caution.
Indicator 4: SK Hynix Price-to-Book (P/B) Ratio vs. Historical Range SK Hynix's current P/B ratio is 9.6x, the highest point in five years. Just four years ago, its P/B ratio was only 0.8x. The historical median P/B ratio is 1.8x. The market holds extremely high expectations for SK Hynix's future profitability, thus assigning a substantial premium.
Indicator 5: SK Hynix's Market Capitalization Share in the South Korean Stock Market Over the past year, SK Hynix's market capitalization has increased approximately ninefold. Currently at $825 billion (approximately 5.6 trillion RMB), it is the world's 16th largest company by market cap among listed firms. The two major South Korean semiconductor groups, Samsung and SK, accounted for only 22.6% of the South Korean stock market at the beginning of 2025. By May of this year, these two companies comprised 66% of the South Korean stock market. Two-thirds of the market is dominated by just two companies! The third-largest company by market cap in South Korea is Hyundai Motor, with a weight of only 5% in the index!
An interesting phenomenon unique to the South Korean market involves the permitted maximum holding limit for a single stock. South Korean fund regulations stipulate that a fund's holding in a single stock cannot exceed 10%. However, an exemption allows adjustment based on the average weight data for the past three months published by the Korea Financial Investment Association. This association regularly publishes data, and funds use this data as the upper limit for stock purchases. The association releases this data at the beginning of each month. This mechanism led to a peculiar scene in the South Korean market in 2026. When Samsung and SK Hynix stock prices surged mid-month, some funds might not immediately increase their allocations, waiting instead for the adjustment at the beginning of the next month. If investors anticipate that the holding limits for Samsung and SK Hynix will be raised at the start of the next month, retail investors may buy these stocks in advance at the end of the month. Then, at the beginning of the month, when the holding limits are raised, passively following funds are forced to buy. This allows retail investors to effectively "front-run" the institutions. It's not difficult to see that this mechanism can also become an amplifier during a price decline. Once stock prices fall and the permitted holding limits are lowered, any fund that hasn't adjusted in time will be forced to sell when the investment association publishes the new data at the month's start.
Will the 6 Million RMB Average Bonus Be Realized? The 6 million RMB average bonus is predicated on the company's projected profit for 2027. If realized, it would mean the positive stimulus from this semiconductor boom cycle for SK Hynix would continue for another two years. This outcome is viewed with skepticism. The semiconductor cycle has never disappeared; only the driving factors differ each time. This memory super cycle, driven by AI capital expenditure, differs from historical cycles in the authenticity of demand and the concentration of supply. It may last longer than any previous cycle. However, no matter how deep the competitive moat, how领先 the technology, or how significant the alpha, they are ultimately just waves, large or small, in the tide of the era.
Comments