The ascent is rooted in a long, risky pivot away from conventional memory chips toward a then-niche technology that would later become essential to the artificial intelligence boom.
A risky acquisition once dismissed as reckless
Back in 2012, SK Group’s acquisition of Hynix Semiconductor was widely seen as a gamble bordering on recklessness. At the time, Samsung Electronics dominated the global memory market, valued at more than ten times its smaller rival and firmly established as the leader in DRAM—critical components powering everything from smartphones to laptops.
SK Hynix, by contrast, was still rebuilding its identity after years of instability and financial distress. The memory market itself was also brutally cyclical, marked by price crashes and overcapacity that had already pushed the company to the edge of collapse more than once.
Even credit agencies were uneasy. Standard & Poor’s issued warnings about SK Group’s exposure to a capital-intensive and volatile industry, questioning whether the acquisition could pay off at all.
The bold bet on an unproven technology
Rather than compete head-on with Samsung in mainstream DRAM, SK Hynix sought differentiation through a less-established segment: high-bandwidth memory (HBM). At the time, HBM was considered a niche product—technically promising, but far from mainstream adoption in data centers or consumer devices.
The company took an early leap. In 2014, it delivered the world’s first HBM product in collaboration with Advanced Micro Devices. However, progress was uneven. The second generation stumbled, and by the late 2010s SK Hynix had fallen behind Samsung in development.
Inside the company, the setback triggered serious internal debate over whether to abandon HBM entirely.
“We believed that it would be impossible to overcome Samsung in commodity DRAM products,” said Hyun Sun-yeop, a former SK Hynix HR executive. “We were desperate to change the market dynamics. We needed a breakthrough.”
The decision that changed everything
Instead of retreating, SK Hynix doubled down. Executives chose to continue investing heavily in HBM, betting that data-heavy computing—and eventually artificial intelligence—would require a new kind of memory architecture.
That conviction led to massive capital spending, including an 880 billion won ($640 million) investment in a packaging facility in Icheon and other infrastructure.
Shim Dae-yong, who led HBM development at the time, recalled the uncertainty that followed.
“It was a headache back in 2019,” he said. “It was obsolete.”
The timing looked disastrous. Demand from Nvidia and cryptocurrency miners dropped sharply, leaving parts of the new capacity underutilized and raising questions about whether the investment had been a costly miscalculation.
From underused factories to AI-critical infrastructure
Everything changed with the global artificial intelligence surge triggered by OpenAI’s ChatGPT in 2022. The rapid expansion of generative AI systems created explosive demand for high-performance computing hardware—particularly Nvidia’s AI accelerators, which rely heavily on HBM to process vast amounts of data.
Suddenly, SK Hynix found itself in a position few competitors anticipated: the central supplier of a technology that had become indispensable.
“No one expected the HBM market would post such explosive growth,” Shim said. “But we were ready in terms of performance and capacity.”
Today, SK Hynix is widely regarded as Nvidia’s primary HBM supplier, placing it at the heart of the AI infrastructure boom powering data centers worldwide.
Boom, bust, and a historic turnaround
The company’s rise did not come in a straight line. After surviving earlier near-bankruptcies in the early 2000s—when creditor banks had even explored selling it to Micron Technology—the firm continued to oscillate with the memory market’s violent cycles.
A major downturn in 2023 underscored that volatility, pushing SK Hynix into an operating loss of 7.73 trillion won. Yet the rebound was just as dramatic: by 2024, it posted record operating profits, fueled by soaring demand for AI-related chips.
By 2025, it had briefly overtaken Samsung as the world’s leading DRAM maker, a symbolic reversal of roles in a market Samsung had long dominated.
“No one would ever have imagined that SK Hynix would overtake Samsung,” said Shin Jae-yong, a professor of business administration at Seoul National University. “It is almost impossible for a runner-up to catch up with the market leader in this capital-intensive industry.”
A shifting rivalry at the top
Samsung is now working to regain momentum, leveraging its in-house foundry capabilities in the broader race for HBM competitiveness. SK Hynix, meanwhile, relies on Taiwan Semiconductor Manufacturing Company for key production steps, highlighting the increasingly global and fragmented nature of chip supply chains.
The rivalry has become a defining feature of the AI hardware era, with both firms racing to secure dominance in a market reshaped by machine learning demand.
Economic ripple effects and global ambitions
SK Hynix’s rise has extended beyond corporate rankings. Its performance has helped energize South Korea’s stock market and boosted national investor sentiment. The company’s employees have even become cultural symbols of success in South Korea’s job market.
At the top of the company, expectations have also escalated. SK Group Chairman Chey Tae-won previously suggested ambitions of reaching a 1 quadrillion won market capitalization—and eventually doubling that target.
“What I really wanted to accomplish when we acquired Hynix was to transform it from a commodity memory producer into a mainstream semiconductor company whose products are indispensable,” Chey said in remarks published in a book released in January.
That vision now appears closer to reality than ever. On Monday, SK Hynix reached a market value of nearly 2.1 quadrillion won, briefly making it South Korea’s most valuable listed company.
In Chey’s words, it marks a transformation driven by ambition, endurance, and timing—but also by a wager that only fully paid off when the world itself changed.