AI Chip Demand Is Set to Double
— and the Entire Industry Is Already Sold Out

Opening the Floodgates

On July 20, 2026, SK Group Chairman Chey Tae-won stood before the Korea Chamber of Commerce and Industry’s Jeju Forum and delivered a warning that should have been front-page news everywhere.

AI semiconductor demand next year will grow by 60 to 100 percent, he said. Total memory semiconductor demand will rise by at least 50 to 60 percent. Supply is not keeping pace. “No company has meaningful new capacity coming online next year,” Chey said.

The global scramble for memory supply has reached what he called “panic” level.

But Chey’s warning was not simply about scarcity. He explicitly stated that current memory prices are “abnormal” and not a healthy market signal. He warned that PC and phone manufacturers cannot indefinitely pass storage cost increases on to consumers, and that if prices remain too high, new competitors will enter the market and governments may intervene.

“Prices must come down,” Chey said. “Current prices are not normal.”

This is not a short-term cycle. This is a structural transformation of the global semiconductor industry — and it is only just beginning.

The Numbers Behind the Panic

The demand numbers are significant. Customers have asked SK Hynix for 60 to 100 percent more AI memory in 2027 than this year. Because AI now accounts for more than half of overall semiconductor consumption, Chey put total market demand growth at a minimum of 50 to 60 percent — a figure that applies to the entire memory industry, not just the AI segment.

Gartner forecasts that worldwide semiconductor revenue will grow 64 percent in 2026, exceeding $1.3 trillion for the first time. Memory revenue alone is expected to triple amid what analysts are calling “memflation.” AI semiconductors are expected to account for approximately 30 percent of total semiconductor revenue in 2026 and will remain the driving force behind overall industry growth.

Morgan Stanley has been even more aggressive, predicting the memory market will surge from roughly $220 billion in 2025 to $890 billion in 2026. Within just three months, the firm raised its 2026 DRAM price forecast by 71 percent — an adjustment that reflects how rapidly the market is tightening.

On the supply side, the picture is grim. Chey noted that almost no manufacturer has substantive plans to expand capacity next year. Advanced AI memory capacity expansion requires massive capital investment and long lead times, while facing constraints in equipment and personnel. “We are doing our best to maximize supply, but demand is growing much faster than we can keep up,” Chey said.

The result: SK Hynix’s entire 2026 output of HBM, DRAM, and NAND — every single chip the company can produce — is already sold out, with the crunch expected to extend into 2027.

Who Controls the Market

The memory market is dominated by three players, and the concentration is extreme.

In the first quarter of 2026, SK Hynix held 58 percent of the global HBM market by revenue, according to Counterpoint Research. Samsung Electronics and Micron each held 21 percent, tied for second.

In the broader DRAM market, Samsung led with 38 percent, followed by SK Hynix at 29 percent and Micron at 22 percent. Chinese manufacturer ChangXin Memory Technologies (CXMT) has increased its DRAM market share to 8 percent — a sign that China is slowly building a presence in a market long dominated by Korean and American players. The need to make the U.S. chip supply chain “more robust” has become a central theme in trade negotiations, with Commerce Secretary Howard Lutnick applying pressure on Korean chipmakers to invest more heavily in American manufacturing.

Chey was dismissive when asked about Lutnick’s pressure. Lutnick “has been this way” since the early Trump administration, Chey said. “The question is always the same: how much will you invest in the U.S., how much will you reshore. This time isn’t different.”

But the pressure is real — and it’s coming from multiple directions. Chey warned that foreign governments have begun treating memory access as a matter of “economic security,” and that the situation has produced what he called “near-chaotic lobbying” from both corporate customers and governments. “Right now, companies absorb the pressure. Governments will start pressuring other governments soon,” he said.

He warned that sustained high prices would invite geopolitical retaliation of the kind other Asian exporters have faced before. He cited Tesla CEO Elon Musk’s stated interest in chip manufacturing as an early sign of that pull.

Chey described the current moment in stark terms: “Building where we can, as fast as we can, has almost become the lifeline of Korea’s semiconductor industry.”

The Expansion That Isn’t Fast Enough

SK Hynix is racing to expand.

The company plans to double its memory wafer production capacity over the next five years. It has pulled the first clean room at its Yongin cluster forward to February 2027 from May, and committed an additional $14.52 billion (21.6 trillion won) in March of this year. It is converting its Cheongju M15X plant into a dedicated DRAM base for high-bandwidth memory.

A larger long-term expansion into the country’s southwest, jointly planned with Samsung Electronics, will be sited at the Gwangju military airfield.

And in the U.S., SK Hynix is building a $3.87 billion HBM advanced packaging and R&D facility in Indiana, with mass production expected to begin in the second half of 2028. The Indiana project has been awarded up to $458 million in CHIPS Act funding and could receive up to $500 million in loans. Air Liquide has also invested over $170 million to build two industrial gas production units in Indiana to support the facility.

The company raised a record $26.5 billion in a U.S. stock offering in July 2026 — the biggest first-time share sale by a foreign company in U.S. history. Analysts estimate SK Hynix can generate more than $300 billion in free cash flow this year and next.

But even this may not be enough. SK Hynix CEO Kwak Noh-Jung warned that the memory-chip shortages will probably persist beyond 2030. Customers are signing long-term contracts because “they believe that the shortage situation will last for longer,” Kwak said in his first-ever English-language interview.

What This Means

The AI chip shortage is not a temporary supply-demand imbalance. It is a structural transformation of the semiconductor industry — and it is happening faster than the industry can respond.

Chey framed the challenge in terms that go beyond any single company or country. He warned that if prices remain too high, new competitors will enter the market, and governments will intervene. He argued that semiconductor companies should not maintain high prices by restricting supply — that expanding supply to grow the market is the more sustainable long-term strategy. “Even if profit margins are slightly compressed, expanding supply to grow the market itself will yield greater returns in the long run,” he said.

He also warned of broader bottlenecks. The AI industry now faces a comprehensive shortage of GPUs, memory, and electricity — “core infrastructure.” He predicted that beyond memory, other unexpected bottlenecks will emerge.

“This is a fundamental structural change in the industry, not an ordinary economic cycle,” Chey said. “The supply shortage is likely to continue in the short term.”

The question is not whether the shortage will end. The question is who will still be standing when it does — and who will have been priced out of the market entirely.