If you needed a sign of how seriously global capital markets take the AI buildout, consider what happened this week in New York. SK Hynix — a South Korean chipmaker most business leaders have never heard of — raised $26.5 billion in a US stock listing, making it the largest foreign IPO in American history. Its permanent ticker, SKHY, starts trading on Nasdaq today.

That number beats Alibaba’s $25 billion US debut in 2014. It was more than seven times oversubscribed and drew demand from over 500 investment firms. The market is not being subtle about where it sees the next decade of growth.

Who Is SK Hynix and Why Does It Matter to AI?

SK Hynix makes memory chips — specifically a type called High-Bandwidth Memory, or HBM. If that sounds obscure, here’s the context that matters: HBM is the component that makes NVIDIA’s AI GPUs work. Every H100 and GB200 server that trains a large language model, runs an inference pipeline, or powers an AI agent fleet depends on HBM to move data fast enough to keep the GPU busy.

SK Hynix controls roughly 60% of the global HBM market by revenue. There is no realistic AI at scale without them.

The company priced 177.9 million American depositary shares at $149 each. Shares rose 12.8% on their first day of trading last Friday. Today’s opening under the permanent SKHY ticker marks the full transition to US markets.

Where the Money Goes

The $26.5 billion raised is not going into dividends or buybacks. It is going into production capacity:

• A new memory fabrication facility in South Korea, being built specifically to address the global HBM shortage caused by AI demand
• A new chip packaging facility to handle the complex stacking process that HBM requires
• Next-generation EUV lithography machines — the scanners that make advanced chips physically possible

This is the kind of infrastructure investment that takes two to three years to come online. The companies placing bets on SK Hynix today are betting that AI demand will still be accelerating in 2028 and 2029.

The US Government Is Paying Attention

The IPO came with an unusual request from US officials, who urged SK Hynix to use some of the capital raised to build new fabrication facilities on American soil. The US government has been pushing to reduce dependence on Asian chip manufacturing since 2022, and an AI memory manufacturer with 60% global market share is exactly the kind of strategic asset Washington wants closer to home.

Whether SK Hynix ultimately builds US fabs will be one of the more interesting supply chain stories to watch over the next two years.

The HBM Shortage Is Real

Behind the IPO is a supply problem that has constrained AI deployments across the board. HBM production is extraordinarily complex — chips are stacked vertically and connected with microscopic through-silicon vias, which requires precision that limits how fast capacity can be added. Even with record investment, HBM supply has been unable to keep pace with the number of AI server builds happening globally.

That shortage has had direct effects: it has slowed NVIDIA’s ability to ship AI servers, created long waitlists for cloud GPU access, and kept the cost of AI compute higher than it would otherwise be.

The new SK Hynix fabs being funded by this IPO are the industry’s primary answer to that bottleneck.

What This Means for Business

If you are a business leader thinking about AI, the SK Hynix IPO tells you a few concrete things.

First, the AI buildout is not slowing down. Five hundred investment firms do not collectively pour money into an oversubscribed chip IPO because they think AI is a passing trend. They do it because the fundamentals — compute demand, enterprise adoption, government investment — all point in the same direction.

Second, the cost of AI compute will continue coming down, just not instantly. As new HBM capacity comes online over the next two to three years, the per-unit cost of running AI workloads will fall. Businesses that are building AI capabilities now, before the infrastructure catches up, will have an advantage over those who wait until it feels fully proven.

Third, AI capability is deeply tied to physical infrastructure that takes years to build. The “AI is just software” framing misses something important: the models that power AI agents, voice assistants, and automated workflows exist because of fabs in South Korea and Taiwan. Supply chain events in those regions affect every business using AI, directly or indirectly.

For teams at Enterprise DNA building data skills or deploying AI agents, the broader lesson is that AI is becoming a long-term infrastructure bet — not just a tool you turn on and off. The investors behind SKHY are making that bet at a scale of billions. The question for every business is what their version of that commitment looks like.

Want to understand how AI fits into your business strategy? Book a discovery call with Sam McKay to map it out.