Written by Silvia Pavelli
The market still talks about artificial intelligence as though the most important assets are the models, the apps, and the cloud platforms closest to the user. That language captures the consumer face of the boom, but it increasingly misses where the price discovery is happening. In the latest stage of the AI cycle, the marginal signal is not coming from chatbots or product demos. It is coming from memory. More specifically, it is coming from the Asian companies that sit closest to the supply bottlenecks now determining how fast the entire AI stack can expand.
The most revealing recent proof arrived in Business Insider’s account of Goldman Sachs’ latest South Korea call. Goldman lifted its 12-month target on the Kospi from 8,000 to 9,000 and described South Korea as its highest-conviction market in Asia. That alone would be enough to attract attention, but the rationale matters far more than the target. The bank’s argument, as reported, is that the Korean rally is being driven by AI-fueled demand for memory semiconductors dominated globally by Samsung Electronics and SK Hynix. In other words, investors are no longer treating memory as a background component in the AI story. They are treating it as one of the main places where the story gets priced.
That is a profound change. For decades, memory was the part of technology investors were trained to fear. It was cyclical, commoditized, and usually associated with volatile margins rather than durable market power. AI is rewriting that script because the current demand surge is not merely incremental. It is infrastructure-deep, capex-heavy, and tied to a compute race that still has no obvious ceiling. When that kind of demand collides with constrained supply, the firms controlling the bottleneck stop looking cyclical and start looking strategic.
The Goldman call, as summarized by Business Insider, is explicit about the scale of the repricing. The article says South Korea’s benchmark index is trading around 7,400 after having crossed 7,000, and that it is up around 77% so far this year. More importantly, Goldman expects hardware and semiconductor stocks to drive 300% earnings growth for South Korean corporates in 2026. That kind of number only makes sense if the market believes memory producers are sitting on a profit pool that remains both supply-constrained and hard to substitute away.
The deeper point is that Asia is no longer just participating in the AI boom. It is increasingly setting the marginal price of it. The reason is straightforward. AI models may be trained in Silicon Valley narratives, but they run through a supply chain whose scarcest pieces remain heavily concentrated in Asia. If the decisive economic constraint is memory availability, then the market must eventually grant more pricing power, valuation support, and strategic importance to the companies that control memory output.
Business Insider’s report also underscores why this is happening now rather than later. Goldman points to record shortfalls in DRAM and NAND relative to demand growth from hyperscaler investment. That phrase matters because it captures the industrial character of the current boom. Hyperscalers are not just buying software exposure. They are consuming physical components at a pace fast enough to tighten entire memory markets. Once that happens, the old hierarchy of glamour versus commodity starts to invert. The “commodity” supplier becomes the strategic winner precisely because nothing at the top of the stack works without it.
This has consequences far beyond South Korea. It changes how investors should think about geography, leadership, and even geopolitical leverage inside the AI trade. For years, the dominant assumption was that U.S. software and platform companies would inevitably capture the lion’s share of value because that is where the user relationship sits. That assumption now looks too narrow. In a supply-constrained phase of technological expansion, the highest-quality earnings may instead belong to the firms sitting at the hardest-to-replace point in the chain.
That is why Samsung and SK Hynix matter so much in this moment. They are not merely beneficiaries of more server shipments. They are increasingly the companies through which the market measures whether the AI buildout remains bottlenecked, how durable pricing will be, and whether the economics of compute can continue supporting premium valuations elsewhere in the ecosystem. Their importance is diagnostic as much as financial. When memory prices stay strong, they tell you the AI machine is still being fed faster than it can be supplied.
There is also a subtle political-economy layer to this. South Korea’s rally is not happening in a vacuum. The Business Insider piece notes that reforms aimed at reducing the old “Korea discount” have helped reinforce the move. That matters because governance improvements make it easier for global investors to own strategic exposure without demanding the same historical discount. When governance reform meets AI-linked scarcity, the result can be a much more aggressive rerating than either force would produce alone.
Still, the most important lesson is not about country allocation. It is about where the market now locates necessity. AI used to be priced mainly through the imagination of software. It is increasingly priced through the arithmetic of hardware. That arithmetic favors the firms with direct exposure to memory tightness, long-term supply agreements, and the physical constraints that separate hype from actual compute. Asia, and especially the Korean memory complex, sits squarely inside that zone.
This does not guarantee an uninterrupted rise. Memory has always been capable of looking permanently scarce right before a new capacity wave changes the balance. But that objection misses the current point. Right now, the market is not paying for an abstract future. It is paying for present scarcity and the earnings durability that scarcity creates. That is why the recent Korean rerating looks less like a speculative sideshow and more like a signal about where the AI trade is being anchored.
The sharpest reading of the past two days, then, is that Asia’s memory suppliers are no longer downstream participants in an American AI story. They are among the institutions telling the market what that story is worth. Goldman’s raised target, the emphasis on memory profits, and the focus on supply shortfalls all point in the same direction. In this phase of the cycle, memory is not just part of the AI boom. It is setting the price of it.
