SoftBank plunges 13% and SK Hynix drops 10% as AI cost fears hit Asia

SoftBank Group plunged 13% and helped drag a broad selloff across Asia technology stocks, while SK Hynix slid 10%. The common thread is blunt: mounting concerns over the rising cost of artificial intelligence infrastructure. This is not a niche trade. It is the market repricing what it will take to build and scale AI, and how quickly those costs can translate into revenues.

The first-order takeaway for executives is that capital markets are currently treating AI infrastructure as an expensive, time-sensitive problem. When SoftBank takes the lead to the downside, it is a signal that investors are willing to cut risk across the AI supply chain, not just in one corner of the sector. And with SK Hynix down 10%, the selloff is explicitly touching the hardware layer that underpins compute and memory capacity. In other words, the “AI costs” worry is not staying abstract. It is showing up in the prices of companies positioned to supply the machinery of AI.

To understand why this matters, zoom out to how AI infrastructure is funded and valued. AI is a capex-heavy buildout. You do not just buy software. You fund data centers, networking, specialized chips, memory, and the engineering to keep everything running. The market dynamic is simple and brutal: if investors believe the cost to expand AI capacity keeps rising, they often demand either faster monetization or lower risk. If neither happens quickly, the valuation math compresses.

That is why “infrastructure cost” concerns can cascade. When uncertainty rises around payback periods, the market tends to de-risk across the board. Asia trades often respond quickly to U.S. signals, and this move did not happen in a vacuum. The selloff in Asia tech stocks tracked declines in the U.S., which is an important reminder for boards and treasury teams: global risk appetite is still chained together. You can have different company fundamentals, but liquidity and sentiment can dominate short-term performance.

There is also a corporate governance angle worth watching, especially for major investment platforms and holding companies. SoftBank is the kind of name that can be sensitive to shifting expectations about how and when its portfolio companies and strategic bets will monetize. When the stock drops sharply in a session like this, it does not only reflect current operating results. It also affects how investors think about future capital deployment, exit timelines, and the credibility of growth narratives. Even if the underlying business performance is unchanged, a sharp repricing can force new constraints: investors may ask for clearer execution milestones or adjust the discount rate applied to long-dated AI claims.

For SK Hynix, the 10% slide is a reminder that infrastructure cost worries can hit suppliers even before any single customer order is publicly known. Memory and related components are often tied to demand expectations for compute expansion. If AI capex expectations soften, the supplier story can still get hit first, because revenue expectations often lag. The second-order effect is that boards may scrutinize whether investment plans, capacity expansion, and supply commitments line up with the new, more cautious view of demand. In tight markets, suppliers do not just compete on product. They compete on timing.

Now layer in the practical reality for decision-makers: when the market focuses on AI infrastructure cost, it can change how executives allocate budgets internally. Finance teams start pushing on unit economics: what does it cost per inference, per training run, per deployed model, per customer? Procurement teams ask whether existing agreements should be renegotiated, and engineering teams prioritize efficiency, whether through better utilization, stronger scheduling, or reduced waste. Even without new regulation in the immediate news cycle, the market is effectively acting like a regulator of sorts, by setting the cost of capital.

So what should peers take from this? If SoftBank can plunge 13% and SK Hynix can drop 10% on infrastructure cost concerns, then the bar for AI optimism just moved. Companies that are planning aggressive AI buildouts need to show credible paths to monetization or improved cost curves. Otherwise, the next leg of the selloff can stay contagious across Asia and into U.S. trading again. In an AI arms race, the weapons are expensive. The market is currently asking: who can afford them, and who can prove the ROI faster?