Nutstuff loves AI but at the same time has said, “Dance hard, but do so near the fire exits.” One of the things making me increasingly cautious on semiconductors is that the market seems to be extrapolating today's shortages while largely ignoring tomorrow's supply. The latest example comes from South Korea, where China's YMTC has re-entered the market through its Zhitai brand with PCIe 5.0 SSDs that are no longer competing at the low end, but directly against Samsung's premium offerings. That is significant because it challenges one of the market's most comfortable assumptions: that China remains years behind in advanced memory technology. If YMTC can deliver competitive performance at lower prices while simultaneously bringing vast new capacity online, then the industry's economics may prove far less durable than investors currently assume. At the same time, Wall Street's optimism has become increasingly concentrated. The oft-quoted claim that Micron represents 30% of 2027 S&P earnings appears overstated, but the reality is still extraordinary. Current estimates suggest Micron alone accounts for roughly 7% of expected S&P 500 EPS growth in 2027, while Nvidia and Micron together drive around one-third of projected index earnings growth. That is a remarkable concentration of expectations in two companies whose fortunes are heavily tied to the continuation of today's AI infrastructure boom. My concern is not that AI fails. Quite the opposite. My concern is that AI succeeds, attracts unprecedented amounts of capital, and in doing so triggers the same cycle that has defined semiconductors for decades. Shortages create extraordinary profits. Extraordinary profits attract enormous investment. Enormous investment eventually creates excess capacity. The Korean YMTC news matters because it provides a glimpse of what that future may look like. Investors have become fixated on demand for AI chips, memory and compute, while paying far less attention to the sheer volume of supply now being built globally. The irony is that the more successful AI becomes, the greater the incentive for competitors, governments and new entrants to flood the market with capacity. China's ambitions extend well beyond mature-node chips. If YMTC's expansion plans materialise, and if software innovations from the likes of DeepSeek and others continue reducing memory requirements, then today's assumptions around pricing power, margins and earnings durability could prove wildly optimistic. Investors are increasingly valuing semiconductor companies as though scarcity is permanent. History suggests scarcity is usually the most temporary thing in markets. In brutally simple terms, the AI boom has convinced investors that every shovel seller will become rich. The YMTC development is a reminder that China is busy building shovel factories. Lots of them. If they work, the technology revolution will remain real, but the profits available to today's incumbents may look very different from the ones currently embedded in consensus forecasts and market valuations.
2June 4, 2026 245 1
