📰 Key Takeaways

Toshiba once had grand ambitions to dominate both memory chips and nuclear power, but after a management crisis broke out in 2016, it was forced to sell off these two core businesses one after another. The memory unit changed hands at a relatively depressed valuation at the time, and after integration was rebranded as Kioxia; the US nuclear subsidiary Westinghouse Electric filed for bankruptcy protection under financial pressure. However, ironically, about a decade later, these two businesses that Toshiba gave away have both seen explosive growth driven by the AI wave. Kioxia’s market cap has now climbed to around $350 billion, making it a major global NAND flash memory supplier, benefiting from the huge demand for storage capacity in AI training and inference; Westinghouse Electric has also seen its value surge as countries reassess nuclear power as a stable power source for AI data centers. Toshiba’s history has become a classic example of missing out on the era’s opportunities amid crisis. See the original article for details.

💬 JudyAI Lab Perspective

Toshiba’s story teaches us one thing: assets forced to be sold during a crisis are often exactly what the next era needs most.

From an AI builder’s perspective, there’s an industry reality worth noting: infrastructure assets are often hardest to price precisely when demand hasn’t yet emerged. Kioxia’s memory business and Westinghouse’s nuclear assets were sold by Toshiba at depressed valuations; about a decade later, AI training’s massive demand for storage capacity and AI data centers’ craving for stable power have caused both businesses to explode. Toshiba’s mistake wasn’t misjudging the technology itself—it was being forced to miss the timing due to financial pressure from the 2016 management crisis. When we look at AI supply chains, maybe we need to pay more attention to those “currently平淡估值，卻默默支撐AI運算的底層資源” (currently平淡估值 = currently flat/low valuation).

下次評估某項看起來不起眼但支撐AI運算的基礎資源時，值得多停一秒問：它的需求曲線，在五年後長什麼樣子？

📅 Source Info

🔗 Further Reading