Jukan
Semiconductor & AI Infrastructure Research at Citrini
Ajinomoto, like Nittobo, is too conservative and cares too much about customer relationships and market share to exploit its pricing power. Despite a severe shortage of high-end glass fiber for AI servers, Nittobo says it has no plans to raise T-glass prices and will instead expand capacity to defend share, which is why its stock has moved sideways. Unlike memory companies aggressively raising prices, these Japanese suppliers are leaving profits on the table. x.comÂ
Nittobo and Ajinomoto are unlikely to deliver alpha because they will not monetize their monopoly pricing power, putting customer relationships ahead of shareholders. SK hynix and Micron raise prices when demand exceeds supply and reinvest profits to widen their technology lead; per @jukan05, Nittobo has no intention of raising prices. In fast-moving chips, customer loyalty does not merit a high valuation premium, even with Ajinomoto holding 95% of ABF film and no qualified Chinese rival yet. … Show more
Furukawa Electric is not worth chasing because companies like it will fail to raise prices on high-demand products and stay committed to low-margin, low-growth segments. This is not a rejection of growth. A post in the thread notes that end customers such as Nvidia and Micron earn 75% gross margins, yet Japanese suppliers prefer to let Americans take the margin. x.comÂ
Ibiden is overpriced and set to lose market share, so AT&S, Kinsus and Ajinomoto are the better substrate positions. This runs against FundaAI (@FundaAI), which argues that high-end ABF capacity is failing to keep pace with AI silicon and sees rerating room for Ibiden at about 45 times FY3/28 earnings. x.comÂ
Memory makers' pricing power looks unsustainable because betting on a long-term bottleneck is betting against human ingenuity. Memory players are squeezing customers in a way ASML and TSMC historically chose not to; Micron may have the right to get back at Apple, but that hurts long-term relationships, and disregarding the non-AI business could come back to bite. Even if inference demand drives memory prices, as @bnemethbrown argues, whether regular DRAM gets commoditized is still open. x.com … Show more
Memory stocks should do well, though they are businesses hated by their own customers, who are desperate for workarounds. Per @DeepSailCapital, hyperscalers are paying a memory tax and have a huge incentive to escape it through China's memory ramp, different designs or different models. x.comÂ
Kioxia faces eroding pricing power as surging chip production capacity threatens the storage-chip boom. TSMC's signal of a hike in 2026 capex to $64 billion fed fears that new capacity will erode the pricing that made memory a gold mine, amid broader doubts that AI spending turns into stable profits. The stock has fallen 52% from its June highs. After a 600% climb on retail leverage, the descent is rarely gentle, and in cyclical industries supply-side realities catch up to the hype. x.comÂ
