Odaily reported that Serenity posted an AI supply chain view on X, arguing that demand for AI infrastructure is still pushing several parts of the market into a long expansion cycle. The list included storage, advanced packaging, compute financing, optical communications, power supply systems, and electronic components. In Serenity’s view, the AI supply chain remains in a rapid growth stage.
Storage remains a central bullish theme
In storage, Serenity cited a UBS forecast that gross margin at traditional DRAM makers could keep rising. Serenity said manufacturers such as Micron may see gross margin reach an unprecedented 95% in 2027, a level that could even exceed margins from HBM products.
Serenity also said SanDisk’s long-term agreements already cover about two-thirds of its 2028 production capacity. The minimum contracted revenue tied to those agreements stands at $93 billion, while the company’s current market capitalization is about $239 billion. Serenity used that comparison to argue that the company may not be easily treated as a traditional cyclical stock, as its future revenue targets could extend for years.
Cloud infrastructure names may benefit from longer GPU use
On cloud infrastructure, Serenity said CoreWeave has signed an agreement to use Nvidia A100 GPUs through 2029. According to Serenity, that is positive for newer cloud computing companies such as Nebius and Iren, and it also weakens part of the bearish case from investors who expected older GPUs to depreciate quickly.
AI model companies are still growing faster than expected
Serenity said growth at AI model companies continues to beat expectations, with frontier AI labs still expanding at a very fast pace. Serenity added that a slowdown in that growth would be the more concerning signal.
Market estimates now suggest Anthropic could reach $190 billion to $200 billion in revenue by 2028. Serenity pointed to that range as another sign that demand tied to the model layer is still climbing rather than leveling off.
Advanced packaging and semiconductor infrastructure stay constrained
At the same time, Serenity said advanced packaging and semiconductor infrastructure remain the core bottlenecks in the sector. A TSMC executive in charge of advanced packaging said the industry may face not only memory shortages in the next few years, but also tight supply of ABF substrates.
Serenity’s conclusion was that the AI infrastructure supply chain is still expanding across GPUs, memory, advanced packaging, power, optical communications, and electronic components, with long-term demand growth visible in each of those segments.