Serenity: Don't label all AI infrastructure as a bubble; current phase is deleveraging, not demand collapse

November 8 — Serenity posted that the media's repeated use of "AI bubble" is inappropriate. Among infrastructure names, memory companies from Samsung to ESMT trade at forward P/E ratios of only about 1x to 3x, and HBM makers have signed multi-year long-term agreements with hyperscalers that include price floors. Optical companies such as POET even trade at a market cap about $97 million below net cash, effectively assigning a negative valuation to the operating business. Meanwhile, AI-related companies that still carry premiums continue to record demand for the next 3 to 5 years, for example long-term agreements of about $20 billion with Coherent and about $5 billion with Lumentum.

Serenity said it is hard for the media to directly characterize the current high-volatility optical stocks with negative enterprise value, and the low-valuation memory stocks that have locked in multi-year demand and price floors and remain highly profitable, as a bubble. There is clear divergence within the theme: some startups such as Instinct have private valuations that are somewhat bubbly, some overly leveraged Neocloud platforms with improper financing, and the valuations and commitments of some AI labs all warrant caution. However, the market often habitually uses "leverage" or "debt" issues to scare investors.

Serenity believes that although valuations may compress in phases due to the macro environment, regulatory uncertainty, or capex cuts, for companies that truly underpin AI buildout, this looks more like a 2026 "great deleveraging" rather than fundamental demand being destroyed. With subsequent earnings disclosures and free cash flow accumulation, Serenity is bullish on valuation repair in bottleneck segments, and expects the market will eventually distinguish what is sustainable growth from what is genuine bubble valuation.