Hello, my friends! Welcome back to my latest analysis report.
In this report, I'll be discussing whether there is AI bubble in the market or we are just seeing some AI air pockets.
People are discussing whether we are currently in a massive AI bubble which is going to burst like 2000 since tons of semiconductors are surging since 2025 and KOSPI is rising in a parabolic move. But at the same, some argue that we are merely at an early stage of AI development and those hyperscalers will continue invest substantially for AI infrastructure.
The AI pockets we are seeing
1. The Physical Infrastructure Bottleneck
While we see a severe shortage of memory chips lasting until late 2027 or beyond, the real physical bottlenecks are power availability and AI data center construction. It is reported that nearly half of all planned data center projects are facing delays or cancellations due to local NIMBY protests, labor shortages, and inadequate power supplies. Consequently, even if you could procure the necessary chips within three to six months, the actual facilities will not be ready for two to five years. This stark mismatch could ultimately force a pause in chip procurement altogether.
2. The Chip Depreciation
Right now, leading chipmakers like Nvidia and AMD are releasing massive technological upgrades every 12 to 18 months. For instance, from Hopper (2022), Blackwell (2024) and then Vera Rubin (2026). Since the market will immediately rush to adopt the latest generation of AI chip, older premium chips quickly become functionally obsolete. As a result, an AI chip begin losing its value the moment it leaves the factory, much like a new car.
That said, we are likely still in the early stages of AI development. What we are experiencing is an AI air pocket rather than a structural bubble. Hyperscalers like Google or Microsoft are fighting for their long-term survival; they cannot afford to slow their capital expenditure because they know that losing the AI race means being eliminated from the market entirely. For these companies, stopping AI development is simply "game over."
What are your thoughts? Let me know where you find any flaws in the comments below!
