Harry Stebbings: AI Infrastructure is Not a Bubble
Harry Stebbings shared Andrew Feldman's view, stating that the claim of a bubble in AI infrastructure is incorrect.
Feldman emphasized that the current pace of data center construction is far behind demand, with companies "chasing demand" rather than building ahead.
He pointed out that the shortage of AI computing power remains severe, and there is still significant room for infrastructure spending.
Source: Public Information
ABAB AI Insight
Andrew Feldman, as the founder of an AI hardware company, has been at the core of the computing power supply chain for a long time. His statement continues his assessment that demand far exceeds supply, consistent with the "Power Wall" perspective emphasized by institutions like a16z.
In terms of capital flow, large cloud providers and massive AI companies are continuously investing hundreds of billions of dollars into data centers, electricity, and chips, concentrating resources on suppliers that can deliver computing power the fastest, rather than slowing down. The motivation is to create long-term barriers by securing scarce electricity and land resources.
Similar to the rapid reversal from "overcapacity concerns" to shortages in cloud computing from 2021-2022, and the current ongoing GPU shortages, AI infrastructure is currently in a stage of tight balance where demand is explosively growing far beyond the pace of construction.
Essentially, this is a concentration of capital: the demand for AI training and inference is growing exponentially, while the long construction cycles of data centers and slow electricity approvals lead to a severe lag in supply. The mechanism is that the actual token and model size demand continues to exceed expectations, forcing capital to concentrate more intensely on the AI infrastructure sector, rather than entering a correction before a bubble bursts.
ABAB News · Cognitive Law
A true bubble occurs when too much is built, not when demand fails to keep up with construction.
When everyone is shouting about a bubble, supply is actually still lagging significantly.
Demand leading construction is the most reliable pricing signal in the AI era.