Jensen Huang: Extremely Low Probability of an AI Bubble in the Next Five Years
nashnova research
Nvidia CEO Jensen Huang said an AI infrastructure bubble is possible long-term but highly unlikely within five years, because physical bottlenecks — chips, memory, power, land — keep capital from overheating.
Why does Huang rule out a bubble for five years?
Huang acknowledged AI infrastructure could eventually see a bubble, but judged the probability very low for at least the next five years.
This means → his confidence is not "AI will never overheat" but "the conditions for overheating do not yet exist."
What is physically blocking a bubble?
The binding constraint is not a lack of capital but build speed — chips, HBM memory (high-speed stacked memory designed for AI), land, power, and data-center construction capacity are all in shortage.
In plain terms = even if capital wanted to flood in, factories cannot produce enough chips and grids cannot deliver enough electricity — the money simply cannot be spent fast enough.
This reflects a shift in AI's real bottleneck: from "who is willing to invest" to "who can actually build."
What does this mean for the market?
Across-the-board supply constraints mean AI investment can only land gradually, not surge into severe oversupply.
This means → physical bottlenecks act as a natural throttle, pushing any potential bubble further into the future.
In plain terms = it is not that a bubble will never come — it is that until chip and power capacity are dramatically unlocked, the market lacks the ability to overbuild.
市场有风险，内容仅供研究参考，不构成投资建议。