Nvidia CEO Jensen Huang has once again addressed the topic of AI bubble. Recenty, Huang talked about the issue that whether artificial intelligence is moving towards a bubble reminiscent of the dot-com crash. Speaking at the 2025 US-Saudi Investment forum, Huang delivered a three-minute response that reframed the conversation. Instead of focusing on the market valuations, Huang argued that the investors are worried about a bubble may be overlooking the fundamental technological shift which is already reshaping global computing. Speaking at the event, Huang started pointing to a structural constraint in computer science: the end of Moore’s Law, the decades-long trend of transistor density doubling every two years. Huang added that the trend has now plateaued, creating a widening gap between skyrocketing demand for computing power and the industry’s ability to supply it through traditional chips. “This isn’t speculation—it’s a fundamental challenge driving the need for new computing paradigms,” Huang said.This gap, he explained, is why companies are not merely experimenting with AI infrastructure but committing massive capital expenditures to it out of necessity.
The supercomputer revolution already happened
Huang also stressed on the fact that the AI boom is not about betting on future technology but it is about relocating computing resources today. “The world is voting with real capex,” he said, noting that industries are pouring investment into GPUs because traditional CPUs cannot efficiently handle the costs of AI training. Without this shift, global computing demands would require trillions in annual spending.From finance and healthcare to research and manufacturing, accelerated computing is already embedded in daily operations, proving that this is not hype but an ongoing transformation.
Huang’s three-part case for why AI demand is real
Huang said the industry is undergoing structural shifts that have little to do with hype. According to him, companies are moving away from traditional CPU-based systems, which can no longer keep up with the heavy computing demands of AI.He told analysts that data processing, search, advertising and engineering workloads are increasingly shifting to GPU-based infrastructure because they “need the AI,” as per CNBC.He added that AI is no longer just improving existing applications but is creating entirely new categories of software. Huang also pointed to the rise of “agentic AI” — systems capable of reasoning and operating with limited user input — which will require even greater computing power in the coming years.Nvidia, he said, is the only company positioned to power all three transitions.
Growth challenges ahead, despite strong demand
Huang avoided naming a single constraint to Nvidia’s growth when asked by analysts, instead highlighting the scale and “newness” of the AI transformation, reported Reuters.Analysts, however, point to external roadblocks. eMarketer’s Jacob Bourne told Reuters that physical constraints — such as power availability, land and grid capacity — could limit how quickly cloud providers can turn GPU demand into real revenue.Nvidia, largely restricted from selling advanced chips to China due to US export rules, is now looking to the Middle East as an emerging market. The US commerce department has authorised the export of up to 35,000 Nvidia Blackwell chips to two firms in Saudi Arabia and the UAE — a shipment worth more than $1 billion by market estimates.
