The AI industry faces its biggest challenge: a power shortage, not a chip shortage.

DNVN - In a recent podcast interview with Sam Altman on Bg2 Pod, Satya Nadella, CEO of Microsoft, revealed a paradox in the artificial intelligence (AI) industry: instead of a chip shortage, it's the power supply issue that's the real bottleneck. He stated that there are GPUs sitting in storage because they can't be plugged in due to insufficient power to operate the infrastructure.

Tạp chí Doanh Nghiệp•06/11/2025

From Satya Nadella's perspective, the AI industry today isn't hampered by a lack of components – but rather by the limited power supply capacity for data centers. In the Bg2 Pod podcast, when asked about Nvidia CEO Jensen Huang's opinion that there wouldn't be an oversupply of computing power in the next 2-3 years, Nadella responded: “Our biggest problem right now isn't a shortage of AI chips, but a lack of power to plug them in. I even have GPUs in storage that I can't use simply because I don't have enough power infrastructure.”

(Illustrative image)

Mr. Nadella also pointed out that many "data center shells"—buildings prepared for the installation of electricity, water, and server infrastructure—are under construction but cannot yet operate due to unreliable power supply.

This reality reflects a turning point in the perspective on AI development: While hardware like chips and GPUs has seen significant investment and supply chains are stabilizing, power infrastructure has become a new barrier to the speed and scale of AI deployment. Many large technology corporations are now seeking to address this issue by investing in proactive power sources such as small modular reactors (SMRs) to ensure a reliable power supply for their data centers.

Furthermore, the issue of electricity consumption not only affects large companies but also poses socio-environmental challenges: the enormous electricity consumption of data centers has led to higher electricity bills in some areas of the US, sparking debate about the impact of AI on life and the environment.

Nadella and Altman also discussed the future when advanced language models like GPT-5 or GPT-6 could run locally on personal devices with significantly lower power consumption than current models. Altman stated, "The day will come when we can create a consumer device that can run the GPT-5 or GPT-6 model entirely locally with low power consumption."

From these insights, it's clear that the AI industry is entering a new phase: it needs not only powerful chips, large data networks, and sophisticated algorithms, but also a stable, sustainable, and sufficiently large power supply to ensure that technological development is not hampered by… power shortages.

Vietnam and other developing countries, when setting digital transformation goals or developing data centers, need to consider the lessons from the global situation: electrical infrastructure – from power supply to cooling – must go hand in hand with investment in hardware and algorithms, otherwise, they will face the situation of "slowing down due to… power shortage".

Nguyen Bach

Source: https://doanhnghiepvn.vn/cong-nghe/nganh-ai-doi-mat-thach-thuc-lon-nhat-thieu-dien-chu-khong-phai-thieu-chip/20251106103714043