The world faces the risk of an 'electric shock' from AI.

Behind the billions of calculations performed every second lie inherent risks to digital infrastructure and energy security as traditional data centers continue to expand into enormous structures.

The "Spikes" phenomenon and the risk of macro-level load disruption.

Unlike traditional data centers (DCs) which consume electricity at stable and predictable levels, AI operational infrastructure has a much more volatile and unpredictable load characteristic. Speaking at Computex 2026, Yin Zheng, Executive Vice President for East Asia and China at Schneider Electric, emphasized that the nature of IT loads serving AI requires entirely new governance mechanisms due to the sudden changes in power supply.

The cause of the aforementioned fluctuations stems from the way artificial intelligence models operate. According to Himanshu Prasad, Senior Vice President of Schneider Electric, during data training or inference, thousands of GPUs are activated to operate synchronously at the same time. This process creates extremely strong sudden power surges in the system, resulting in localized load spikes ("Spikes"). Without a mechanism to smooth and control the load, this synchronization will cause violent fluctuations, directly threatening the stability of the power line.

Mr. Himanshu Prasad shared this at Computex 2026.

The surge in electricity consumption has also propelled technological infrastructure into an unprecedented era. The global data industry is witnessing a dramatic shift from 10-100 megawatt facilities to "mega-projects" with capacities up to 1 gigawatt, equivalent to the electricity consumption of a medium-sized city.

You may also like

Free Gemini users receive a range of premium features.Google is offering Gemini's new inference modes for free, helping users receive more in-depth answers.

According to Doug Warren, Senior Vice President of AVEVA, at this scale, the conventional concept of a "data center" no longer accurately reflects reality. Modern AI infrastructures have a level of complexity, energy consumption, and technical requirements comparable to heavy industrial complexes such as aluminum smelters or semiconductor mega-factories. The system must operate continuously 24/7 and absolutely cannot tolerate any disruption.

This enormous scale also leads to a systemic collapse risk. Mr. Himanshu Prasad warned that at Gigawatt-scale power plants, even a brief disruption on the grid that disconnects the database, causing a massive amount of power to suddenly disappear, would bounce back through the transmission system, creating a disproportionate drop and potentially causing the entire regional power grid to collapse.

Solving operational problems with a "Grid Awareness" software solution.

Given the enormous increase in heat energy and the numerous complex requirements for electromechanical systems, maintaining manual operation methods is completely outdated. Yin Zheng asserts that multi-gigawatt systems with high variability cannot be managed solely by human effort. These systems must utilize automation, artificial intelligence, and intelligent software to monitor and maintain reliability throughout the design lifecycle and into actual operation.

To mitigate risks early on, experts recommend that data centers implement "Grid-aware operations" strategies. Doug Warren shared that real-time data governance software solutions need to continuously monitor changes in AI workloads, thereby providing accurate predictions of their corresponding impact on the national power grid.

At the same time, when an AI factory operates at peak capacity, potentially triggering thousands of system alerts simultaneously, the application of intelligent alert management technology is essential. This system helps to categorize and group error notifications, assisting operating engineers in making timely and accurate technical interventions.

You may also like

Safety guidelines for small traders and vendors at Thanh Xuan Market.Based on a simulated fire scenario at a kiosk in Thanh Xuan Market, more than 130 traders and local fire fighting forces participated in a fire fighting, rescue, and relief drill under the guidance of the Hanoi Fire and Rescue Police Force.

The explosion of the artificial intelligence era shows that simply focusing on designing more powerful generations of GPUs is not enough. This new wave of technology will not materialize if countries and businesses cannot solve a crucial problem: building highly compatible and durable AI super-factories and establishing a safe "coexistence" mechanism with the national power grid infrastructure.

According to Thanh Nien Newspaper

Source: https://baoangiang.com.vn/the-gioi-doi-mat-nguy-co-soc-dien-vi-ai-a487803.html