We have entered what I like to call the “demand decade” – a period when electricity demand will grow faster than at any other time in recent memory. Electrified transport, the rise of data centers and artificial intelligence (AI), industrial decarbonization, and population growth are all converging to reshape how we generate and consume energy.

I’ve spoken about this inflection point at the recent Reuters Global Energy Transition event in New York City and again during a live webinar because I believe we’re facing both a challenge and an extraordinary opportunity. The challenge is simple: the grid, as it stands today, wasn’t designed for what’s coming. The opportunity hence lies in how we respond – by building systems that are not only larger, but secure, faster, and more resilient.

A Less Forgiving Grid
The energy transition has made the grid less forgiving. As renewables replace conventional generation, we lose the stabilizing inertia that large rotating machines once provided. With lower inertia, voltage and frequency can swing more rapidly and what used to unfold in minutes now happens in milliseconds.

This calls for a fundamental rethinking of how we design and operate our networks. Stability can no longer rely solely on mass and momentum, but rather on flexibility, near real-time intelligence, and the ability to act almost instantaneously.

AI, edge computing, and digital twin technology are helping us move from reactive control to predictive systems. They give operators the visibility to anticipate disturbances before they cascade. It’s not a theoretical ideal anymore – it’s happening now, in utilities that have begun embedding digital intelligence at every layer of the grid.

From Centralized Power to Distributed Intelligence
Two key forces are converging: rising demand and the decentralization of power generation. This means we need to shift from a hierarchal grid to a distributed one. Power will increasingly flow in two directions, and decisions need to be made closer to where data and events originate.

This is where digital substations become essential. They are no longer just physical assets, but adaptive digital nodes that are reconfigurable, software-defined, and capable of analyzing and acting locally. By embedding sensors, analytics, and edge control, substations can evolve from silent transfer points to active participants in maintaining balance and reliability.

Modeling and Managing Risk
Some recent global outages have shown just how quickly a local fault can cascade. To manage this new level of complexity, it’s important to move from static contingency analysis to dynamic modeling.

Through digital twins and near real-time simulation, operators can continuously stress-test their systems under real conditions. Combined with AI and high-resolution sensing, these tools enable a much faster, automated response. Our Zonal Autonomous Control (ZAC) technology embodies this principle, allowing parts of the grid to operate semi-autonomously while remaining harmonized with central control functions.

Breaking Down Data Silos
AI’s greatest limitation today is not the algorithm – it’s actually the data. Much of the information we need remains siloed at the edge or lost before it can be used. In my estimate, the next breakthrough will come from data fusion, combining simulations with AI to produce hybrid models that are more accurate and explainable.

Edge commuting will make this possible: by processing data where it’s generated, it’ll reduce latency, preserve context, and unlock predictive capabilities at scale. And this, well this is the key to a grid that learns as it operates.

Modernization Without Reinvention
Many utilities tell me they feel trapped between urgent modernization needs and the realities of aging infrastructure or regulatory inertia. My advice is: don’t wait for a full reset. Build evolutionarily.

Digital solutions like asset performance management and device management can be layered onto existing grid units. These tools extend equipment life, improve performance, and help direct capital to the highest-impact upgrades.

I often point to the examples of Transelec in Chile, which supplies electricity to more than 96% of the country’s population, and Tohoku in Japan - the largest APM Enterprise solution with 20M assets. By adopting GE Vernova’s GridBeats™ portfolio, specifically GridBeats™ Asset Performance Management (APM), they gathered asset-condition data, assessed risk proactively, planned maintenance effectively, and integrated with other IT systems. The result included fewer unexpected outages and higher network availability and the highest impact of their capital investment and maintenance focus. This showcases resilience in action, with data and foresight working hand in hand.

However, equally important is collaboration. Innovation that brings together utilities, regulators, and technology partners can accelerate pilots and provide evidence-based approaches. Because modernization doesn’t have to be disruptive. It can be deliberate, measurable, and fast.

As data centers, industrial players, and communities deploy their own generation and storage, grid operators are evolving into the role of system orchestrators. Their job is to integrate diverse resources (both centralized and distributed) while ensuring fairness, affordability, security, and reliability.

The Vision for a Resilient Grid
At the end of the day, a truly resilient grid is adaptive, intelligent, and collaborative. It blends technology with teamwork. It senses, predicts, and restores. And it evolves continuously to meet the changing landscape.

That spirit of collaboration is at the core of our mission. GE Vernova’s recent five-year, $50 million partnership with MIT is part of that commitment – advancing innovation in electrification and driving the energy transition while developing the next generation of energy leaders.

Electricity underpins everything in our modern lives. As we enter the decade of demand, we have the tools, the knowledge, and the imagination to build grids worth of the future they will power. To hear more, check out the webinar here.

The dawn of the Artificial Intelligence (AI) era is upon us, and with it comes a seismic shift in how we build, deploy, and power our digital infrastructure. AI is the present and the future of augmented innovation across industries—from defense to healthcare, finance to software engineering, transportation, and manufacturing. As AI continues to evolve, so too must the infrastructure that supports it. Enter the AI factory: a new generation of data centers designed specifically to handle the massive computational demands of large-scale AI workloads.

The Rise of AI Factories

An AI factory is not an average cloud-related data center. They are purpose-built to support the high-density, high-frequency, and high-performance computing requirements of AI applications. These facilities are characterized by their gigawatt-scale capacity, drawing from thousands of kilowatts to ultimately one megawatt per rack power loads. According to recent data, the existing 12,000 data centers accounting for a total of 135 gigawatts worldwide should grow towards 300 gigawatts of capacity by 2030, with an upgraded installed base of 15,000 data centers[1]. Within this growth, the average data center size is expected to increase by 72%, with more than 20 gigawatt-scale AI factories – each ranging from 1 to 5 gigawatts – projected to be built by 2030. The International Energy Agency further projects that global data center electricity consumption will more than double by 2030, with the U.S. accounting for nearly half of that growth[2].

The Challenge: Powering Gigawatt-Scale AI Factories

The rapid proliferation of AI workloads is straining existing power grids, revealing significant challenges in capacity, reliability, and infrastructure deployment. Limited transmission capacity, aging substations, and lengthy permitting timelines are just a few of the hurdles facing the connection of new AI factories to national grids. Moreover, the dynamic and unpredictable nature of AI workloads – characterized by frequent, large swings between high and low power consumption and high frequency load variability – poses unique stability challenges for the grid and on-site generation power quality.

Furthermore, traditional data center requirements for power supply—reliability, latency, affordability, and sustainability—are amplified in the context of AI factories. The intermittent nature of renewable energy sources, the need for rapid energy injection during power network disruptions, and the requirement for load voltage ride-through capability further complicate the design and operation of AI factory power systems.

The Vision: New Power Architecture

To meet these challenges head-on, GE Vernova is pioneering new power architecture for AI factories. This architecture is built on five main elements:

1. Grid Infrastructure: High-voltage to medium-voltage transformers, breakers, switchgear, and protection and control systems that enable seamless interconnection from the point of power generation to the AI factory.
2. Generation: A mix of baseload, standby, and black start generation capabilities, including gas turbines - CO₂ capture ready, renewable sources, and small modular reactors for a low-carbon energy strategy.
3. Stability Solutions: A combination of hardware and software solutions, including Battery Energy Storage Systems, advanced STATCOMs with supercapacitor energy storage, and digital platforms for grid orchestration and management.
4. Energy Management System: High-resolution, low-latency systems that empower operators to monitor, control, and improve energy consumption in near real-time.
5. Bulk Grid: The main grid supply point that provides power at a capacity and voltage level negotiated based on the AI factory’s requirements.

In collaboration with key data center stakeholders, GE Vernova has developed three reference designs to guide the deployment of AI factories: grid-connected, islanded, and bridging power. These standardized designs enable faster deployment, greater scalability, and improved reliability, ensuring that the demanding requirements of AI workloads are met.

Looking Ahead: 800 VDC Distribution

As we look towards the future of AI factories, it's important to consider the evolving landscape of power distribution with the escalation of power up to 1 megawatt per server rack. While traditional alternating current systems have served us well, the dynamic and high-density power demands of AI workloads are pushing the boundaries of what these systems can handle, especially in terms of plot, footprint, and energy efficiency. In response, we are engaging with the transition to 800 VDC distribution systems, and exploring even higher-voltage optionality, a shift that is fundamental to powering the rising demands of AI workloads. GE Vernova is at the forefront of this transition, leveraging its experience in designing and delivering higher-voltage direct current distribution systems from other industrial applications – a must-have for meeting the energy efficiency challenges of electro-intensive industries. This shift to DC architecture offers several strategic and operational benefits, including global flexibility, alternative semiconductor and magnetic materials, and space efficiency. While this transition requires careful planning and validation, it promises to enhance efficiency, scalability, and resilience in AI factory power infrastructure. As we continue to innovate and collaborate with industry stakeholders, we are committed to paving the way for the next generation of AI factories, ensuring they are powered by the most advanced and reliable systems available.

The Urgency for Action

The strategic urgency for adopting new power architecture in AI factories cannot be overstated. As the demand for AI-ready data center capacity grows at an average annual rate of approximately 33% between 2023 and 2030[3], the need for reliable, scalable, and high-performance power infrastructure becomes increasingly critical.

The future of AI is bright, but it requires a robust, innovative, and seamless “power-to-rack” architecture to support it. GE Vernova is committed to collaborating with the AI gigafactory ecosystem to enable the next generation of AI factories through rigorous validation, targeted design improvements, leadership in establishing industry standards, and ensuring that AI gigafactories can operate with maximum efficiency, resilience, and room for future growth.

For an in-depth look at the trajectory of AI factories, check out our latest whitepaper  “AI Factory: Reference Designs”.