AI×Energy: Unlocking a New Power System

2026-10-26

AI×Energy: Unlocking a New Power System

We have seen that in many fields, AI brings the anxiety of being replaced, such as coding or autonomous driving. But in our power industry, AI is creating tremendous demand. So what we should really focus on is not the risks of being replaced, but how to turn uncertainties into opportunities.

The exponential growth of AI is heavily reliant on five critical power elements: robust installed capacity, a resilient power grid, abundant green energy, affordable electricity tariffs, and quick access to electricity.

Recently, a wideband oscillation triggered by AIDC in Shanxi Province caught our attention. It serves as a clear signal for us to do more for our system stability going forward.

So we need deep collaboration with AIDCs across planning, construction, dispatch and market.

At the same time, AIDC is not only a troublemaker, but also a huge potential driver for our growth, which comes from two aspects: increased power consumption, and monetization via tokens.

Beyond supplying power, we can realize monetization via tokens, leveraging the five critical elements mentioned above and two additional competitive strengths, including OPGW telecommunication and asset-heavy operation experience.

The strategy is simple—play to your strengths:

Generators should focus on large scale AIDC: Move the data to the power side.

Grid Operators should focus on edge AI computing: Power distributed data centers for time-sensitive services.

Of course, it is necessary for us to build strong customer-facing, digital operation capabilities. 

AI×Energy: Unlocking a New Power System

Although the fundamental physics of this industry remain unchanged, the game — the production and operation model of power systems — will be totally changed when our century-old power system meets an era of renewables and intelligence.

Everyone and every organization across the industry will feel the impact of this trend. And AI has become a survival essential for the future power system. We firmly believe AI will reshape the whole process from power generation to consumption, reshape full-life asset cycles, reshape customer services, and finally reshape our entire management system.

AI×Energy: Unlocking a New Power System

Two years ago at the opening ceremony of the CIGRE Paris session, an IEA speaker remarked that grid infrastructure is becoming the key bottleneck for the global energy transition.

Today, everyone is talking about AI. In my perspective, we need to pay more attention to the low-voltage side—and it's also AI’s main arena. Let me show you what's happening there.

In Europe, E.ON has been positioning prosumers as the backbone of the energy transition.

In China, distributed PV reached 530 gigawatts last year and will exceed 1,000 gigawatts by 2030.

Globally, about 700 million people still live without access to electricity.

These challenges stem mainly from the low-voltage side. To bridge this century-old gaps, we need to make the low-voltage side transparent to our management to ensure it is observable, measurable, controllable, adjustable, tradable, and traceable.

Low-voltage management must become systematic to ensure long-term sustainability.

First, production relations are critical. Professional capabilities still matter, but how do we make our expertise to be flexibly assembled to meet various business needs. The operating mechanism is just as important — how people collaborate, how responsibilities are defined, and how incentives are aligned.

Next, the communication target network is key. Without communication, nothing will happen—and we must solve two major challenges: low-voltage transparency and medium-voltage convergence.

Third, we need to move from traditional vertical, silo-based architecture to a layered, decoupled, horizontal one that supports lego-style modular composition and on-demand cloud-edge-device collaboration.

Finally, based on real-time, high-quality data, AI is deeply integrated into three business flows and innovation.

AI×Energy: Unlocking a New Power System

Huawei Kirin 9050 Pro, launched on September 7, is the world's first commercial flagship chip built on Tao's Law.

The Tao’s Law marks a paradigm shift — from shrinking geometry to shrinking time. From "making things smaller" to "making things faster." Speed is the ultimate goal. Logic folding, full-stack optimization, and system restructuring are the keys.

The direction of low-voltage management is simple: to get high-quality data faster with higher cost efficiency.

For vertical folding, in the past, meter data went through over 20 steps to reach the main station, spanning data acquisition, transmission, storage, and application. That chain is long. It usually takes more than 24 hours.

First, make the flat three-dimensional. Real-time and event-driven services close the loop locally. Non-real-time services get a clear policy. And we decide what actually goes up to the main station.

Second, go from serial to parallel. Push the relevant data conversion, protocol parsing, and computation to the edge. There is no need to send everything back to the main station. We can handle it locally at the nearest edge node. That way, massive concurrent streams flow fast, and critical data reaches the main station quickly.

Third, let data self-govern and self-heal. Quality checks move upstream, and anomalies are handled locally. Data is clean when it's created, and trustworthy when it flows.

Fourth, keep optimizing protocols, standards, processes, and rules for smooth data transfer.

For horizontal folding, with ontology modeling and a multimodal data lake, we break down silos and connect time-series, object, and knowledge vector data across domains.

AI×Energy: Unlocking a New Power System

From an application architecture perspective, six building blocks make up the system: IoT management, data platform, enterprise data exchange space, data management, AI platform, and security. I'll highlight just two, data management and security.

We should focus on the quality and operations of data flow from end to end.

And also pay close attention to massive connectivity risks in low-voltage scenarios. Optimize zoning. Enable logical multi-tenant isolation. A local problem should never become a global collapse.

AI×Energy: Unlocking a New Power System

At the same time, we can observe that low-voltage intelligence requires six core technologies.

First, a reliable low-voltage communication network is the cornerstone. Powered by MIMO, Bitloading, and OFDMA, super PLC enables 99.9% reliability. In field tests, reading 500 meters under one distribution transformer takes less than one minute, and distributed PV control completes within seconds. Get it right the first time. Unfortunately, too many system-level resources wasted on errors.

Second, super PLC also supports automatic low-voltage topology identification, eliminating the need for manual meter archive management and laying a solid foundation for full-scenario operations.

Third, a wireless private network is necessary for us to support 99.99% reliability in medium-voltage backhaul communication, addressing the grid’s weakest point.

Fourth, cloud-edge collaboration enables edge computing units to download required apps from the cloud, supporting distributed PV, charging piles, three-phase imbalance and reverse power flow management. With AI, it advances line-loss management, load forecasting and transformer district autonomy.

Fifth, on-premise cloud supports northbound enablement in five areas: R&D, app, data, AI, and integration. In the southbound, it enables synergy in four aspects: application, data, AI, and O&M. In the eastbound, it empowers digital and intelligent operations. In the westbound, it ensures business continuity management and trustworthiness.

Finally, with massive low-voltage connections, we must move from the “coconut-shell” security model to the “onion-like” layered defense model, supported by a dynamic Security Operations Center, to keep cybersecurity risks away from the core business.

AI×Energy: Unlocking a New Power System

Today, our biggest challenge is making AI agents work in core production scenarios.

We're tackling this through real-world projects like losses reduction. Working closely with our partners, we solve key issues around AI computing servers and model fine-tuning, and we pack these tools and capabilities into our Phecda platform.

Phecda matches each agent's inference profile and offers tailored suggestions—like deterministic reasoning. This helps us quickly find the best balance across cost, speed, and accuracy, so agents can run faster and deliver better results.

Finally, industry know-how is hard to come by. That's why we're open-sourcing Phecda platform. We want to empower our partners and work together to help power companies roll out power AI agents at scale.

AI×Energy: Unlocking a New Power System

Here, I’d like to highlight five key points, talent, ecosystem, standards, enterprises, and regulation.

Point 1: Agents, skills, and data now become indispensable to the power industry. To achieve sustainable development, digital and intelligent capabilities must be built in-house.

In order to bridge this talent gap and align with the “digital talent first” strategy, we launched the Wittgenstein Studio talent development program with two goals:

First, shift mindsets to explore uncertainty and inspire innovation. Our industry is built on a strong engineering culture, which is crucial to the industry yet a little bit conservative in its nature. This is not conducive to innovation.

Second, lower the threshold to digitalization and intelligence, to spark creativity among engineers and cultivate them from passive response into proactive mini-CEOs.

Point 2: In the era of digitalization and AI, no one can win alone. Huawei focuses on what we are good at and thriving with ecosystem partners to bring industry expertise and cross-domain capabilities together, better serving the power sector.

Point 3: We need open, consistent, forward-looking standards to drive innovation, industrial upgrading and ecosystem prosperity with a three-layer framework: metrics, architecture and technology.

First, metrics and evaluation unify industry understanding and steer high-quality business development.

Second, an open architecture supports plug-and-play, ensuring cost sharing and seamless heterogeneous combination while avoiding vendor lock-in.

Third, technology and product standards keep advanced technologies accessible and evolvable.

Point 4: Digital infrastructure first for enterprises. They are so many complicated things waiting for us, such as how to address the challenges efficiently, how to keep Spanish blackout away from our business, how to seize the new opportunities quickly, and how to bridge the gaps between our strategies and core capabilities. This is why I firmly believe that investing 5% of the total utility revenue into digital infrastructure is necessary for the long-term sustainability.

Point 5: Digital RAB first for regulators. Regulators should include digital infrastructure within regulated asset base. According to IEA reports, the lifespan of power equipment varies widely. Although the lifespan of the digital infrastructure is shorter, it can deliver significant value to inspire innovation and accelerate energy transition.

Disclaimer: The views and opinions expressed in this article are those of the author and do not necessarily reflect the official policy, position, products, and technologies of Huawei Technologies Co., Ltd. If you need to learn more about the products and technologies of Huawei Technologies Co., Ltd., please visit our product pages or contact us.

Inicio