The Third Energy Revolution: Powering the AI Era

By Robert Taylor, General Manager of Power Design Services and Power Delivery, and Henrik Mannesson, General Manager of Energy Infrastructure at Texas Instruments.

A paradigm shift is unfolding before our eyes. In the 18th and 19th centuries, the United Kingdom harnessed coal to power the Industrial Revolution — the world’s first energy revolution — transforming manual labor into machine-driven manufacturing. The 20th century brought the second energy revolution in the United States, as oil fueled an unprecedented boom in automobiles and electricity.

Now, the rapid advancement of artificial intelligence (AI) is driving the third energy revolution — one defined by the energy required to power and process the massive volumes of data our digital world consumes. Generating, converting, and efficiently distributing this energy — from the grid to the processor gate — has become one of the most exciting engineering challenges of our time.

Redefining Power Distribution

As the computing demands of AI data centers grow exponentially, their power architectures are undergoing a fundamental transformation. Traditionally, servers in a rack are powered by a bottom-mounted power supply unit (PSU), which converts alternating current (AC) into 48V, then steps down to 12V before reaching the processors.

However, with the rise of generative AI and large language models (LLMs), power requirements have surged dramatically. For example, processing a single query in an LLM can consume ten times the energy of a typical search engine request. These escalating power levels are pushing conventional rack designs to their limits.

To address this, data center operators are exploring new architectures — moving AC-to-DC conversion out of individual racks entirely to free up valuable space for additional servers. One approach is the use of sidecar racks, which house PSUs adjacent to IT racks. The long-term vision, however, is a dedicated power room distributing high-voltage DC power throughout the server floor.

Tapping Solar Energy to Meet Growing Demand

As data centers consume ever more power to support complex AI workloads, renewable energy sources such as solar are emerging as key enablers. Solar power has become increasingly viable and cost-effective in many parts of the world.

Moreover, many data center customers have pledged to run entirely on renewable energy — commitments that must extend to the facilities they operate. Solar energy not only helps meet sustainability goals but also provides a scalable, deployable source of additional power.

At the heart of solar energy conversion are semiconductors. Advanced power conversion and precision sensing technologies are essential to make solar energy a stable, reliable contributor to the grid — and by extension, to data centers.

Energy Storage: Unlocking Solar’s Full Potential

While data centers operate around the clock, solar energy is only available during the day. To ensure continuous power availability, energy storage systems (ESS) play a vital role.

Batteries have become integral to modern grids, storing and discharging large amounts of energy as needed — and are now increasingly dedicated to data centers. Battery management systems (BMS) within ESS monitor individual cells, tracking voltage, charge status, and overall health to ensure stable, dependable power delivery.

Powering the Future of Data

In the age of AI, data is the new currency, and its value continues to grow. Just as coal once fueled factories and oil powered automobiles, renewable energy will sustain the digital engines of tomorrow — helping data centers meet the soaring energy demands of the AI revolution.