Telecom’s DC power legacy can offer a blueprint for AI data centres: Vertiv’s Greg Funk
Funk said data centre operators can draw on benefits of efficiency and reliability as AI concentrates more computing power into smaller spaces.

Greg Funk, vice-president of modular power converters at Vertiv
Vertiv supplies power, cooling and IT infrastructure equipment and services for data centres and communications networks.
“We still kind of see them (telecom networks and AI data centres) as separate businesses, but the technology is certainly converging,” Funk told ET AI during the India Mobile Congress (IMC) 2026. He said telecom’s progress towards greater efficiency, higher power density and voltages provide lessons for data centres, where AI has accelerated the need for change.
Lessons from telecom
Traditionally, Funk said, data centres built their infrastructure around alternating current (AC), while telecom networks focused on DC systems.
DC flows in one direction, while alternating current reverses direction periodically. DC systems simplify battery and energy storage connections by linking solar arrays and batteries to a shared DC bus, avoiding repeated conversions between DC and AC.
Telecom’s approach reduces the number of power conversions, helping improve both efficiency and reliability, Funk said. “So the efficiency and reliability naturally goes up,” he said.
Funk said data centre operators can draw on these benefits as AI concentrates more computing power into smaller spaces. He added that AI servers need more electricity in the same amount of space.
Cooling was the first challenge as AI chips generated more heat than air cooling can handle, prompting the move towards liquid cooling, Funk said. He said power delivery becomes the next constraint as servers pack in more performance.
“Copper can only get so large before we run out of space,” Funk said.
Need for higher voltage
Funk said the operators cannot indefinitely increase the amount of copper needed to supply powerful racks. Higher voltage allows them to deliver power with less current and therefore less copper, he said.
“Increasing the voltage reduces the current, which reduces the amount of copper required,” Funk said.
By utilising Ohm’s and Joule’s laws, increasing the voltage allows the system to push the same power using only a fraction of the current. This lower current generates far less electrical friction, letting energy safely glide through razor-thin copper wires instead of heavy, expensive cables.
Funk said the industry has focused on 800-volt DC as an approach to support denser AI racks. He added that increasing voltage starts to make sense as rack power requirements move towards 300 kilowatts, allowing operators to reduce the copper needed to supply a row of servers.
The transition could happen in stages, with operators converting selected rows that house their most demanding equipment rather than entire facilities, Funk said. He said a converter at the end of a row could feed an 800-volt power distribution system, which operators could expand as their needs grow.
“For a new facility, or a greenfield project, where the operator decides to use 800-volt DC, the question is how to deploy it quickly,” Funk said.
Funk said modular systems can help operators make these changes quickly, whether for upgrading infrastructure or building a new facility. He said Vertiv offers prefabricated blocks sized for different GPU deployments, which can simplify installation and repairs. This can bring operational gains along with the electrical efficiency improvements from changing voltages and reducing conversion stages, he said.
Moving conversion equipment outside the rack can also free space for more computing equipment, Funk said. He said graphics processing units (GPUs) work together like a large parallel computer and their performance partly depends on proximity.
Packing more chips into a smaller space raises power density but can improve performance per watt and dollar, Funk said.
Planning for the next generation
Existing data centres can adopt higher-voltage DC, although facilities that retain conventional systems may eventually face limits on density or efficiency, Funk said.
Funk said the operators are planning for power constraints expected from the new generations of chips based on roadmaps of the chipmakers. He added those constraints are not yet widespread in mass-produced systems. “We are trying to solve problems we believe are going to be coming in the future.”
Operators need to consider infrastructure investments before new generations of servers arrive, he said.
“The last thing they want is to be sitting on millions of dollars’ worth of GPUs that they cannot turn on,” Funk noted.
He said facility designers must consider both current and future electricity needs when choosing infrastructure.
“Everything starts at the load…You need to have your load in mind… as to what you might want to do today, or what you’re going to be trying to do tomorrow,” he added.
The Economic Times Business News App for the Latest News in Business, Sensex, Stock Market Updates & More.