Black storage and processing unit, large AI processing server room, data center storage and network processing room, SSD and hard disk drive storage, and processor cooling system,3D rendering.
04 Nov 2026

As AI Racks Move Toward Megawatt-scale Power, the way Data Centers Distribute Electricity Needs to Change

For years, data-center power architectures have been designed around relatively predictable rack densities. AI is changing that equation.

The increasing power requirements of GPUs and accelerated computing are pushing traditional power distribution architectures toward their practical limits. As a result, the industry is beginning a significant transition toward 800 VDC distribution for next-generation AI data centers.

So, why 800 VDC, and what does this change mean for the equipment that will power tomorrow's AI infrastructure?

Why Higher Voltage Matters

The principle is straightforward: for the same amount of power, increasing voltage reduces the current required.

Delivering 1 MW at 800 VDC requires approximately 1,250 A. At 54 VDC, the same power would require approximately 18,500 A.

That difference becomes increasingly important as AI rack densities grow. Lower current can reduce conductor and busbar requirements, copper usage, electrical losses and the amount of heat that needs to be managed. Higher-voltage DC distribution can therefore make delivering hundreds of kilowatts, and potentially megawatts, to high-density computing infrastructure considerably more practical.

It's More Than a Change in Voltage

The bigger shift is actually the power architecture around it.

Today's data centers may involve several electrical conversion stages before electricity ultimately reaches the processors. An 800 VDC architecture provides an opportunity to simplify that journey by distributing high-voltage DC closer to the computer’s equipment and converting it to the lower voltages required by the electronics nearer the load.

The result can be a simpler, more compact and more efficient electrical infrastructure, better suited to the rapidly increasing power density of AI computing.

But there is another side to this transition.

New Architecture, New Compliance Considerations

Moving toward 800 VDC changes the electrical environment in which data-center power equipment operates.

Busways, switchgear, power conversion equipment, protection devices, insulation systems, connectors and integrated power modules all need to perform safely and reliably at these higher DC distribution voltages.

Existing standards already provide important foundations. For example, UL 857 and CSA C22.2 No. 27 address busway systems up to 1,000 V, while IEC 61439-6 covers busbar trunking systems up to 1,500 VDC. Evaluation can include temperature rise, dielectric withstand, short-circuit performance, grounding and bonding, mechanical performance and environmental considerations.

The challenge also extends beyond individual products. As AI data centers increase in scale, the interaction between medium-voltage switchgear, transformers, low-voltage distribution, UPS, BESS, controls, modular power blocks and onsite generation becomes increasingly important. In effect, these facilities are beginning to resemble small utility-scale power systems rather than conventional IT installations.

Preparing for the Transition

This is where considering safety, performance and compliance early in the design process can make a significant difference.

As 800 VDC architecture develops, manufacturers and data-center stakeholders may need to consider not only whether an individual component meets the relevant requirements, but how that equipment performs as part of a much larger electrical ecosystem.

Intertek can support that journey from early design assessment through testing, certification and system-level validation. This can include evaluating busways and power-distribution equipment against relevant North American and international standards, assessing short-circuit, dielectric and thermal performance, and reviewing how switchgear, transformers, UPS, BESS, and other elements operate together within modular power systems. For larger installations, support can extend into factory acceptance testing, protection and relay validation, commissioning readiness, and the documentation needed to support utility and Authority Having Jurisdiction acceptance.

Taking this broader approach can help identify potential compliance or performance challenges earlier, when they are generally easier to address, while providing manufacturers and project stakeholders with a clearer pathway from design to certification and ultimately energization.

Conclusion

800 VDC is not simply about increasing voltage. It represents a fundamental change in how power could be delivered to the next generation of high-density AI infrastructure.

As computing moves toward megawatt-scale racks, data-center power distribution needs to become more efficient, compact and scalable. At the same time, higher-voltage DC brings new considerations around electrical safety, fault protection, insulation, thermal performance and system integration.

For manufacturers and data-center stakeholders, the opportunity is to address these requirements while the technology and architecture are being developed, rather than waiting until the point of certification or installation.

That early understanding of safety, performance and compliance could prove just as important as the move to 800 VDC itself

Pierrick Balaire headshot
Pierrick Balaire

Global Business Director

Pierrick Balaire is a Global Business Director specializing in industrial machinery and energy with comprehensive expertise including ASTA certification (LV to HV type test) and strategic business development. He develops tailored plans for assigned industries, focusing on portfolio optimization and profitable growth while collaborating with key functional leaders on competitive positioning and new business initiatives.

You may be interested in...