The rapid expansion of artificial intelligence has created an urgent challenge for data center operators: power. As AI accelerators grow more powerful and racks become more densely packed, the traditional electrical infrastructure that has supported data centers for decades is reaching its limits. Google, Microsoft, and Nvidia are now pushing for a new common approach, supporting an 800-volt direct current (800VDC) standard through the Open Compute Project (OCP).

The three companies are working with the broader OCP ecosystem to define shared requirements for 800VDC power distribution, including power conversion, power quality, system interfaces, and safety. The goal is to create an open, industry-wide standard that can handle the power demands of next-generation AI data centers while simplifying infrastructure and reducing costs.

Key facts

• Google, Microsoft, and Nvidia are supporting 800VDC as an open standard through the Open Compute Project.
• The standard addresses AI-driven power consumption, particularly the high draw of GPUs.
• 800VDC could cut copper usage by 50 to 80 percent and reduce annual energy-related operating expenses by 8 to 12 percent.
• AI-first facilities could save 4 million to 8 million dollars in capital expenses per 10 MW build.
• Amazon Web Services has not committed to 800VDC, but is redesigning its data centers for higher density. Its internal Titus initiative is focused on next-generation infrastructure for Nvidia GB200-class systems and future Vera Rubin hardware.

The power problem driving the shift

AI workloads require levels of compute density that were uncommon just a few years ago. GPUs and other accelerators draw enormous amounts of power, and they are being deployed in configurations that strain existing electrical systems. Traditional data center power is based on alternating current (AC), distributed at lower voltages, with multiple conversion stages and significant infrastructure overhead.

In an AC system, power must be transformed, switched, and converted multiple times before it reaches servers. The losses from these conversions add up, especially as power demand climbs. In a high-density AI environment, the inefficiencies become even more pronounced. Racks that once consumed a few kilowatts now demand tens or even hundreds of kilowatts, and future systems will push that number higher.

Direct current (DC) offers a different path. By keeping power in DC form and distributing it at a higher voltage, operators can reduce conversion losses, lower current, and simplify the electrical path from the grid to the servers. The result is a more efficient and cost-effective power chain.

Why 800 volts?

Higher voltage allows more power to be delivered with less current. Current is the element that generates heat in conductors, so lowering current reduces thermal stress on cables and components. 800VDC has emerged as the industry convergence point because it can move substantial power while reducing the amount of copper and conductor infrastructure required.

Compared with traditional AC distribution, 800VDC uses fewer wires. An AC system typically requires four wires, while a DC system needs only two. That difference has a major impact on the cost and complexity of data center construction. Less copper means lower material costs, reduced weight, and easier installation.

The change also affects day-to-day operations. Lower current means less heat, which can reduce the burden on cooling systems. Data centers that run on 800VDC can therefore run cooler and more efficiently. This makes the standard particularly attractive for AI facilities, where power and cooling are among the largest operational costs.

Financial impact

The savings from 800VDC go beyond technical convenience. According to the companies involved, the standard could reduce copper usage by 50 to 80 percent. For a one-gigawatt data center, that is potentially millions of pounds of copper wire saved.

Annual energy-related operating expenses could also fall by 8 to 12 percent through lower conversion and distribution losses. These savings are significant in an industry where power costs are rising. For an AI-first facility, the capital expense savings are notable as well. Projects could see 4 million to 8 million dollars in savings per 10 MW build by reducing upstream AC infrastructure.

These financial benefits are likely to accelerate adoption. Data center operators have been under pressure to contain costs while scaling capacity for AI. A standard that lowers both capital and operating expenses, while improving performance, is difficult to ignore.

Work through the Open Compute Project

The effort is being coordinated through OCP, a community that designs and shares open data center technologies. Google, Microsoft, and Nvidia began presenting their 800VDC work through OCP in 2025. Since then, the effort has moved from demonstrating technical feasibility to establishing specifications that can be adopted across the data center supply chain.

The companies have initiated an OCP workstream focused on converting medium-voltage AC power directly to 800VDC. They are also aligning requirements that cover power quality, power smoothing, and end-to-end system interfaces. This collaborative approach is intended to ensure that different vendors build compatible systems.

One of the most important aspects of this work is standardization. Instead of each hyperscaler designing a proprietary power system, equipment manufacturers can build products to a common specification. This is similar to how servers have been standardized in the OCP ecosystem. A common power standard could reduce fragmentation and speed up deployment of new AI infrastructure.

Supply chain implications

The move toward 800VDC could have far-reaching implications for the data center supply chain. Power equipment makers, electrical contractors, and component suppliers will need to adapt to the new standard. But the open approach means they will not have to design separate systems for every major customer.

For manufacturers, a common spec lowers engineering costs and allows for economies of scale. It also makes it easier to develop standardized products that can be deployed by multiple operators. This is a meaningful shift from the current environment, where power distribution systems are often customized for individual facilities.

Standardization also encourages innovation. With a clear target specification, vendors can focus on improving efficiency, reliability, and cost. The open nature of OCP means proposals are transparent, and participants can contribute to the development of the standard.

AWS and the competitive landscape

Notably absent from the 800VDC initiative is Amazon Web Services (AWS). The company has made no commitment to 800VDC and has not announced its own comparable power distribution plan. That absence is significant because AWS is one of the largest data center operators in the world. Its choices have historically influenced the broader industry.

But AWS cannot ignore the growing power density challenge. AI accelerators are forcing every hyperscaler to rethink how data centers are built. AWS is already aggressively redesigning its AI data centers for much higher-density systems. Its internal Titus initiative is reportedly focused on next-generation infrastructure capable of supporting Nvidia GB200-class systems and future Vera Rubin hardware. The initiative is also aimed at increasing data center capacity while improving power and cooling efficiency.

The fact that AWS is not part of the open 800VDC effort does not mean it is idle. It may be developing proprietary approaches, or it could be waiting for the OCP specification to mature before making a commitment. Given the scale of AWS, its eventual choice will have a major impact on the future of data center power.

Broader industry momentum

OCP has a track record of turning collaborative engineering into widely adopted standards. Its work on server hardware, storage, and networking has reshaped the data center industry. The 800VDC initiative is a natural extension of that mission, applying open design principles to the electrical backbone of the data center.

The involvement of Google, Microsoft, and Nvidia adds significant weight to the project. These companies occupy critical positions in the AI ecosystem, from cloud platforms to accelerator hardware. Their shared support for 800VDC could push the standard into the mainstream.

There are also environmental implications. Less copper usage reduces the material footprint of data center construction. Lower energy losses mean less wasted electricity, which is important as AI drives overall data center power consumption upward. More efficient power distribution could help operators meet sustainability targets while supporting continued expansion.

Technical and operational considerations

Transitioning to 800VDC is not without challenges. Existing data centers were built around AC distribution, and retrofitting them is difficult and expensive. New construction, however, can adopt 800VDC from the start. The greater opportunity lies in building AI-focused facilities with the new standard from day one.

Safety is another consideration. DC voltage at 800 volts requires robust protection systems, proper connector designs, and careful maintenance procedures. The OCP workstream includes safety as one of its key focus areas, along with power quality and system interfaces. These technical details will determine whether the standard can be deployed reliably at scale.

Power conversion efficiency is also critical. Moving from medium-voltage AC directly to 800VDC reduces the number of conversion steps. Each step introduces losses, so fewer steps mean higher overall efficiency. The ongoing work is aimed at optimizing these conversions and ensuring that power quality remains stable under varying loads.

Energy storage is another area where 800VDC could be useful. Data centers are increasingly integrating batteries and other storage systems to manage peak demand and backup power. A DC-based architecture could simplify the connection of storage assets, since batteries operate on DC and can be integrated with fewer conversions.

What comes next

The OCP 800VDC initiative is still taking shape. Specifications must be finalized, prototypes must be tested, and the broader supply chain must adjust. But the direction is clear: high-density AI data centers need a more efficient power standard, and the industry is converging around 800VDC.

For Google, Microsoft, and Nvidia, this is an attempt to create a common electrical foundation for AI factories before increasingly dense systems force every hyperscaler and equipment supplier to develop its own solution. The collaborative approach reduces risk and ensures interoperability across the ecosystem.

As AI workloads continue to grow, power infrastructure will become as important as compute performance. The 800VDC standard represents a major step toward a more efficient, scalable, and sustainable data center power architecture. By working together through OCP, the companies are attempting to set the course for the next generation of AI infrastructure.

Source:Network World News