Google, Microsoft, and Nvidia are collaborating to establish an 800V direct current (800VDC) standard for high-density AI data centers through the Open Compute Project (OCP). This initiative aims to streamline power delivery systems while reducing operational costs amidst increasing power demands driven by AI workloads, particularly from GPUs.
The 800VDC standard seeks to alleviate the complexity of existing infrastructures, which struggle to accommodate the substantial power required by modern AI applications. By shifting from alternating current (AC) to direct current (DC), the infrastructure can optimize power delivery. A typical DC system requires fewer wires than AC, resulting in significant savings on wiring costs and a reduction in heat generation, which enhances cooling efficiency.
This transition could yield substantial reductions in resource use; data centers might observe a 50% to 80% reduction in copper requirements and an 8% to 12% decrease in annual energy expenses due to minimized conversion and distribution losses. For facilities adopting AI at a larger scale, this change may lead to capital expenditure savings of $4 million to $8 million for every 10 MW of power capacity.
The rise in energy consumption is swiftly outpacing the capabilities of conventional low-voltage AC systems. Higher voltage DC allows for higher power density configurations with a reduced need for extensive conductor infrastructures, which the industry is converging toward due to its efficiency.
In 2025, Google, Microsoft, and Nvidia began presenting their findings on the 800VDC approach and are now focused on creating comprehensive specifications to be widely adopted in the data center supply chain. They are currently developing a workstream dedicated to converting medium-voltage AC power into 800VDC while also harmonizing standards for power quality and system interfaces.
The open initiative promises significant changes for power equipment manufacturers, who could standardize their designs, enabling more efficient production and deployment across multiple data centers, much like server manufacturing. This development underscores a shift towards establishing a unified electrical foundation for AI processing demands before the anticipated spike in system density necessitates custom solutions from each provider.
Notably absent from this initiative is Amazon Web Services (AWS), which has yet to commit to the 800VDC standard or outline any parallel strategies. However, it is reportedly redesigning its AI data centers to accommodate denser systems, as demonstrated by its internal Titus initiative, which aims to upgrade infrastructure for next-generation AI technologies.
In summary, the collaboration between Google, Microsoft, and Nvidia marks a significant stride towards modernizing data center power systems, a necessity driven by the burgeoning demands of artificial intelligence technologies.
