Power Integrations, a US based semiconductor manufacturer specializing in power conversion technology, has announced a partnership with Nvidia to help advance the transition of artificial intelligence data centers toward 800 volt direct current power systems.
The collaboration is part of Nvidia’s larger initiative to make massive computing facilities more energy efficient and reliable as global AI demand continues to grow.
Most of today’s large scale data centers operate using alternating current power distributed from the grid. That power is then converted to direct current at the server level, a process that leads to energy losses, extra heat generation, and higher operational costs. Nvidia’s new strategy seeks to simplify this by distributing 800 volt DC power directly within the data center, reducing the need for multiple conversions and improving efficiency.
Power Integrations, which reported annual revenue of about $419 million in its latest fiscal year, designs chips that regulate and convert electrical energy efficiently.
The company’s technology relies on gallium nitride (GaN), a material that offers better performance and efficiency than traditional silicon.
GaN components can operate at higher voltages and frequencies while generating less heat, making them well suited for advanced computing environments. The firm’s automotive grade reliability helped it earn a place on Nvidia’s list of approved suppliers.
That list also includes major competitors such as Infineon Technologies and InnoScience, both of which are developing their own high voltage power solutions for the growing AI infrastructure market.
Roland Saint Pierre, vice president of product development at Power Integrations, said that the company’s success in the automotive sector played a key role in its selection.
The fact that we are automotive qualified, and we already have GaN devices running in cars, was a major factor for Nvidia, he said. “They appreciate the level of reliability and testing that comes with that qualification.”
Industry experts believe this collaboration highlights a broader shift within the technology sector toward more efficient energy delivery systems.
High voltage DC architectures are becoming critical as AI workloads scale up, said Dr. Emily Harris, a senior analyst at Global Tech Insights.
The power requirements for advanced GPUs and AI clusters are skyrocketing, and conventional AC systems can’t keep up without significant losses.
Gallium nitride chips are becoming increasingly popular because they enable compact and efficient power systems.
By switching at higher frequencies, GaN based transistors reduce the size of supporting components like transformers and cooling systems, lowering both cost and space requirements.
Traditional AC based systems often suffer energy losses exceeding 10 percent due to multiple conversion stages. Centralized 800 volt DC systems, however, can reduce those losses dramatically, achieving efficiencies near 98 percent according to engineering estimates.
The shift also reduces the need for heavy copper cabling and bulky transformers, freeing up space inside server racks and improving cooling performance.
Even a few percentage points of efficiency improvement can translate into millions of dollars in annual energy savings for hyperscale data centers that consume as much electricity as small cities.
“Even a modest gain in efficiency has an enormous financial and environmental impact at the scale these facilities operate,” said Aaron Patel, an energy systems consultant at DataGrid Engineering.
The 800 volt DC approach represents a long term investment in sustainability and performance. Data center operators are cautiously optimistic about the change but stress the importance of reliability.
The idea is solid fewer conversions mean less wasted energy, said Carlos Reyes, a facility manager for a major cloud provider in Virginia. “But any shift in the core electrical design of a data center takes years of testing before full adoption.”
Manufacturers in Asia are already preparing for the shift. “We’ve started seeing growing demand for components that support 800 volt configurations,” said Mei Chen, a product engineer at a Taiwanese server rack company.
Nvidia’s leadership in this space is accelerating the entire industry’s transition. A California based systems engineer, who requested anonymity, said the design could also reshape data center layouts.
It’s not just about efficiency, he said. “Using 800-volt DC could allow more compact rack designs, reduce heat output, and improve GPU density which is critical for AI processing.”
Nvidia’s long term goal is to transform traditional data centers into “AI factories” highly efficient, scalable facilities that can support next generation AI workloads. Moving to higher voltage DC infrastructure is an essential part of that plan.
Power Integrations, with its gallium nitride technology and proven reliability in demanding industries, is expected to play a key role in this transition.
Although the company has not provided a specific timeline for when its chips will be deployed, it confirmed that development and testing are ongoing.
Experts predict the adoption of 800 volt DC power systems will take several years, citing challenges such as new safety standards, compatibility with existing AC infrastructure, and the need for large scale testing. Still, momentum is clearly building.
Data centers are approaching the limits of what current AC systems can deliver, Dr. Harris said. “Switching to high voltage DC is one of the few ways to meet future performance and efficiency goals.”
The collaboration between Power Integrations and Nvidia marks a significant step toward reshaping how data centers manage power in the AI era.
By integrating gallium nitride technology into Nvidia’s 800 volt DC ecosystem, both companies aim to create more efficient, compact, and environmentally responsible computing environments.
If successful, the initiative could redefine the global standard for power distribution in next generation data centers, paving the way for a more sustainable and energy efficient digital future.