TI and NVIDIA: 800V Tech for AI Data Centers
Source: cio.eletsonline.com
TI and NVIDIA Collaborate on AI Data Center Tech
Texas Instruments (TI) and NVIDIA are working together on power management and sensing technologies. The goal is to create high-voltage power systems for AI data centers.
This collaboration focuses on developing an 800-volt direct current (DC) power distribution system. This system will be important for powering NVIDIA’s future AI servers. The innovation seeks to provide better scalability, reliability, and energy efficiency for data centers as AI computing grows.
Gabriele Gorla, Vice President of System Engineering at NVIDIA, stated that semiconductor power systems are key for high-performance AI infrastructure. He added that NVIDIA is working with partners like TI to create an 800V DC architecture for AI-powered data centers.
Upgrading Data Centers
The move to 800V DC systems is a notable upgrade from the traditional 48V systems. Scaling 48V systems is becoming difficult; a single 48V system rack can need up to 205 kilograms of copper wiring, which limits scalability, adds weight, and affects thermal performance.
The new 800V architecture uses TI’s expertise in power conversion and sensing to deliver more power with less material and complexity. This will simplify data center design and improve the ability to handle increasing compute demands without needing more physical space.
Jeffrey Morroni, Director of Power Management R&D at Kilby Labs and a TI Fellow, noted that AI data centers are pushing power requirements higher. He stated that while 48V systems were once the focus, TI’s innovation and NVIDIA’s AI leadership are making 800V systems the future of data center power.
The new power distribution system should reduce power losses, improve energy use, and reduce the size of electrical systems in AI data centers. The collaboration shows a shift to high-voltage, high-efficiency infrastructure to support the power needs of AI technologies. The TI-NVIDIA partnership aims to set a standard for energy-efficient data center design.