According to Zhitong Finance APP, the stock price of ON Semiconductor (ON.US), a chip giant focused on the automotive and industrial sectors, rose more than 6% in pre-market trading on Tuesday, after the chip company announced a collaboration with &#34;AI chip leader&#34; NVIDIA (NVDA.US) to accelerate the development of 800-volt direct current (800V DC) power solutions for the next generation of artificial intelligence data centers.
At the core of this transformation is a new type of power distribution system that must deliver large amounts of electricity with minimal loss during each voltage conversion. ON Semiconductor stated that its smart power solutions are the most critical link in providing power supply for the next generation of AI data centers, achieving high efficiency and high power density in energy conversion at every stage.
As AI applications such as ChatGPT, Claude, and DeepSeek sweep the globe, the power demand of large-scale AI data centers worldwide is becoming increasingly enormous. The high-energy-consuming AI data centers, which are expanding exponentially due to the fierce demand for AI chips and other AI computing infrastructure, rely on the core foundation of power supply, which is the origin of the market view that &#34;the end of AI is power.&#34; A forecast report from the International Energy Agency (IEA) indicates that by 2030, the power demand of global data centers will more than double, reaching approximately 945 terawatt-hours (TWh), slightly higher than Japan's current total electricity consumption, with AI applications being the most important driver of this growth; it is expected that by 2030, the overall power demand of data centers focused on AI will quadruple.
The power consumption of global AI data centers is rapidly rising from the traditional 20-30 kW per rack to levels of 500 kW or even 1 MW. To deliver several times the current in the same cabinet space without overheating the busbars, cables, and converters, the industry is beginning to transition from 48V / 400V DC to an 800V high-voltage direct current (HVDC) architecture.
The 800V DC solution jointly announced by ON Semiconductor and NVIDIA is precisely this evolution. By raising the bus voltage of data center rooms from the currently common 48V or 380/400V DC to 800V DC, the current and I²R losses can be reduced to one-tenth at the same power level, significantly decreasing the number of busbars, cables, and conversion stages, thus improving overall power supply efficiency by at least 5 percentage points and leaving power space and heat dissipation margin for NVIDIA's &#34;AI factory&#34; with a single rack of 500 kW-1 MW in the future.
Currently, large-scale cloud computing and most AI training/inference cluster rooms still primarily use 48V DC (such as OCP Open-Rack) or 380/400V DC parallel power architectures, while 800V is just beginning to pilot as the next generation of high-voltage direct current (HVDC) solutions, led by NVIDIA, ON Semiconductor, and others, expected to be demonstrated between 2025 and 2027.
The mainstream power systems currently remain at 48V and some 380V DC; 800V can be said to be in the early adoption phase, targeting the future paradigm of 1MW racks. NVIDIA's single AI GPU server is approaching 1 kW, and a fully equipped NVIDIA NVL AI server cabinet easily exceeds 100 kW; NVIDIA plans to mass-produce 1 MW AI Factory AI server rack clusters starting in 2027, and only an 800V level high-voltage bus can keep the bus size and thermal loss within usable limits
The 800V bus is also easier to integrate with immersion/cold plate liquid cooling, mainly due to characteristics such as low current and fewer hot spots.
For the world's highest market capitalization technology giant, the first &#34;A-chip overlord&#34; NVIDIA with a market value exceeding $4 trillion, a new revenue stream may soon be added—after all, 800V DC high-voltage direct current distribution is likely to become the new power backbone for future megawatt-level AI data centers. This also means that after AI GPUs, high-performance network solutions for data centers InfiniBand, and automotive-grade high-performance SoC platform NVIDIA DRIVE Thor, NVIDIA will welcome a brand new and extremely strong demand growth channel.
For ON Semiconductor, the company is expected to provide 800V-level SiC MOSFETs, solid-state transformers, and high-density DC-DC modules, which are the core power devices/power OEMs of the entire solution. With NVIDIA's exclusive scalable ecosystem deployment, ON Semiconductor can directly secure long-term large orders and strengthen its competitive advantage in the high-voltage SiC power market against competitors like Infineon

NVDA

ON Semiconductor collaborates with NVIDIA to launch an 800V DC power solution aimed at meeting the growing power demands of AI data centers. This solution maximizes power conversion efficiency to support high power density AI infrastructure. It is expected that global data center power demand will double by 2030, with AI applications becoming a major driving force

- On Tuesday, ON Semiconductor's stock rose over 6% after announcing a partnership with Nvidia to develop 800V DC power solutions for next-generation AI data centers.  
- This collaboration aims to enhance power efficiency and density, addressing the growing energy demands of AI applications, which are projected to double global data center power needs by 2030.  
- The shift to 800V DC is expected to significantly reduce power loss and improve overall efficiency, positioning both companies for substantial growth in the evolving AI infrastructure market.

From GPU to Data Center Power Revolution: NVIDIA digs for growth points again, partnering with ON Semiconductor to layout 800V DC