Monday, October 13th 2025
NVIDIA Moves to 800 Volt DC Power Infrastructure to Drive Gigawatt-Scale AI
NVIDIA is implementing a major change in data center power architecture by using 800-volt DC distribution as the practical foundation for next-generation AI systems. Presented at the 2025 OCP Global Summit, the company's specifications and partner commitments frame 800 VDC not as an experimental idea but as a necessary response to racks that already draw well over 100 kW and are headed toward megawatt levels, soon to reach gigawatt of power per cluster. The core problem is simple and physical: low-voltage, high-current delivery forces bulky copper conductors and complex buswork into every rack, wasting space and money while creating thermal and installation headaches. Raising the distribution voltage reshapes those tradeoffs. At 800 volts, the same conductors can carry substantially more power, shrinking copper mass and freeing rack real estate for compute instead of power hardware.
NVIDIA's approach also simplifies the conversion chain by moving the main AC-to-DC rectification to the grid edge and using compact DC-DC modules in racks, reducing the number of conversion stages that generate heat and incur losses. Across large installations that efficiency improvement compounds. Even a few percentage points saved in delivery translates to megawatts less wasted energy at the gigawatt scale. Practical deployment plans tie the electrical advances to cooling and mechanical design: water-cooled busbars and integrated liquid cooling will be central to new rack designs that pack hundreds of GPUs while keeping temperatures and copper needs manageable.NVIDIA is contributing designs to the Open Compute Project and working with power-infrastructure and component manufacturers to standardize connectors, protection, and safety practices. The result is likely a split market: greenfield gigawatt facilities built around 800 VDC, and legacy sites that remain on lower-voltage systems until major refreshes. For AI's appetite for scale, the higher-voltage path looks like the more sustainable way forward. NVIDIA's Kyber rack-scale system is the first in the adoption path. With 800 VDC, the same copper design as previous NVL systems can deliver over 150% more power, removing the need for massive 200-kilogram busbars to supply a single rack.
Source:
NVIDIA
NVIDIA's approach also simplifies the conversion chain by moving the main AC-to-DC rectification to the grid edge and using compact DC-DC modules in racks, reducing the number of conversion stages that generate heat and incur losses. Across large installations that efficiency improvement compounds. Even a few percentage points saved in delivery translates to megawatts less wasted energy at the gigawatt scale. Practical deployment plans tie the electrical advances to cooling and mechanical design: water-cooled busbars and integrated liquid cooling will be central to new rack designs that pack hundreds of GPUs while keeping temperatures and copper needs manageable.NVIDIA is contributing designs to the Open Compute Project and working with power-infrastructure and component manufacturers to standardize connectors, protection, and safety practices. The result is likely a split market: greenfield gigawatt facilities built around 800 VDC, and legacy sites that remain on lower-voltage systems until major refreshes. For AI's appetite for scale, the higher-voltage path looks like the more sustainable way forward. NVIDIA's Kyber rack-scale system is the first in the adoption path. With 800 VDC, the same copper design as previous NVL systems can deliver over 150% more power, removing the need for massive 200-kilogram busbars to supply a single rack.



27 Comments on NVIDIA Moves to 800 Volt DC Power Infrastructure to Drive Gigawatt-Scale AI
This could have been done before, it is only done now because the energy usage is exceeding expectations - because AI, much like RT, is just raw crunching with no upper ceiling. Sky's the limit.
You should take a brief look at the tax rate on energy for larger capacities compared to what you and I pay on it.
www.michigan.gov/mpsc
Like I said, the laws of economics apply everywhere. You always pay for it.
> NVIDIA is contributing designs to the Open Compute Project and working with power-infrastructure and component manufacturers to standardize connectors, protection, and safety practices
Insert 16-pin embarrassment meme, but that aside, that'd be interesting to see in detail
That's some industrial voltage. I've heard stories of 'objects' practically dematerializing in dead shorts.
I wonder what extra safety and training has to be implemented, and if they're straight-borrowing from Industry that utilize HVDC?
Please show where these grants go to operating expenses.
Even if these grants were somehow offsetting electricity prices, the State of Michigan has a population of over 10 million people. So how far do you think a tiny $4.875M is going to offset the cost of power.
These are climate grants. Look where Michigan gets its power from.
So, yeah, government doing what a government is supposed to do. What’s the issue again? I showed you Michigan electricity rates are regulated by the state. Anything after that is you trying to protect your ego.
Increasing power usage is a net loss for anyone that isn't getting a kickback on the price repurcussions, regardless of the innovations made to enable it. HEV-styled high voltage DC cabling is cool and all, but it's being used to enable Colorado-drying levels of power draw instead of reducing whatever marginal losses are caused by unit-scale AC rectification and power delivery-related heat losses. I can't approve of that with any level of conscience.
Forcing 800V on everyone wanting NV-AI, looks to me like easiest "cash grab" to pull of (ie. making buyers get even more NV stuff).
AMD does it too, btw. It's not just Nvidia with massive power needs. They have a MI355X that can draw 1400W
The AI Bubble's Impossible Promises
www.wheresyoured.at/the-ai-bubbles-impossible-promises/
Speaking of which:
China's underwater data center facility could save 122 million kWh of electricity and 105,000 tons of water each year
www.techradar.com/pro/a-mini-data-center-village-under-the-sea-china-sinks-tens-of-thousands-of-powerful-servers-in-fresh-seawater-as-it-grapples-with-demand-for-more-power
I like Gary's example of AlphaFold: