🔍 Read the full analysis: A New Direction For AI Data Center Power: 800-Volt DC on ThorstenMeyerAI.com
TL;DR
Vantage reserved more than a gigawatt of factory capacity for Heron Power solid-state transformers that convert grid electricity to 800-volt DC. The move reflects a broader effort to serve higher-density AI racks and manage power demand, but efficiency and cost benefits cited so far are largely vendor claims. Grid connections, equipment lead times and who pays for new power infrastructure remain major issues.
Data center operator Vantage reserved more than a gigawatt of factory capacity for Heron Power solid-state transformers that convert grid electricity directly into 800-volt direct current, a development that highlights the infrastructure shift underway as AI racks demand more power. The reservation does not establish how many transformers will be delivered or when they will be installed; operators are also pursuing new generation and batteries to address electricity supply constraints.
The reservation, reported by ThorstenMeyerAI.com on October 27, is for manufacturing capacity, not a stated delivery of more than one gigawatt of operating power. Heron’s transformer is designed to convert grid power to 800-volt DC and includes energy storage to help buffer changes in data center demand. The source does not give a delivery schedule, final order value or the specific Vantage facilities involved.
A separate deal announced five days before the Vantage reservation links NiSource and Microsoft. The 20-year agreement adds about 1,100 megawatts of generation and a 600-megawatt battery, according to the source. It is an example of a power supply arrangement tied to data center demand, though the supplied material does not specify the generation technology, project timetable or how costs will be allocated.
The technical case for 800-volt DC is that it can reduce the number of power conversions between the grid and servers, while carrying more power through conductors than lower-voltage systems. NVIDIA’s reference design converts 13.8-kilovolt grid power to 800-volt DC near a facility and targets one-megawatt racks from 2027. NVIDIA claims 45% less copper than 415-volt AC distribution and up to 5% better end-to-end efficiency than current 54-volt systems. Those figures are company engineering claims, not independent measurements cited in the source.
Infographic 01 · From grid to rack
AI racks near a megawatt are outgrowing the AC power chain.
Why the industry is moving to 800-volt direct current: fewer conversions, less copper, batteries on the main path. The numbers are vendor figures with different baselines, so they cannot be added up.
Today: many conversions
Typical AC chain from the grid to the rack
- Medium-voltage AC from the grid
- Transformer steps it down to low-voltage AC
- UPS converts to DC for the batteries and back to AC
- Distribution units feed the racks
- Every server power supply converts AC to DC again
Each stage loses energy, and almost all of the loss becomes heat the cooling system must remove.
800 V DC: convert once, outside
NVIDIA reference design, full-scale production planned with Kyber racks in 2027
- 13.8 kV AC from the grid
- Converted once to 800 V DC at the edge of the site, batteries on the same DC bus
- 800 V DC busway to the racks
- Stepped down inside the rack, for 1 MW racks and beyond
Bridge for existing halls: a „sidecar“ cabinet next to the rack converts AC to 800 V DC, while the old chain stays in place.
Power lost per MW delivered to the rack
kW, SMA white-paper model vendor model
What NVIDIA says 800 V DC changes
NVIDIA’s engineering claims claim
Sources: NVIDIA technical blog (20 May 2025, updated 31 Jul 2025); SMA America, sponsored content in Utility Dive (26 Oct 2026). Rack density: IEA, Key Questions on Energy and AI (Apr 2026) (AI server power density 11× from 2020 to 2025, a further 4× by 2027).
Powered by Thorsten Meyer AI
Power Delivery for Megawatt Racks
The change matters because AI infrastructure is increasing the amount of electricity demanded at a single rack, putting pressure on equipment originally designed for far smaller loads. The International Energy Agency says AI server power density rose 11-fold between 2020 and 2025 and is projected to quadruple again by 2027. It estimates a rack could then draw as much power at peak as 65 households. These are projections and comparisons in the IEA material, not a description of every rack in operation.
Higher-voltage DC distribution could reduce conductor requirements and conversion losses, freeing space and potentially reducing the heat that cooling systems must remove. But the performance and cost gains depend on the system design and need to be tested in deployed facilities. Moving to DC also does not create electricity or remove the need for grid connections, substations, backup systems and cooling.
Batteries are part of the same planning problem. AI workloads can cause rapid shifts in demand, and storage may help a site manage those fluctuations or respond to grid disruptions. The IEA projects 20–25 gigawatts of batteries could be installed in data centers by 2030, while saying they could support the wider grid if incentives are in place. That is a forecast, not a confirmed buildout.
Popular right now
Breville Bambino Espresso Machine, Brushed Stainless Steel
Dyson Ball Animal Total Clean Upright Vacuum
Roborock RockNeo Q110H Wire-Free Robot Lawn Mower
As an affiliate, we earn on qualifying purchases.
From AC Conversion to 800 Volts
Most data centers have traditionally received alternating current, stepped it down through transformers and power systems, and converted it to low-voltage DC near or inside servers. Each conversion can lose energy, much of which becomes heat. The proposed 800-volt approach moves conversion closer to the facility’s incoming power supply and distributes DC toward high-demand racks. Operators may also use a “sidecar” conversion cabinet beside a rack to support new equipment in halls built around AC infrastructure.
The scale of electricity demand is part of the backdrop. The IEA estimates global data center electricity use at 485 terawatt-hours in 2025 and projects roughly 950 terawatt-hours by 2030. Berkeley Lab estimated U.S. data centers used 4.4% of national electricity in 2023 and could account for 6.7% to 12% by 2028. Those figures refer to different regions and periods and are estimates or projections, not current readings from every data center.
Physical infrastructure can also take longer to secure than computing equipment. The source cites IEA estimates that transformers can take two to three years to arrive and grid connections five to ten years in many locations. Berkeley Lab reported that the U.S. interconnection queue held about 2,300 gigawatts at the end of 2024; only 13% of earlier requests were ultimately built. Queue capacity is not the same as power that is approved, financed or under construction.
„NVIDIA says its design can move 85% more power through the same conductor size.“
— NVIDIA, in its 800 VDC technical blog
Costs, Deliveries and Grid Impacts
Key commercial details remain unavailable. The source does not report the value of Vantage’s capacity reservation, the number of transformers covered, delivery dates or which sites would use them. It also does not say whether the NiSource-Microsoft agreement’s generation and battery are operational, under construction or at an earlier development stage.
The expected savings from 800-volt DC are not yet established here through independent, field-based measurements. NVIDIA’s copper, efficiency and maintenance figures are vendor claims. The Utility Dive analysis cited in the source is sponsored by SMA America, whose model compares power losses under specified configurations. Actual outcomes could vary by facility, workload, equipment and conversion design.
It is also unclear how much of the cost of new generation, storage and grid infrastructure will be borne by technology companies, utilities or other customers. The source cites U.S. household electricity prices averaging 18.21 cents per kilowatt-hour in 2026, compared with 17.30 cents in 2025, and a 41% rise in wholesale prices in the PJM region. Those figures do not by themselves establish that data center growth caused the changes or show how costs will affect individual bills.
Projects Must Reach Deployment
The next evidence will be whether the reserved transformer manufacturing capacity becomes equipment delivered and installed at data center sites, and whether 800-volt systems meet their stated efficiency and space targets in operation. NVIDIA has said full-scale production for its Kyber systems is planned for 2027. The source provides no comparable deployment date for the Heron-Vantage arrangement.
For the NiSource-Microsoft deal, the key milestones will include the development and commissioning of the generation and battery capacity, along with the terms governing the 20-year supply arrangement. More broadly, operators will need to secure grid connections and equipment before new power designs can support planned AI capacity. Details on project schedules, costs and customer impacts remain to be reported.
Key Questions
What was Vantage’s announcement?
Vantage reserved more than one gigawatt of factory capacity for Heron Power solid-state transformers designed to convert grid electricity directly to 800-volt DC. The source does not say how many units will be delivered or when.
Why are data centers considering 800-volt DC?
Higher-voltage DC can carry more power through conductors than lower-voltage distribution and may reduce conversion stages between the grid and servers. The expected efficiency and cost benefits cited so far include vendor claims that have not been independently verified in the source.
How do batteries fit into the design?
Batteries can store energy and may help manage rapid shifts in data center demand or grid disturbances. Heron’s transformer is described as having storage built in, while the NiSource-Microsoft agreement includes a 600-megawatt battery.
Does the Vantage reservation mean the equipment is already installed?
No. The reported development is a reservation of factory capacity. The source gives no installation date, named site or confirmation that the equipment is operating.
Will these deals lower household electricity bills?
The source does not establish that. It says electricity prices have risen and discusses uncertainty over who pays for new power infrastructure, but it does not attribute household price changes directly to these agreements or predict their effect on bills.
Source: ThorstenMeyerAI.com
