Why 800-Volt DC Is Gaining Attention In AI Data Centers
AIThis post was created with the assistance of artificial intelligence (AI).

🔍 Read the full analysis: Why 800-Volt DC Is Gaining Attention In AI Data Centers on ThorstenMeyerAI.com

Before you orderOffer from Amazon

Get smart everyday buys delivered free with Prime

  • Fast, free delivery on millions of items
  • Prime Video, Amazon Music and more included
  • Member-only deals all year
Start your free Prime trial Free trial for eligible customers · Cancel anytime
As an affiliate, we earn on qualifying purchases.

TL;DR

Vantage reserved more than a gigawatt of factory capacity for a solid-state transformer designed to supply 800-volt DC, while NiSource and Microsoft signed a long-term power deal that includes new generation and a battery. The developments reflect growing interest in ways to serve dense AI racks and manage power swings, but efficiency and cost advantages cited by vendors have not been independently established in the source material.

Data center operator Vantage reserved more than a gigawatt of factory capacity on October 21 for a solid-state transformer designed to convert grid electricity directly into 800-volt direct current, a system being developed for high-demand AI facilities. The move, alongside a separate Microsoft-NiSource power agreement that includes a 600-megawatt battery, highlights how data center expansion is increasingly shaped by power equipment and grid access as well as chips.

Vantage’s reservation concerns manufacturing capacity for a transformer from Heron Power, according to the October 27 report by ThorstenMeyerAI.com. The source says the equipment is designed to convert grid power directly to 800-volt DC and includes energy storage to help buffer changes in data center demand. A reservation of factory capacity is not the same as a confirmed installation: the source does not identify a deployment site, delivery schedule or final installed capacity.

On October 16, NiSource signed a 20-year power agreement with Microsoft, the source reports. The deal adds about 1,100 megawatts of generation and includes a 600-megawatt battery. The figures describe the planned power arrangement; the material does not specify the battery’s duration, project location, construction timeline or how its costs will be allocated.

The technical case for 800-volt DC is that it could reduce the number of power conversions between the grid and AI servers, a topic explored in analysis of AI data center power bottlenecks. Data centers commonly distribute alternating current through the facility, then convert it to lower-voltage DC inside server equipment. NVIDIA’s reference design proposes converting 13.8-kilovolt grid power to 800-volt DC near the facility perimeter and distributing DC to racks designed for roughly one-megawatt loads. NVIDIA says the design could use 45% less copper and improve end-to-end efficiency by up to 5% compared with current 54-volt systems. Those are vendor estimates, not independent measurements, and full-scale production is planned for 2027.

At a glance
reportWhen: Developments reported October 16–27, 20…
The developmentA Vantage transformer capacity reservation and a NiSource-Microsoft power agreement have brought renewed attention to 800-volt DC systems and batteries for AI data centers.
Thorsten Meyer AI
Power architecture

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

  1. Medium-voltage AC from the grid
  2. Transformer steps it down to low-voltage AC
  3. UPS converts to DC for the batteries and back to AC
  4. Distribution units feed the racks
  5. 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

  1. 13.8 kV AC from the grid
  2. Converted once to 800 V DC at the edge of the site, batteries on the same DC bus
  3. 800 V DC busway to the racks
  4. 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

  • AC chain, conversion at the racktraditional design~80 kW
  • 800 V DC from outside the buildingmedium voltage converted once~20 kW

What NVIDIA says 800 V DC changes

NVIDIA’s engineering claims claim

−45%copper, vs 415 V AC distribution
+85%power through the same conductor size
≤5%better end-to-end efficiency vs 54 V racks
200 kgcopper busbar for one 1 MW rack at 54 V

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

AI computing is pushing rack-level electricity demand beyond the assumptions behind many existing data center designs. The source cites an International Energy Agency report saying AI server power density rose 11-fold from 2020 to 2025 and is projected to quadruple again by 2027. The agency’s projection says a rack could then draw peak power comparable to 65 households. These figures point to a practical constraint: operators need to move much more power into a limited space without allowing conversion losses and heat to grow at the same pace.

Higher-voltage DC could reduce the amount of copper needed to carry a given amount of power and simplify parts of the electrical path. Fewer conversion stages may also reduce energy lost as heat, although actual results will depend on system design and operating conditions. Batteries could help manage rapid load changes from AI workloads and support a site during grid disturbances. The IEA estimates data center batteries could total 20–25 gigawatts by 2030; it says they could also serve the wider grid if suitable incentives are in place.

The commercial stakes extend beyond equipment performance. The source says transformers can take two to three years to arrive and grid connections can take five to ten years in many locations, citing the IEA. Those delays can affect when planned computing capacity becomes usable. For households, the consequences of new power demand depend on who pays for generation, transmission and distribution upgrades. The source reports U.S. residential electricity averaged 18.21 cents per kilowatt-hour in 2026, compared with 17.30 cents in 2025, but that figure alone does not show how much data centers contributed.

Products Worth Considering

Dell XPS 13 9350 AI Professional Laptop 13.4 120Hz FHD+ Display

Dell XPS 13 9350 AI Professional Laptop 13.4 120Hz FHD+ Display

As an affiliate, we earn on qualifying purchases.

From AC Chains to DC Distribution

Most data centers have historically received alternating current and stepped it down through transformers, uninterruptible power supplies and distribution equipment. Server power supplies then convert it to low-voltage DC. Each conversion can waste some energy, which becomes heat that cooling systems must remove. This architecture has served conventional server loads, but the source says it becomes harder to fit the required power equipment into a rack as demand approaches a megawatt.

NVIDIA’s technical material describes the scale of that challenge: it says a one-megawatt rack supplied at 54 volts could require up to 200 kilograms of copper busbar, while power shelves could occupy as many as 64 rack units in its planned Kyber systems. NVIDIA says 800-volt DC can move more power through conductors of the same size. These are company-provided engineering claims. The source also describes a possible intermediate approach: a “sidecar” conversion cabinet beside a rack, allowing DC-ready racks to operate in buildings that retain much of their existing AC infrastructure.

The developments arrive amid broader growth in electricity demand. The IEA projects global data center electricity use will rise from 485 terawatt-hours in 2025 to 950 terawatt-hours in 2030. Berkeley Lab estimates that data centers used 4.4% of U.S. electricity in 2023 and could account for 6.7% to 12% by 2028. Separately, Berkeley Lab reported that U.S. grid interconnection queues held about 2,300 gigawatts of proposed generation and storage at the end of 2024; only 13% of projects in earlier queues were ultimately built. Queue totals represent proposed projects, not guaranteed future supply.

“Leaving no room for compute”

— NVIDIA, describing its 800-volt DC design

Costs, Performance and Grid Effects

The reported deals show growing interest, but they do not establish how widely 800-volt DC will be adopted or how much it will save in operating costs. NVIDIA’s copper, efficiency and maintenance figures are vendor claims. A Utility Dive article dated October 26 was sponsored by inverter maker SMA America; its model estimates that a conventional one-megawatt AC power chain loses about 80 kilowatts, compared with about 20 kilowatts in an 800-volt DC arrangement. Those modeled results are also not independent field measurements.

The available source does not provide Heron Power’s final project delivery details, installation sites or commercial terms. It also does not explain the cost allocation, construction schedule or battery duration for the NiSource-Microsoft agreement. More broadly, it remains unclear which data center projects will receive grid connections on schedule, how much new generation will be built, and how resulting infrastructure costs will affect electricity customers in particular regions.

Deployment and Grid Timelines

The next milestones are equipment production, project deployment and evidence from operating systems. NVIDIA’s stated target is full-scale production of its 800-volt design with Kyber systems in 2027. Vantage’s reservation may support future supply, but the source does not give a date for a specific installation. Details about the Heron system’s output, battery configuration and commercial performance would help show whether the design can meet the demands described.

For the NiSource-Microsoft agreement, construction and commissioning dates for the generation and battery projects are not provided in the source material. Operators, utilities and regulators will also have to resolve grid connection timing and the division of costs for new infrastructure. Independent measurements of efficiency, reliability and maintenance in deployed facilities will be needed to test vendor projections against real-world operation.

Key Questions

What does 800-volt DC mean for an AI data center?

It refers to a proposed system that converts grid electricity to 800-volt direct current near a facility and distributes DC toward high-power racks, rather than relying on multiple conversion stages before power reaches servers.

What did Vantage reserve?

According to ThorstenMeyerAI.com, Vantage reserved more than a gigawatt of factory capacity for a Heron Power solid-state transformer. The source does not say that this capacity has already been installed or identify a delivery schedule.

What is included in the Microsoft-NiSource deal?

The reported 20-year agreement adds about 1,100 megawatts of generation and includes a 600-megawatt battery. The supplied material does not give the battery’s duration or project timetable.

Are the projected efficiency and copper savings proven?

No independent measurements are provided in the source. NVIDIA’s figures, including a claimed 45% reduction in copper and up to 5% better end-to-end efficiency, are vendor estimates.

Could these projects affect household electricity bills?

New generation and grid infrastructure can affect electricity costs, but the source does not establish how much data center demand has contributed to household rate changes. The outcome depends on regional power markets and how project and grid-upgrade costs are assigned.

Source: ThorstenMeyerAI.com

This content is for general information only and is not financial, tax or legal advice. Consult a qualified professional for decisions about your money.
You May Also Like

Two Decades Of RISC OS Open: A Tech Signal Monitoring Perspective

Analysis of 20 years of RISC OS Open highlights how platform evolution impacts small software companies and tech monitoring strategies.

Write Code Like A Human Will Maintain It

Experts highlight the importance of writing code that humans can easily understand and maintain, shaping future software development practices.

Workflow Cloner Tools To Accelerate Helpdesk Platform Migration

New workflow cloning tools aim to streamline helpdesk platform switches by exporting and translating macros, rules, and automations, reducing manual rebuild time.

I Ran 71 AI Configurations Through My News Pipeline—Here’s The Catch

A 71-configuration test found GPT-6 Astra scored highest, but GPT-6 Luna was far cheaper. Two fallback models produced fabricated quotes or unusable output.