How AI Data Centers Are Adopting 800-Volt DC And Batteries
AIThis post was created with the assistance of artificial intelligence (AI).

🔍 Read the full analysis: How AI Data Centers Are Adopting 800-Volt DC And Batteries on ThorstenMeyerAI.com

Buying for a business?Offer from Amazon

Get business pricing on office and shipping supplies

  • Business-only prices and quantity discounts
  • Tax-exempt purchasing
  • Multiple users, one account, clear invoices
As an affiliate, we earn on qualifying purchases.

TL;DR

Vantage reserved more than 1 gigawatt of manufacturing capacity for a solid-state transformer designed to convert grid power directly to 800-volt DC. The move, alongside a Microsoft-NiSource power agreement that includes a 600-megawatt battery, reflects efforts to serve high-density AI facilities; the timing, cost and independent performance of the new systems remain uncertain.

Data-center power infrastructure is drawing new investment as AI workloads push up electricity needs: Vantage reserved more than 1 gigawatt of factory capacity for a solid-state transformer designed to convert grid power directly to 800-volt direct current, while NiSource and Microsoft agreed to a long-term power arrangement that includes a 600-megawatt battery. The deals show companies planning for heavier, less predictable loads, but do not by themselves establish how widely the technology will be adopted or how much it will reduce power costs, a challenge explored in US data-center capacity constraints.

Vantage made its reservation on October 21, according to ThorstenMeyerAI.com. The planned transformer, from Heron Power, is designed to take grid electricity and deliver 800-volt DC to a data center. The source describes the equipment as having energy storage built in, intended to help buffer rapid changes in demand. A factory-capacity reservation indicates planned manufacturing access; it does not confirm that a gigawatt of equipment has been installed or is operating.

On October 16, Indiana utility NiSource signed a 20-year power agreement with Microsoft. The arrangement adds about 1,100 megawatts of generation and a 600-megawatt battery, the source reports. It is a separate development from Vantage’s transformer reservation, though both address the difficulty of supplying large data-center loads and the power bottleneck facing AI data centers. The source does not specify the battery’s energy capacity in megawatt-hours, which would indicate how long it can deliver power at its rated output.

The technical case for 800-volt DC is tied to the amount of electricity required by dense AI server racks and the growing overlap between AI and data-center operations. NVIDIA’s reference design proposes converting 13.8-kilovolt grid power to 800-volt DC near a facility’s perimeter, then distributing DC to racks. NVIDIA says the design could use 45% less copper than 415-volt AC distribution and improve end-to-end efficiency by up to 5% compared with current 54-volt systems. These are company estimates, not independently verified results. The source says full-scale production with NVIDIA’s Kyber systems is planned for 2027.

At a glance
reportWhen: Developments reported October 16–21, 20…
The developmentTwo recent power deals point to data-center operators and utilities preparing for AI loads with 800-volt DC equipment, new generation and batteries.
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 Design Moves Beyond the Server

AI infrastructure planning increasingly depends on more than securing processors. High-density racks require substantial power delivery, and converting electricity repeatedly can add losses and heat that facilities must remove. An 800-volt DC architecture is intended to reduce some conversion steps and carry more power through conductors, while batteries can respond to short-term swings or grid disturbances. Whether the approach delivers the projected gains at operating sites remains to be demonstrated independently.

The projects also point to the grid connection and equipment supply as constraints on new capacity. ThorstenMeyerAI.com, citing the International Energy Agency, says transformers can take two to three years to arrive and grid connections can take five to ten years in some locations. Those timelines can affect when planned data centers become usable, regardless of how quickly servers are procured. The NiSource-Microsoft agreement pairs generation with storage, but the source does not detail how the costs or benefits will be allocated among the utility, the company and other electricity customers.

Electricity demand has broader implications for communities and energy markets, but the effects will vary by location and contract. The source reports an average US household electricity price of 18.21 cents per kilowatt-hour for the year, compared with 17.30 cents in 2025, and a 41% increase in wholesale prices in the PJM region. These figures describe different price measures and markets; they do not establish that data centers caused either increase.

Apple AirPods Pro 3 Wireless Earbuds with Active Noise Cancellation

Apple AirPods Pro 3 Wireless Earbuds with Active Noise Cancellation

KitchenAid Artisan Series 5-Quart Tilt-Head Stand Mixer, Dried Rose (KSM150PS)

KitchenAid Artisan Series 5-Quart Tilt-Head Stand Mixer, Dried Rose (KSM150PS)

As an affiliate, we earn on qualifying purchases.

Why AI Loads Challenge AC Systems

Conventional data centers generally receive alternating current, then step down and convert it through transformers, uninterruptible power supplies and distribution equipment before servers use low-voltage DC internally. Each conversion can lose energy. At the much higher rack loads associated with AI, the size and efficiency of power-delivery equipment become more consequential.

The International Energy Agency estimates global data-center electricity use will rise from 485 terawatt-hours in 2025 to 950 terawatt-hours in 2030, roughly doubling. It also reports that AI server power density rose 11-fold from 2020 to 2025 and is projected to quadruple again by 2027. The IEA’s projection is not a measurement of future consumption, and the scale of the increase will depend on deployment and efficiency.

NVIDIA says a one-megawatt rack supplied at 54 volts could require up to 200 kilograms of copper busbar. In its proposed design, higher-voltage DC is distributed closer to the rack. Operators may not need to replace entire facilities at once: an industry article published by Utility Dive on October 26, sponsored by inverter maker SMA America, describes a “sidecar” conversion cabinet that could let 800-volt racks operate in existing AC halls. Its modeled power-loss comparison is a vendor-sponsored estimate, not an independent field result.

“45% less copper”

— NVIDIA, in its 800 VDC technical blog

Deployment Costs and Results Remain Open

The reported reservation does not establish when Heron Power’s equipment will be delivered, how much of the reserved factory capacity Vantage will use, or whether the transformer and integrated storage have been tested at commercial scale. The source also does not give the contract’s financial terms or explain how the agreement defines the capacity being reserved.

Independent operating data on 800-volt DC systems is not provided. NVIDIA’s efficiency, copper and maintenance figures are projections from the company; SMA’s power-loss figures come from sponsored material and a modeled comparison. Actual outcomes could depend on facility layout, equipment choices, load patterns and local power conditions.

The source does not explain the NiSource-Microsoft agreement’s detailed cost allocation, the battery’s duration, or how the added generation will be delivered. Nor does it establish that data-center demand caused the household or PJM price movements cited. Those questions matter to determining who ultimately bears the cost of new generation, storage and grid infrastructure.

Watch for Delivery and Field Data

The next milestones are the delivery and deployment of the reserved transformer capacity and progress toward NVIDIA’s stated 2027 production plan for Kyber systems. Operators’ choices between new DC-built facilities and retrofit equipment such as sidecars will help show whether 800-volt distribution can fit existing buildings as well as new ones.

For the NiSource-Microsoft deal, further details about generation schedules, battery duration and cost responsibility would clarify what the 20-year agreement means for the grid and customers. Independent measurements of energy losses, cooling needs, reliability and maintenance at operating sites will be needed to test vendor projections. Until those details emerge, the deals show preparation for AI power demand, not proof that the new designs will deliver a specific cost or efficiency outcome.

Key Questions

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

It describes a proposed way to distribute 800-volt direct current through a facility to high-power racks, rather than relying on several stages of AC conversion and lower-voltage distribution. The potential efficiency and material savings cited in the source are vendor projections.

What did Vantage reserve?

Vantage reserved more than 1 gigawatt of factory capacity for a Heron Power solid-state transformer designed to convert grid power directly to 800-volt DC. The reservation is not confirmation that the equipment has been delivered or installed.

What is included in the NiSource-Microsoft agreement?

The reported 20-year agreement adds about 1,100 megawatts of generation and a 600-megawatt battery. The source does not provide the battery’s energy duration or detailed terms for allocating costs.

Do batteries mean a data center can run independently of the grid?

Not on the information provided. Batteries can store and deliver electricity and may help manage load swings or disturbances, but the source does not say these projects are designed to operate data centers independently of grid power.

When will the new systems be in use?

NVIDIA has said full-scale production with its Kyber systems is planned for 2027. The source does not give a confirmed operating date for Vantage’s reserved transformer equipment or for the battery included in the NiSource-Microsoft arrangement.

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

The SSD Squeeze: Why Storage Joined the Party

Enterprise and consumer SSD prices surge as NAND supply tightens due to AI demand and factory competition, impacting the entire storage market.

Where AI’s 176GB Disappears To: The Hidden Memory Drain

Exploring the unseen memory costs in AI inference, focusing on the overlooked impact of the KV cache and other factors on large models’ memory use.

How AI Will Impact Daily Life In 2026: Top 6 Insights

Exploring how artificial intelligence will reshape everyday activities by 2026, with six key insights on technology, work, and society.

AI, Not Just Acceleration, Is The Story Of Tesla’s New Roadster

Tesla showed its second-generation Roadster in Texas, pitching extreme performance while pointing toward a future of AI and autonomous vehicles.