The New Power Mix Behind AI Data Centers: 800-Volt DC And Batteries
AIThis post was created with the assistance of artificial intelligence (AI).

🔍 Read the full analysis: The New Power Mix Behind AI Data Centers: 800-Volt DC And Batteries 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

Two October power developments highlight how AI data centers are changing their electricity infrastructure: Vantage reserved more than a gigawatt of production capacity for Heron Power solid-state transformers, and NiSource signed a long-term deal with Microsoft that includes new generation and a 600 MW battery. The shift toward 800-volt DC and storage is intended to address high rack loads and power swings, but deployment costs, performance and effects on electricity bills remain uncertain.

Vantage reserved more than one gigawatt of factory capacity on October 21 for Heron Power solid-state transformers designed to convert grid electricity directly to 800-volt direct current, while a separate NiSource-Microsoft agreement includes about 1,100 megawatts of generation and a 600-megawatt battery. The developments point to a growing effort to match power infrastructure to energy-intensive AI data centers, though the reservation is not evidence that all the equipment is installed or operating.

Heron Power’s solid-state transformer is designed to convert incoming grid power to 800-volt DC, with battery storage built into the system, according to the source report. Vantage’s reservation concerns factory capacity; publicly supplied details do not establish how many transformers will be delivered, where they will be installed or when they will enter service.

On October 16, Indiana utility NiSource signed a 20-year power agreement with Microsoft. The deal adds about 1,100 megawatts of generation and a 600-megawatt battery, the report says. It does not give further details on the generation mix, project schedule, the battery’s duration or how costs will be allocated between the companies and other customers.

The backdrop is rising electricity demand and rack density, amid power bottlenecks for AI data centers. The International Energy Agency estimates data center electricity use will rise from 485 terawatt-hours in 2025 to 950 TWh in 2030. Its April 2026 report says AI server power density increased elevenfold from 2020 to 2025 and is projected to quadruple again by 2027. These are agency estimates and projections, not measurements of every facility.

At a glance
reportWhen: Developments announced October 16 and O…
The developmentA transformer capacity reservation for Vantage and a NiSource-Microsoft power agreement show data center operators and utilities pursuing new ways to supply AI facilities.
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 Supply Shapes AI Expansion

For data center developers, the issue is no longer just securing processors and buildings: they also need large, reliable power connections, equipment to distribute electricity inside facilities and ways to manage rapid changes in demand. A move to higher-voltage DC could reduce the amount of copper and the number of conversion stages needed for very high-power racks, while batteries could help buffer short-term fluctuations. How much those approaches save in practice will depend on facility design and operating conditions.

Grid access is also a planning constraint. The source cites IEA estimates that transformers can take two to three years to arrive and connections can take five to ten years in many locations. Those waits could affect when new computing capacity comes online. The costs of new generation and transmission also matter beyond the data center: who pays is being determined through individual utility and customer agreements, and effects on household bills may vary by location.

Some efficiency figures circulating in the debate are vendor claims. NVIDIA says its planned 800 V DC design could use 45% less copper than 415 V AC distribution and improve end-to-end efficiency by up to 5% compared with current 54 V systems. Those figures should not be treated as independently verified results or as guaranteed savings across data centers.

Products Worth Considering

Logitech MX Master 4 for Mac, Ergonomic Bluetooth Mouse with Haptic Feedback, Ultra-Fast Scrolling, USB-C Charging, for MacOS, iPadOS, Windo

Logitech MX Master 4 for Mac, Ergonomic Bluetooth Mouse with Haptic Feedback, Ultra-Fast Scrolling, USB-C Charging, for MacOS, iPadOS, Windo

As an affiliate, we earn on qualifying purchases.

From AC Distribution to 800 V

Most data centers use alternating current from the grid, then step voltage down and convert power through several stages before servers use it as low-voltage DC. Losses at those stages become heat that cooling equipment must remove. The existing design was developed for lower-power racks; the source report says the push toward racks approaching a megawatt is prompting interest in a different layout.

NVIDIA’s reference design proposes converting 13.8-kilovolt grid power to 800 V DC near the facility perimeter, then distributing DC to high-density racks. The company says a 1 MW rack supplied at 54 volts could require up to 200 kilograms of copper busbar. It plans full-scale production with its Kyber systems in 2027. These are design details and claims from NVIDIA, not independent field measurements.

A Utility Dive article published October 26, sponsored by inverter maker SMA America, presented a vendor white-paper model comparing power losses. It estimates losses of about 80 kilowatts in a traditional AC chain delivering 1 MW, versus about 20 kW with conversion to 800 V DC outside the building. SMA also describes a “sidecar” conversion cabinet that could allow 800 V racks to operate in existing AC data halls. The comparison is a modeled case from sponsored material, not a reported result from a deployed site.

Deployment and Costs Remain Open

The announced developments do not establish how quickly the equipment and power projects will be built, or how much capacity will reach operating data centers. Vantage’s reservation is for manufacturing capacity, not confirmation of a completed installation. The NiSource-Microsoft agreement’s project locations, generation technology, construction timetable and battery operating duration are not specified in the source material.

It is also unclear how 800 V DC systems will perform across different sites compared with existing AC designs. The efficiency and copper-reduction figures cited here come from NVIDIA and SMA-sponsored material; independent operating data were not provided. The IEA’s 20–25 GW estimate for data center batteries by 2030 is a projection, not a tally of installed systems.

Electricity costs for other customers are another open issue. The source reports average US residential electricity prices of 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 alone do not show what share of the changes is caused by data centers; the report does not supply a causal breakdown or explain how the figures were calculated.

Projects and Standards Ahead

Near-term milestones include further details on Heron Power’s delivery schedule and Vantage’s intended installations, as well as project filings and construction plans related to the NiSource-Microsoft agreement. Those details will show whether the announced capacity turns into operating supply and when the battery could be used.

NVIDIA has said full-scale production of its 800 V DC design is planned with Kyber systems in 2027. Operators, utilities and equipment suppliers will also need to establish how new DC systems connect safely and reliably to existing facilities. As more projects come online, independently measured efficiency, battery performance and cost allocation will help clarify whether the approach can ease power constraints and who bears the expense.

Key Questions

What did Vantage reserve?

Vantage reserved more than one gigawatt of factory capacity for Heron Power solid-state transformers designed to convert grid power to 800 V DC. The report does not say that this equipment has been delivered or installed.

What is included in the NiSource-Microsoft agreement?

The reported 20-year agreement includes about 1,100 MW of generation and a 600 MW battery. The source does not specify the generation mix, project schedule or battery duration.

Why are data centers considering 800 V DC?

Higher-voltage DC distribution is intended to supply very high-power AI racks while reducing conversion stages and the amount of copper needed. Specific efficiency and material savings cited in the report are company or vendor claims, not independently verified outcomes.

Will batteries lower household electricity bills?

The announcements do not establish that. Batteries may help data centers manage power fluctuations, but the effect on household bills depends on project costs, utility arrangements and local electricity markets. The source does not identify a direct bill impact from these deals.

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 Skills Marketplace, Six Months Later: Predicted vs Actual

An analysis of the skills marketplace six months after predictions, highlighting confirmed developments, structural complexities, and future outlooks.

What Would AI Policy Look Like In A Canada-EU Model?

Analysis of how a Canada-EU AI policy framework might look, comparing models, licenses, and strategic differences shaping future AI regulation and deployment.

Replace AI Agent Check-Ins With Decision-Focused Reports

A ThorstenMeyerAI.com article argues that AI agent updates should be limited to decisions people can act on, citing Anthropic workforce data.

How 10 Mobile Workstation Laptops Suit Professional AI Work In 2027

A source-based comparison of 10 workstation configurations, their listed hardware, and the questions professionals should verify for AI workloads.