🔍 Read the full analysis: How Data Centers Are Reworking Power For AI With DC And Batteries on ThorstenMeyerAI.com
Get audio and creator gear delivered free with Prime
- Fast, free delivery on millions of items
- Prime Video, Amazon Music and more included
- Member-only deals all year
TL;DR
Two deals this month underscore how data center power planning is shifting toward high-voltage direct current and large batteries. Vantage reserved more than a gigawatt of factory capacity for Heron Power solid-state transformers, while NiSource and Microsoft signed a 20-year power agreement that includes about 1,100 megawatts of generation and a 600-megawatt battery. The deals point to new approaches, but do not establish how widely they will be deployed or who will bear related grid costs.
Data center power plans are shifting toward high-voltage direct current and batteries as AI workloads raise demand and strain grid connections. On October 21, operator Vantage reserved more than a gigawatt of factory capacity for Heron Power solid-state transformers; five days earlier, NiSource and Microsoft signed a 20-year power deal that adds about 1,100 megawatts of generation and a 600-megawatt battery. The agreements show investment in new supply and power equipment, but do not by themselves establish how much electricity will be delivered to data centers or how costs will be shared.
Heron Power’s solid-state transformer is designed to convert grid electricity into 800-volt direct current (DC), a format that could reduce the number of conversion steps between the grid and computing equipment. Vantage’s reservation is for factory capacity exceeding one gigawatt, not a stated amount of electricity already supplied or a count of transformers installed. The source material does not specify delivery dates, sites, or the full commercial terms.
The NiSource-Microsoft agreement is a separate development. NiSource signed a 20-year power deal that adds roughly 1,100 megawatts of generation and a battery rated at 600 megawatts. The reported figures describe the deal’s planned additions; further details about the generation mix, project schedule, battery duration, and allocation of costs were not provided in the source material.
The wider technical case is based on the increasing power demand of AI systems. The International Energy Agency (IEA) says AI server power density rose 11-fold from 2020 to 2025 and is projected to quadruple again by 2027. NVIDIA’s reference design calls for converting 13.8-kilovolt grid power to 800-volt DC for racks drawing up to one megawatt, with full-scale production planned alongside its Kyber systems in 2027. NVIDIA’s estimates of lower copper use and improved efficiency are company claims, not independent measurements.
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 Equipment Sets AI’s Pace
For data center developers, the bottleneck is increasingly not just access to computing chips but the equipment and infrastructure needed to power them. High-density racks require substantial electrical capacity in limited space. A DC distribution design could reduce some conversion losses and the copper needed to carry power, while batteries could help manage rapid changes in demand. These potential gains matter to operators facing long waits for transformers and grid connections.
The proposed changes also have consequences beyond data center campuses. The IEA estimates global data center electricity use could rise from 485 terawatt-hours in 2025 to 950 terawatt-hours in 2030. More generation, wires, and storage may be needed to serve that demand. Who pays for those additions—and whether households see higher bills—depends on project contracts and utility rules. The two deals alone do not show how costs will be allocated or establish a direct effect on consumer electricity prices.
Products Worth Considering
Anker SOLIX C200 DC Portable Power Station, 192Wh, 200W
Jackery HomePower 3600 Plus Portable Power Station, 3584Wh, 3600W
Apple iPad Air 11-inch (M2), 512GB, Wi-Fi, Space Gray
As an affiliate, we earn on qualifying purchases.
From AC Conversion to 800-Volt DC
Most data centers use alternating current (AC) from the grid, then step it down and convert it multiple times before servers use DC internally. Each conversion can waste some energy as heat, which cooling systems must remove. Traditional designs were built around lower-power racks; proposed AI systems are pushing toward rack loads near one megawatt, changing the scale of the electrical hardware required.
NVIDIA says delivering one megawatt at today’s 54 volts could require up to 200 kilograms of copper busbar. Its 800-volt approach is intended to carry more power through conductors of a given size. The company 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 vendor estimates. Separately, an October 26 Utility Dive article sponsored by inverter maker SMA America modeled losses of about 80 kilowatts in a traditional one-megawatt AC chain, versus about 20 kilowatts in a proposed 800-volt DC design. That comparison is also a sponsored vendor model, not an independent field test.
Operators may adopt DC in stages rather than rebuild entire facilities. SMA describes a “sidecar” conversion cabinet placed beside racks, which could let new 800-volt equipment operate in halls designed for AC. Batteries are also DC devices, so a DC-based system may make it easier to connect storage along the main power path. The IEA projects that 20–25 gigawatts of batteries could be installed in data centers by 2030; this is a forecast, not a confirmed deployment total.
Deployment and Costs Still Unsettled
The reported deals do not establish when Heron’s equipment will be delivered or installed, how much of Vantage’s reserved factory capacity will be used, or which data center sites will receive it. The NiSource-Microsoft agreement’s reported capacity figures do not specify the battery’s storage duration, the generation sources, or the projects’ construction schedules. Those details affect how much dependable power the projects can provide and when.
Performance claims for 800-volt distribution also need to be distinguished from measured results. NVIDIA’s copper and efficiency figures are vendor claims, while the SMA loss comparison comes from sponsored content and a modeled scenario. The source material supplies no independent, operating-site comparison of the systems. It is also unclear how utilities and customers will allocate the cost of new generation, batteries, transformers, and grid upgrades.
Grid constraints remain substantial, but their scale varies by location. The source material cites an estimated 2,300 gigawatts of projects in U.S. interconnection queues at the end of 2024, and Berkeley Lab’s finding that 13% of earlier requests were built. Those figures describe queued proposals and the historical conversion of requests, not a forecast of power that will reach data centers. Their relevance to either deal depends on the projects’ locations and connection status.
Project Schedules and Grid Connections
The next milestones are likely to be project-specific: further information on Heron equipment deliveries and Vantage deployments, and details on the generation and battery projects tied to the NiSource-Microsoft agreement. NVIDIA has said full-scale production of its 800-volt design is planned with Kyber systems in 2027. That date is a stated plan, not confirmation that the equipment will be installed or operating at data centers by then.
For operators and electricity customers, the key questions are whether new designs perform as projected at operating sites, how quickly utilities can provide connections, and which parties fund the required infrastructure. Until those details emerge, the deals are evidence of changing power strategies—not proof that grid constraints or electricity-price pressures have been resolved.
Key Questions
What did Vantage agree to?
On October 21, Vantage reserved more than one gigawatt of factory capacity for Heron Power solid-state transformers. The report does not say that this is one gigawatt of electricity already delivered to data centers.
What is included in the NiSource-Microsoft deal?
The companies signed a 20-year power agreement that adds about 1,100 megawatts of generation and a 600-megawatt battery. The source material does not provide the battery’s duration or the projects’ schedule.
Why are data centers considering 800-volt DC?
AI racks are drawing more power, and DC distribution could reduce conversion steps and the amount of copper needed to carry electricity. Claimed efficiency and copper savings are vendor estimates and are not presented here as independently verified results.
How could batteries help an AI data center?
Batteries could help buffer rapid changes in computing demand, provide power during grid disturbances, and support a controlled return to full load. The scale of future installations remains a projection: the IEA estimates 20–25 gigawatts by 2030.
Do these deals mean household electricity prices will rise?
The reported agreements do not establish a direct effect on household bills. Costs depend on the projects, utility arrangements, and rules for allocating expenses for generation and grid upgrades.
Source: ThorstenMeyerAI.com
