🔍 Read the full analysis: The AI Race Is Reshaping Data Center Power—and Electrical Jobs on ThorstenMeyerAI.com
Get the latest gadgets 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
Vantage reserved more than 1 gigawatt of factory capacity for Heron Power solid-state transformers, while NiSource signed a 20-year deal with Microsoft that includes about 1,100 megawatts of generation and a 600-megawatt battery. The deals highlight how AI data center growth is driving investment in electrical infrastructure, even as grid connections, equipment lead times and cost allocation remain unresolved challenges.
Data center operator Vantage reserved more than 1 gigawatt of factory capacity for Heron Power solid-state transformers on October 21, while utility NiSource signed a 20-year power agreement with Microsoft five days earlier that adds about 1,100 megawatts of generation and a 600-megawatt battery. The agreements reflect a shift in AI infrastructure investment from chips alone toward electricity supply, transformers, storage and grid access.
Heron Power’s transformer is designed to convert grid electricity into 800-volt direct current (DC) and integrate battery storage. Vantage’s reservation is for factory capacity, not a confirmed deployment of that full amount of equipment at a particular data center. The source material does not give a delivery schedule or identify the sites involved.
NiSource’s agreement with Microsoft is a separate development: it covers about 1,100 megawatts of generation and a 600-megawatt battery under a 20-year arrangement. The available details do not specify the plants, construction timetable, customer cost allocation or how much power will serve individual Microsoft facilities.
The deals come as operators examine electrical designs for increasingly dense AI computing. NVIDIA’s 800-volt reference design, planned for Kyber systems in 2027, would convert power near a facility’s perimeter and distribute DC to high-demand racks. The company says the approach can use 45% less copper than 415-volt AC distribution and improve end-to-end efficiency by up to 5% versus current 54-volt systems. Those are NVIDIA vendor estimates, not independent measurements of operating data centers.
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 Infrastructure Becomes an AI Bottleneck
The immediate significance is that AI data center expansion depends on equipment and connections that cannot be added as quickly as computing hardware. The International Energy Agency says transformer deliveries can take two to three years, while grid connections take five to ten years in many locations. Those timelines can limit when new facilities open, even if operators can obtain chips and build the structures.
Higher rack loads also change the economics and design of facilities. The IEA reported that AI server power density rose 11-fold from 2020 to 2025 and is projected to quadruple again by 2027. It says a rack could then reach peak demand comparable to 65 households. That projection helps explain interest in higher-voltage DC distribution, which may reduce the amount of copper and conversion equipment needed. The savings figures currently cited by manufacturers remain claims that require real-world verification.
More generation and storage can help meet data center demand, but their costs and effects on other customers depend on local rules and contracts. U.S. residential electricity averaged 18.21 cents per kilowatt-hour in 2026, compared with 17.30 cents in 2025, according to the source material citing the Energy Information Administration. That national change alone does not establish that data centers caused higher household bills. Who pays for new plants, batteries and grid upgrades is decided through utility arrangements and regulatory processes.
Products Worth Considering
Jackery Explorer 300 Portable Power Station, 292Wh
Anker SOLIX C200 DC Portable Power Station, 192Wh, 200W
DJI Osmo Mobile 8 3-Axis Smartphone Gimbal Stabilizer
As an affiliate, we earn on qualifying purchases.
From AC Server Rooms to 800-Volt Racks
Conventional data centers receive alternating current (AC), step it down through transformers and power equipment, then convert it to low-voltage DC inside servers. Each conversion loses some energy as heat, which cooling systems must remove. This design served lower-density racks, but AI systems concentrating more compute in a single rack increase the amount of power that must be delivered through a limited space.
NVIDIA’s technical proposal would convert incoming 13.8-kilovolt AC to 800-volt DC near the facility edge. The company says delivering 1 megawatt at 54 volts could require as much as 200 kilograms of copper busbar, while higher-voltage distribution reduces conductor requirements. Operators may also adopt intermediate “sidecar” conversion cabinets beside racks, allowing new equipment to work in halls designed around AC systems. These design descriptions and performance estimates come from NVIDIA and power-equipment suppliers, not a reported independent comparison.
Storage is part of the same discussion because AI workloads can produce rapid changes in electricity demand. The IEA estimates that data center installations could include 20 to 25 gigawatts of batteries by 2030. That is a forecast, not installed capacity. Batteries could buffer swings or support a facility during grid disturbances, but their grid role depends on operating arrangements and incentives.
“The company says its design could use 45% less copper than 415 V AC distribution and deliver up to 5% better end-to-end efficiency than current 54 V systems.”
— NVIDIA, in its 800 VDC technical blog
Delivery Dates and Customer Costs
The announced agreements do not establish when Heron Power will deliver transformers, where Vantage will use them or how much of the reserved factory capacity will become installed equipment. The source material also does not provide the full terms of NiSource’s Microsoft agreement, including project locations, construction milestones or the division of costs among the utility, Microsoft and other electricity customers.
It is also uncertain how quickly 800-volt DC systems will become common. NVIDIA’s full-scale production plan points to 2027, but the performance and maintenance savings cited by vendors have not been independently verified here. A Utility Dive article published October 26 was sponsored by inverter maker SMA America; its model of lower losses for an 800-volt system should be read as a vendor-backed estimate, not a neutral field result.
National and regional electricity figures do not show, on their own, how much data center growth affects a particular household’s bill. The source material cites a 41% increase in wholesale prices in the PJM region but does not provide the comparison period in the supplied text. The causes of price changes and the eventual allocation of grid costs vary by location and remain unsettled.
Projects, Permits and Grid Connections
The next evidence will come from project timelines: whether reserved transformer capacity turns into equipment deliveries, when NiSource’s generation and battery resources are built, and how the power agreement is reflected in local utility plans. Regulators and utilities will also determine how new infrastructure costs are assigned and whether storage can provide services to the wider grid.
For data center operators, the near-term question is whether equipment supply and grid interconnection schedules can match planned AI campus growth. For electricity customers, the relevant milestones are utility filings, approved construction costs and rate decisions—not announcements alone. Independent operating data will be needed to test whether high-voltage DC systems deliver the efficiency, copper and maintenance savings their proponents project.
Key Questions
What did Vantage announce?
Vantage reserved more than 1 gigawatt of factory capacity for Heron Power solid-state transformers. The announcement concerns manufacturing capacity; the supplied information does not identify deployment sites or delivery dates.
What is included in NiSource’s deal with Microsoft?
The 20-year agreement includes about 1,100 megawatts of generation and a 600-megawatt battery. Details on project locations, schedules and cost allocation were not provided in the source material.
Why are data centers considering 800-volt DC power?
AI racks are drawing more power in concentrated spaces. Higher-voltage DC distribution may reduce the copper and conversion equipment required, but savings cited by NVIDIA and suppliers are vendor claims that need independent operational verification.
Will these projects raise household electricity bills?
The announcements do not establish how they will affect household rates. The result depends on local power markets, utility contracts, regulatory decisions and who pays for generation and grid upgrades.
What is the main constraint on new AI data centers?
In many places, access to electricity infrastructure can take longer than building computing equipment. The IEA cited transformer delivery times of two to three years and grid connection waits of five to ten years in many locations.
Source: ThorstenMeyerAI.com
