🔍 Read the full analysis: Why AI Data Centers Are Looking Beyond Traditional Power Setups on ThorstenMeyerAI.com
Get business pricing on office and shipping supplies
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
TL;DR
A Vantage reservation for more than 1 GW of solid-state transformer production and a NiSource-Microsoft power agreement point to changing infrastructure for AI data centers. Operators are exploring 800-volt direct current, batteries and dedicated generation as rack demand rises, but claimed efficiency gains and the wider effect on electricity bills remain uncertain.
Two October agreements show how the growth of AI data centers is prompting changes beyond buying more electricity: Vantage reserved more than 1 gigawatt of factory capacity for solid-state transformers, while utility NiSource signed a 20-year power deal with Microsoft that adds about 1,100 megawatts of generation and a 600-megawatt battery. The deals arrive as operators consider 800-volt direct-current systems and storage to handle increasingly power-hungry AI racks and delays in grid connections.
Vantage reserved the transformer production capacity on October 21, according to ThorstenMeyerAI.com. The solid-state transformer is designed to convert grid electricity directly into 800-volt direct current. The report does not specify how much of the reserved capacity will be deployed, where it will be installed or when the equipment will enter service.
Five days earlier, NiSource and Microsoft signed a 20-year power agreement. The reported deal adds about 1,100 MW of generation and includes a 600-MW battery. The source does not give further details on the generation technology, project location, contract terms or how the electricity and storage will be allocated.
The technical rationale is that conventional data centers convert electricity several times between its arrival at a facility and its use inside servers. At higher rack loads, equipment and wiring designed for lower-voltage distribution can take up space and add conversion losses. NVIDIA’s design proposes converting 13.8-kilovolt grid power to 800 V DC at the facility perimeter, then distributing it to racks. The company says its system 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. Those figures are vendor 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 Design Is Becoming a Capacity Constraint
Electricity infrastructure is increasingly part of the timeline and cost of building an AI data center. The International Energy Agency says transformer deliveries can take two to three years, while grid connections may take five to 10 years in many places. A design that makes better use of available power, or storage that buffers short-term demand, could help operators manage constraints. Neither approach, by itself, creates new grid capacity or guarantees a faster connection.
The issue extends beyond data center companies. The IEA estimates global data center electricity use could rise from 485 terawatt-hours in 2025 to 950 TWh in 2030. Berkeley Lab estimated data centers used 4.4% of U.S. electricity in 2023 and could account for 6.7% to 12% by 2028. These are projections with uncertainty, not settled outcomes. As facilities seek more power, the cost of generation and network upgrades—and who pays—can affect utilities, customers and local communities.
Storage could also change how a data center interacts with the grid. The IEA says batteries may help manage rapid changes in demand and could become a grid asset if appropriate incentives are in place. Whether that happens depends on operating arrangements, compensation and the ability to dispatch the batteries when the wider system needs them.
Products Worth Considering
Samsung 27 Odyssey OLED G5 (G50SF) QHD QD-OLED Gaming Monitor
HP Sprocket Photo Booth Machine Instant Color Photo Printer with 10.1 Touchscreen, LED Ring Light, 250+ Frames, 3×4 ZINK Paper, QR Code Shar
LG 34GX90SA-W 34-inch Ultragear WQHD OLED Curved Gaming Monitor
As an affiliate, we earn on qualifying purchases.
From Server Loads to Megawatt Racks
Traditional data centers generally receive alternating current and pass it through transformers, uninterruptible power supplies and distribution equipment before converting it to the lower-voltage direct current used by servers. Each conversion loses some energy, much of which becomes heat that cooling systems must remove. This arrangement was developed for rack loads far below the levels now associated with large AI systems.
The IEA’s April 2026 report says AI server power density increased 11-fold from 2020 to 2025 and could quadruple again by 2027. It projects that a rack could then draw as much power at peak as 65 households. NVIDIA says powering a 1-MW rack at 54 volts could require up to 200 kilograms of copper busbar; it plans its 800-volt design for Kyber systems in 2027. The design date and performance figures come from NVIDIA, rather than a completed independent evaluation.
Not every operator can replace existing electrical infrastructure at once. Utility Dive’s October 26 article, sponsored by inverter maker SMA America, describes a “sidecar” conversion cabinet that could let 800-volt racks operate in halls built around AC systems. Its model estimates losses of about 80 kilowatts when delivering 1 MW through a traditional chain, compared with about 20 kW in an 800-volt DC arrangement. Those numbers are from a sponsored vendor white paper and should not be treated as independently verified results.
“Moves 85% more power through the same conductor size”
— NVIDIA, describing its 800-volt design
Deployment, Savings and Costs Remain Open
The reported deals do not establish how much new 800-volt equipment will be installed or when it will operate. The source gives no project-level schedule or detailed terms for the Vantage reservation, and limited information about the generation and battery components of the NiSource-Microsoft agreement.
Efficiency and copper savings figures cited for 800-volt systems come from NVIDIA and SMA-related material. Independent operating data comparing those systems with conventional power designs is not provided. It is also unclear how widely operators will adopt full facility redesigns rather than incremental equipment such as sidecar cabinets.
The effect on household electricity costs is unsettled and may vary by region and by how utilities allocate the cost of generation and grid upgrades. The source reports an average U.S. residential price 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 establish that data centers caused the changes or show how costs will be shared.
Watch for Projects and Grid Terms
The next evidence will come from project details and operating results: where the reserved transformers are deployed, when they are commissioned, and whether measured losses and copper requirements match vendor estimates. NVIDIA has said full-scale production of its 800-volt design is planned with Kyber systems in 2027.
For the NiSource-Microsoft agreement, further information on generation projects, battery operation and how the 20-year contract affects local power supply would clarify what the deal means in practice. Utilities and regulators will also shape whether data-center batteries can provide services to the grid, and how the costs of new plants and wires are assigned. Until those details emerge, the agreements signal a shift in planning priorities, but not a settled model for powering AI facilities.
Key Questions
Why are AI data centers considering 800-volt DC?
Higher-power racks can require substantial wiring and conversion equipment. Direct-current distribution at a higher voltage is intended to carry more power through conductors with less copper and reduce conversion losses. Claimed savings still need to be checked against independent operating data.
What did NiSource and Microsoft agree to?
The source reports a 20-year power deal that adds about 1,100 MW of generation and includes a 600-MW battery. It does not provide details about the generation technology, location or contract pricing.
Can batteries supply a data center with all its power?
The reported role is to buffer rapid demand changes, help manage grid disturbances and potentially support the wider grid. The source does not say the batteries in the NiSource-Microsoft deal are intended to provide a data center’s full power supply.
Will these power changes lower household electricity bills?
That is not established by the reported deals. Electricity costs depend on regional supply, grid investment and how utilities allocate costs. The available figures do not show that data centers caused the reported price changes.
Source: ThorstenMeyerAI.com
