🔍 Read the full analysis: How 800-Volt DC And Batteries Could Support Growing AI Demand 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
Vantage reserved more than 1 gigawatt of factory capacity for Heron Power solid-state transformers that convert grid electricity to 800-volt DC. A separate 20-year NiSource-Microsoft deal includes about 1,100 megawatts of generation and a 600-megawatt battery. The developments reflect efforts to manage AI data centers’ growing power needs, but efficiency figures remain vendor claims and the projects do not resolve broader questions about grid capacity and costs.
Vantage has reserved more than 1 gigawatt of factory capacity for a solid-state transformer designed to convert grid power directly into 800-volt direct current, while a separate NiSource-Microsoft agreement pairs new generation with a large battery. The deals, reported in October, show how data center operators and power suppliers are addressing rising AI electricity demand, though they do not by themselves establish how much the technology will improve efficiency or lower costs.
Vantage’s reservation, reported on October 21, is for factory capacity for Heron Power’s solid-state transformer. The source account says the equipment is designed to convert grid electricity to 800-volt DC and incorporate energy storage. A factory-capacity reservation is not the same as a report that more than a gigawatt of equipment has already been installed or is operating.
On October 16, Indiana utility NiSource signed a 20-year power deal with Microsoft that adds about 1,100 megawatts of generation and a 600-megawatt battery, according to the source. The agreement ties new power supply and storage to a major technology customer, but the available account does not specify all project locations, construction milestones or how costs will be allocated.
The discussion gained attention again on October 26, when Utility Dive published a piece examining DC power for AI data centers. The source identifies it as sponsored content by inverter maker SMA America, so its modeled comparisons should be treated as a vendor’s case, not independent measurements. The article’s model estimates losses of about 80 kilowatts when delivering 1 megawatt through a conventional AC conversion chain, compared with about 20 kilowatts using an 800-volt DC approach. Those figures are model results, not verified performance from the Vantage or NiSource projects.
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 Delivery Is Becoming a Constraint
The underlying pressure is the rapid growth in the power required by AI computing. The International Energy Agency says AI server power density rose 11-fold from 2020 to 2025 and is projected to quadruple again by 2027. It estimates a future rack could draw at peak as much electricity as 65 households. These figures describe power density and projected peak demand, not average consumption by every data center rack.
Higher rack demand makes electrical distribution a practical constraint alongside access to chips and computing space. NVIDIA’s reference design proposes converting 13.8-kilovolt grid power to 800-volt DC near a facility’s perimeter, then distributing DC to high-power racks. The company says the design could use 45% less copper than 415-volt AC distribution and deliver up to 5% better end-to-end efficiency than current 54-volt systems. Those are NVIDIA’s engineering claims; the source does not cite independent operational measurements confirming them.
Batteries could help manage short-term changes in demand as well as provide backup or support during grid disturbances. The IEA says data centers could have 20–25 gigawatts of batteries by 2030, and says storage might also serve the grid if suitable incentives are in place. Whether that potential is realized will depend on project design, regulation and commercial arrangements. For electricity customers, the central issue is not only whether a data center uses power more efficiently, but also who pays for generation, wires and grid upgrades.
Products Worth Considering
Dell UltraSharp 27 Thunderbolt Hub Monitor (U2726DER)
Dell UltraSharp U2725QE 27-Inch 4K UHD IPS Black Monitor
LG UltraGear 27-Inch QHD OLED Gaming Monitor, 240Hz, 0.03ms, G-SYNC Compatible, FreeSync Premium Pro, HDMI 2.1
As an affiliate, we earn on qualifying purchases.
From AC Conversion to High-Voltage DC
Most data centers have traditionally received alternating current and converted it through multiple stages before servers use low-voltage DC. Each conversion can lose some energy, with the resulting heat adding to cooling needs. The move to higher-voltage DC aims to reduce conversion steps and avoid the large conductors required to carry very high current at low voltage.
NVIDIA’s technical material says a 1-megawatt rack supplied at 54 volts could require up to 200 kilograms of copper busbar. Its proposed 800-volt design is intended for future Kyber systems, with full-scale production planned for 2027, according to the company. A nearer-term approach described by SMA is a “sidecar” conversion cabinet beside a rack, allowing DC equipment to work in existing AC-oriented halls while much of the older power system remains in place.
The broader build-out is substantial. The IEA estimates global data center electricity use could rise from 485 terawatt-hours in 2025 to 950 terawatt-hours in 2030. Separately, 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 forecasts and estimates with stated periods, not a measure of current use at any particular facility.
“leaving no room for compute”
— NVIDIA, describing its proposed 800-volt DC design
Efficiency and Cost Claims Need Testing
The reported developments do not establish how much energy an operating 800-volt DC data center will save, whether those savings will outweigh equipment and conversion costs, or how quickly facilities can adopt the design. The efficiency, copper and maintenance figures cited for NVIDIA’s system are company claims; SMA’s loss comparison is a sponsored vendor model. Independent results from deployed systems are not provided in the source account.
Project details also remain limited. The account does not specify when Heron Power’s reserved factory capacity will translate into deliveries, how much storage is included in the transformer arrangement, or the precise timetable and cost allocation for NiSource’s generation and battery projects. Nor does it establish how much either agreement will change electricity prices for other customers.
Grid access remains a separate challenge. The source cites IEA estimates that transformers can take two to three years to arrive and grid connections can take five to ten years in many places. Berkeley Lab reported about 2,300 gigawatts of proposed capacity in U.S. interconnection queues at the end of 2024, while only 13% of earlier requests were built. Queue totals are proposed projects, not assured future supply.
Projects and Grid Plans Will Test Designs
The next milestones are the delivery and deployment schedules for Heron Power’s equipment, the construction progress of NiSource’s generation and battery projects, and NVIDIA’s stated plan for full-scale production with Kyber systems in 2027. Those developments should provide more information about whether the designs can be installed at the scale and pace data center operators need.
Utilities, regulators and customers will also have to determine how new power plants, storage and grid connections are funded. The source reports U.S. household electricity prices averaging 18.21 cents per kilowatt-hour in 2026, compared with 17.30 cents in 2025, and says wholesale prices in the PJM region were up 41%. Those figures do not by themselves show that data centers caused the increases. The connection between specific projects and customer bills will depend on local rules, contracts and grid conditions.
For now, the deals show that power architecture and storage are becoming part of AI infrastructure planning, alongside computing equipment. Whether 800-volt DC materially eases constraints will depend on measured performance, available grid capacity and the practical economics of each project.
Key Questions
What is the reported Vantage development?
Vantage reserved more than 1 gigawatt of factory capacity for Heron Power solid-state transformers designed to convert grid electricity to 800-volt DC. The reservation is not evidence that that amount of equipment is already installed.
What does the NiSource-Microsoft deal include?
The 20-year agreement adds about 1,100 megawatts of generation and a 600-megawatt battery, according to the source. The account does not provide a full construction schedule or cost allocation.
Why are data centers considering 800-volt DC?
Higher-voltage DC can reduce the current and conductor requirements for delivering power to very demanding racks, and may reduce conversion stages. Specific efficiency and copper savings cited in the source are vendor claims, not independently verified results from these projects.
How could batteries help AI data centers?
Batteries can store electricity and respond to brief changes in demand, potentially helping smooth AI load swings or support a site during grid disturbances. Their potential to provide wider grid services depends on design and incentives.
Will these projects lower household electricity bills?
The source does not establish that. The effect on household bills depends on how generation, storage and grid upgrades are funded, as well as local electricity-market rules and conditions.
Source: ThorstenMeyerAI.com
