🔍 Read the full analysis: Why The AI Boom Is Changing How Data Centers Use Electricity on ThorstenMeyerAI.com
Prime made for students and young adults
- Fast, free delivery for dorm and study essentials
- Prime Video and Amazon Music included
- Member-only deals
TL;DR
Two October developments show how AI data center growth is changing power planning: Vantage reserved more than a gigawatt of factory capacity for an 800-volt DC transformer, and NiSource signed a 20-year power deal with Microsoft that includes new generation and a 600-megawatt battery. The deals point to rising interest in direct-current distribution and storage, but grid connections, equipment supply and the cost allocation for new infrastructure remain unresolved challenges.
Data center power planning is shifting as AI workloads push demand beyond the capabilities of designs built for conventional server racks. On October 21, operator Vantage reserved more than a gigawatt of factory capacity for a solid-state transformer designed to deliver 800-volt direct current, while NiSource announced a 20-year power agreement with Microsoft five days earlier that includes about 1,100 megawatts of generation and a 600-megawatt battery.
The agreements address different parts of the power problem. Vantage’s reservation concerns equipment capacity for a transformer that converts grid electricity for data center use. NiSource’s agreement concerns supply: it adds generation and battery storage to support Microsoft’s data center needs. The announcements do not establish that all planned capacity is already operating, or that the projects will serve only one facility.
The technical case for higher-voltage DC distribution is tied to the amount of power concentrated in AI racks. The International Energy Agency reported in April 2026 that AI server power density rose 11-fold from 2020 to 2025 and is projected to quadruple again by 2027. The IEA says a rack could then draw as much power at peak as 65 households. These are projections about peak demand, not a statement that every rack will use that amount continuously.
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 approach 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 NVIDIA’s engineering claims, not independent measurements; the company has said full-scale production with its Kyber systems is planned for 2027.
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 Moves Into the Rack
For data center developers, electricity is becoming a design constraint as much as a running cost. Higher rack demand affects how power enters a building, how it is converted and distributed, and how quickly a site can respond to sudden changes in load. Moving conversion closer to the facility’s edge and using higher-voltage DC may reduce some conversion steps and copper requirements, but the claimed savings need to be weighed against installation costs and the needs of existing sites.
Batteries are also gaining a broader role than backup during outages. AI computing loads can change rapidly; storage may help smooth those swings or provide support during grid disturbances. The IEA estimates that data centers could have 20 to 25 gigawatts of battery capacity by 2030, while noting that the value to the wider grid depends on incentives and operating arrangements. That is a forecast, not a confirmed build-out.
The change matters beyond technology companies. New generation, batteries and grid connections must be planned and paid for, and the allocation of those costs can affect utilities, large customers and other ratepayers. The source material 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 are up 41%. Those figures alone do not show that data centers caused the increases; the specific effect depends on local markets and how projects are financed.
Products Worth Considering
Anker SOLIX C200 DC Portable Power Station, 192Wh, 200W
Samsung 27-inch Odyssey OLED G6 QD-OLED QHD Gaming 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.
Why Conventional Power Designs Strain
Traditional data centers receive alternating-current power and convert it through several stages before servers use low-voltage DC. Transformers, uninterruptible power supplies and distribution equipment each perform part of that process. Conversion losses become heat, which cooling systems must remove. The design has served facilities with racks drawing tens of kilowatts, but the move toward much higher-density AI systems puts more pressure on the electrical path and available space.
Industry proposals are not all the same. NVIDIA describes a facility-level conversion to 800-volt DC, while an October 26 Utility Dive article sponsored by inverter maker SMA America discusses a “sidecar” conversion cabinet that could connect 800-volt racks to existing AC halls. SMA’s white-paper model estimates that a traditional chain supplying 1 megawatt loses about 80 kilowatts, versus about 20 kilowatts with its proposed DC arrangement. Those are vendor-model estimates, not independently verified results from operating sites.
Power equipment and grid access can also take years to secure. The IEA says transformers may take two to three years to arrive and grid connections five to ten years in many locations. Berkeley Lab reported about 2,300 gigawatts of generation capacity in U.S. connection queues at the end of 2024, while only 13% of earlier requests had been built. Queue capacity is not the same as projects certain to be completed, and it illustrates why a data center’s access to power can be a bigger bottleneck than its access to chips.
“A 1 MW rack at today’s 54 volts would take up to 200 kg of copper busbar.”
— NVIDIA, 800 VDC technical blog
Costs and Deployment Still Unsettled
The announcements do not establish when all the generation, battery and transformer capacity will be operating, how much of it will be dedicated to particular data centers, or what the final costs will be. Vantage’s reservation is for factory capacity, not confirmation that a completed transformer has been installed. The NiSource-Microsoft agreement is a 20-year power deal, but the information provided does not detail the contract’s full terms or how its generation and battery will be dispatched.
Efficiency and copper savings also remain partly claims or model results. NVIDIA’s figures are vendor projections, while SMA’s comparison comes from a sponsored white paper. Actual performance will depend on facility design, equipment, operating conditions and the costs of adapting existing infrastructure. It is also unclear how much the proposed DC systems will be adopted beyond early projects.
Finally, broader electricity-price figures do not establish a direct causal link between data center growth and household bills. The effect varies by region, utility planning and contract structure. Who pays for new plants and wires is being resolved project by project, rather than through one universal arrangement.
Grid Deals and DC Systems Move Forward
The next milestones are the delivery and deployment of equipment reserved for new data center projects, further details on the generation and battery capacity tied to the NiSource-Microsoft agreement, and evidence from operating facilities on the performance of 800-volt DC systems. NVIDIA has said full-scale production of its Kyber systems is planned for 2027; that is a company timeline, not proof of broader industry adoption.
Utilities, regulators and data center customers will also need to address connection schedules and the division of costs for new power infrastructure. As projects advance, clearer information about commissioning dates, battery use, electricity contracts and local rate impacts will show whether the new designs can ease power constraints as AI demand grows.
Key Questions
What did Vantage announce?
On October 21, 2026, Vantage reserved more than a gigawatt of factory capacity for a solid-state transformer designed to convert grid power to 800-volt DC. A factory-capacity reservation is not the same as confirmation that the equipment is installed or operating.
What is the NiSource-Microsoft power agreement?
NiSource signed a 20-year agreement with Microsoft that the source describes as adding about 1,100 megawatts of generation and a 600-megawatt battery. Details such as project timing and the full contract terms were not provided.
Why are data centers considering 800-volt DC?
AI racks require increasingly high power. Higher-voltage DC distribution may reduce the amount of copper and the number of conversion stages needed to deliver power, but savings cited by NVIDIA and other vendors are claims or model estimates rather than independent results across operating facilities.
Will data center growth raise household electricity bills?
The source reports higher U.S. average household electricity prices and increased wholesale prices in PJM, but those figures do not prove data centers caused the changes. Local effects depend on power supply, grid investment, contracts and how costs are allocated.
Source: ThorstenMeyerAI.com
