🔍 Read the full analysis: How The AI Race Is Changing Data Center Power And Battery Use on ThorstenMeyerAI.com
Get smart everyday buys 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 October developments show how the AI data center buildout is changing power infrastructure: Vantage reserved more than 1 gigawatt of factory capacity for solid-state transformers, and NiSource agreed to a long-term Microsoft power deal that includes a 600-megawatt battery. Operators are exploring 800-volt direct current and battery storage to handle dense, fluctuating loads, but performance benefits cited by vendors are not independent measurements.
Data center power infrastructure is becoming a constraint in the AI buildout: Vantage reserved more than 1 gigawatt of factory capacity for solid-state transformers on October 21, and NiSource signed a 20-year power agreement with Microsoft five days earlier that includes about 1,100 megawatts of generation and a 600-megawatt battery. The deals point to growing investment in power conversion and storage as AI facilities demand more electricity and manage sharper changes in load, as seen in battery cell projects for AI data centers.
Vantage’s reservation is for factory capacity to produce a solid-state transformer that converts grid electricity directly into 800-volt direct current (DC), according to the source report. The reservation is not the same as a confirmed deployment of that capacity at a specific data center. The report identifies Heron Power as the transformer maker and says its design includes storage intended to buffer changes in AI power demand, a concern also reflected in large-scale AI data center buildouts.
NiSource’s agreement with Microsoft is a separate development. The 20-year deal adds about 1,100 megawatts of generation and a 600-megawatt battery, according to the source material. The stated figures describe the agreement’s planned power resources; further details about construction schedules, costs and how the battery will operate were not provided, making deployment tracking for data center buildouts relevant as projects advance.
The shift toward DC is also being promoted by chip and power-equipment companies. 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-demand racks. NVIDIA says its approach 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. These are company estimates, not independently verified results. Full-scale production with its Kyber systems is planned for 2027, according to NVIDIA.
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 Bottleneck
The change matters because the power demand of AI computing is rising not only across data centers but within individual server racks. The International Energy Agency (IEA) says AI server power density increased 11-fold from 2020 to 2025 and is projected to quadruple again by 2027. It says a rack could then have peak demand comparable to 65 households. That projection underlines why operators are reconsidering systems designed for far lower rack loads.
Electricity supply and grid access can take longer to secure than computing equipment. The IEA says transformers can take two to three years to arrive and grid connections can take five to 10 years in many locations. Berkeley Lab reported that US interconnection queues held about 2,300 gigawatts of proposed generation and storage at the end of 2024, while only 13% of earlier requests were built. These figures describe queue requests, not projects certain to be constructed.
The buildout may also affect electricity customers, though costs and effects vary by region and are not established by these two deals alone. The source report, citing the US Energy Information Administration, puts the average US household electricity price at 18.21 cents per kilowatt-hour in 2026, compared with 17.30 cents in 2025. It also reports a 41% rise in wholesale prices in the PJM region. Who pays for new generation and grid upgrades depends on contracts and regulatory decisions.
Products Worth Considering
HP Sprocket Panorama Portable Color Photo and Label Printer
FLEXISPOT 42-Inch Electric Height-Adjustable Desk Converter with Keyboard Tray
Brother TD-2020A 2‑Inch Direct Thermal Desktop Label Printer with USB
As an affiliate, we earn on qualifying purchases.
From AC Server Rooms to Dense AI Racks
Most data centers have historically received alternating current (AC), then converted it several times before servers use low-voltage DC. Each conversion loses some energy as heat, which cooling systems must remove. That architecture was developed for racks drawing tens of kilowatts, rather than the much higher loads envisioned for future AI systems.
An 800-volt DC design aims to reduce the amount of conversion and heavy cabling needed to supply high-power racks. NVIDIA says a 1-megawatt rack supplied at 54 volts could require up to 200 kilograms of copper busbar, and that power shelves could occupy as many as 64 rack units in its planned system. These figures are part of NVIDIA’s technical case for its design, not independent measurements of operating facilities.
Operators may not replace existing electrical systems all at once. A Utility Dive article published October 26 and sponsored by inverter maker SMA America describes a “sidecar” conversion cabinet that could let 800-volt racks operate in halls built around AC infrastructure. Its white-paper model estimates losses of about 80 kilowatts when delivering 1 megawatt through a traditional chain, versus about 20 kilowatts in an external AC-to-DC design. Those are vendor-model figures; the article’s sponsorship and the lack of independent measurements in the source material warrant treating them as claims, not established industry-wide savings.
““leaving no room for compute””
— NVIDIA, describing its 800-volt design
Deployment, Savings and Costs Remain Open
The announcements do not establish how quickly the transformer capacity will become equipment installed at data centers, or when NiSource’s generation and battery resources will be operating. The source material does not provide project locations, detailed construction timelines, total costs or the terms governing how the battery will be dispatched.
It is also unclear how much efficiency operators will gain from 800-volt DC systems in real-world deployments. NVIDIA’s copper, efficiency and maintenance figures are company claims, while SMA’s loss estimates come from a vendor-sponsored model. Independent operating data comparing these systems with existing AC facilities is not included in the source material.
The IEA projects that data centers could have 20–25 gigawatts of battery capacity by 2030. That is a forecast, not installed capacity. The degree to which batteries reduce grid strain or household bills will depend on how they are used, where projects connect, and how electricity and infrastructure costs are allocated.
Projects and Grid Plans Will Test Designs
The next indicators are whether the reserved transformer capacity translates into manufacturing and deliveries, and whether planned generation and battery projects reach construction and operation on disclosed schedules. NVIDIA has said full-scale production of its 800-volt design with Kyber systems is planned for 2027; deployment and performance data will be needed to test its stated benefits.
Utilities, data center operators and regulators will also have to resolve grid connection timelines and decide how new generation, transmission equipment and storage costs are shared. More detail on project approvals, contracts and measured operating performance should clarify whether DC distribution and batteries become common tools for managing AI facilities or remain options used in particular designs.
Key Questions
What did Vantage announce?
Vantage reserved more than 1 gigawatt of factory capacity for a solid-state transformer designed to convert grid power to 800-volt DC. The source report does not say that all of this capacity has been installed at operating data centers.
What is included in the NiSource-Microsoft agreement?
The companies signed a 20-year power deal that adds about 1,100 megawatts of generation and a 600-megawatt battery, according to the source material. Detailed costs, operating terms and project schedules were not supplied.
Why are data centers considering 800-volt DC?
AI racks can require far more power than older server racks. Proponents say higher-voltage DC distribution can reduce conversion steps and the amount of copper needed, but the reported efficiency and cost benefits are vendor claims that still require independent, real-world testing.
Will data center batteries lower household electricity bills?
The available information does not establish that. Batteries can help manage facility demand or grid fluctuations, but effects on household prices depend on local grid conditions, operating arrangements and how infrastructure costs are allocated.
When could the new power designs be used at scale?
NVIDIA has said full-scale production of its 800-volt design with Kyber systems is planned for 2027. The timing and scale of wider deployment, including the announced transformer and battery projects, remain unclear.
Source: ThorstenMeyerAI.com
