AIThis post was created with the assistance of artificial intelligence (AI).

🔍 Read the full analysis: Data Center Power For AI: Why DC And Batteries Are In The Mix on ThorstenMeyerAI.com

Before you orderOffer from Amazon

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
Start your free Prime trial Free trial for eligible customers · Cancel anytime
As an affiliate, we earn on qualifying purchases.

TL;DR

Two October developments underscore the changing power needs of AI data centers: Vantage reserved more than a gigawatt of factory capacity for solid-state transformers, while NiSource signed a 20-year deal with Microsoft involving new generation and a 600-megawatt battery. The deals point to growing interest in direct-current distribution and storage, but do not by themselves establish how widely those designs will be adopted or who will ultimately pay for grid upgrades.

Data-center power infrastructure is becoming a focus of AI expansion, as Vantage reserved more than a gigawatt of factory capacity for solid-state transformers and NiSource signed a 20-year power agreement with Microsoft that adds about 1,100 megawatts of generation and a 600-megawatt battery. The deals show operators and utilities addressing electricity supply and distribution alongside computing equipment, though they do not establish that all AI facilities will switch to direct-current power or use batteries in the same way.

Vantage’s October 21 reservation is for manufacturing capacity for a solid-state transformer that, according to the source report, converts grid power directly to 800-volt direct current (DC). The reservation concerns factory capacity; the available material does not specify how many transformers will be delivered, where they will be installed, or when they will enter service. Heron Power is identified as the transformer maker.

Five days earlier, NiSource signed a 20-year agreement with Microsoft involving approximately 1,100 megawatts of generation and a 600-megawatt battery. The agreement links a large technology customer’s power needs with utility generation and storage. The source material does not give the project’s full construction schedule, cost allocation, or the battery’s expected operating role.

The technical rationale is that AI racks need far more power than conventional server racks. The International Energy Agency says AI-server power density rose 11-fold from 2020 to 2025 and is projected to quadruple again by 2027. NVIDIA’s reference design proposes converting 13.8-kilovolt grid power to 800-volt DC near the facility perimeter and distributing DC to high-power racks. NVIDIA says this could reduce copper use by 45% compared with 415-volt AC distribution and improve end-to-end efficiency by up to 5% compared with current 54-volt systems. Those are vendor claims, not independently verified results.

At a glance
reportWhen: Developments announced October 16 and O…
The developmentVantage reserved over a gigawatt of factory capacity for solid-state transformers, and NiSource signed a 20-year Microsoft power agreement that includes generation and battery storage.
Thorsten Meyer AI
Power architecture

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

  1. Medium-voltage AC from the grid
  2. Transformer steps it down to low-voltage AC
  3. UPS converts to DC for the batteries and back to AC
  4. Distribution units feed the racks
  5. 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

  1. 13.8 kV AC from the grid
  2. Converted once to 800 V DC at the edge of the site, batteries on the same DC bus
  3. 800 V DC busway to the racks
  4. 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

  • AC chain, conversion at the racktraditional design~80 kW
  • 800 V DC from outside the buildingmedium voltage converted once~20 kW

What NVIDIA says 800 V DC changes

NVIDIA’s engineering claims claim

−45%copper, vs 415 V AC distribution
+85%power through the same conductor size
≤5%better end-to-end efficiency vs 54 V racks
200 kgcopper busbar for one 1 MW rack at 54 V

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 Meets AI Demand

The development matters because electricity supply and equipment lead times can constrain data-center growth, even when operators can obtain the chips and buildings they need. Higher rack power can require thicker conductors, more electrical conversion equipment and more cooling. Moving conversion closer to the grid connection and distributing higher-voltage DC is one proposed way to reduce the number of conversion steps and the amount of copper needed.

Batteries may also serve a role beyond backup during outages. AI workloads can change their electricity draw rapidly, and storage could help smooth those changes or support a site during grid disturbances. The IEA estimates that 20–25 gigawatts of batteries could be installed in data centers worldwide by 2030. That is a projection, not a confirmed buildout, and the IEA says their value as grid assets depends on incentives.

These choices could affect power-system planning and electricity costs. The source material reports that US residential electricity prices average 18.21 cents per kilowatt-hour in 2026, up from 17.30 cents in 2025, and that wholesale prices in the PJM region are up 41%. Those figures do not show that data centers caused the increases. Who pays for new generation, transmission and local infrastructure depends on contracts and regulatory decisions, not simply on a facility’s use of DC power.

Products Worth Considering

Jackery Explorer 1000 v2 Portable Power Station, 1024Wh LiFePO4 Battery

Jackery Explorer 1000 v2 Portable Power Station, 1024Wh LiFePO4 Battery

LG 27GS60QC-B Ultragear 27-inch Curved Gaming Monitor (QHD, 180Hz, 1ms, 1000R)

LG 27GS60QC-B Ultragear 27-inch Curved Gaming Monitor (QHD, 180Hz, 1ms, 1000R)

LG UltraGear 27-inch QHD OLED Gaming Monitor, 240Hz, 0.03ms, G-SYNC Compatible, FreeSync Premium Pro, HDR True Black 400

LG UltraGear 27-inch QHD OLED Gaming Monitor, 240Hz, 0.03ms, G-SYNC Compatible, FreeSync Premium Pro, HDR True Black 400

As an affiliate, we earn on qualifying purchases.

From AC Server Rooms to 800-Volt Racks

Most data centers have traditionally brought in alternating current (AC), transformed it to lower voltages, and converted it to DC for servers. Each conversion can waste some energy as heat, which cooling equipment must then remove. That established design was developed around racks drawing tens of kilowatts, rather than the much higher loads projected for future AI systems.

The IEA estimates global data-center electricity use at 485 terawatt-hours in 2025 and projects about 950 terawatt-hours by 2030, roughly 3% of worldwide electricity demand. Separately, Berkeley Lab estimated that US data centers used 4.4% of national electricity in 2023 and could account for 6.7% to 12% by 2028. These are estimates and projections with different geographic scopes and time periods.

NVIDIA’s 800-volt design is planned for full-scale production with its Kyber systems in 2027, according to the company. An industry article published by Utility Dive on October 26 describes a possible interim approach: a “sidecar” conversion cabinet placed next to a rack, allowing higher-voltage DC equipment to operate in a hall originally built around AC. The article was sponsored by inverter maker SMA America, and its modeled energy-loss comparisons should be read as a vendor-supported case rather than independent field measurements.

“Leaving no room for compute”

— NVIDIA, in its 800 VDC technical blog

Costs and Deployment Still Unclear

The announced reservation and utility agreement do not answer how quickly the equipment and power projects will be built, how much they will cost, or which customers will bear those costs. Details about Vantage’s reserved manufacturing capacity, including delivery timing and deployment sites, are not provided in the source material. NiSource’s deal is described as a 20-year agreement, but the available account does not specify how its generation and battery will be financed or dispatched.

The performance case for 800-volt DC also remains partly prospective. NVIDIA’s efficiency and copper figures are company claims, while the loss comparison in Utility Dive comes from an SMA-sponsored model. Real-world results may vary by facility design and operating conditions. It is also unclear how quickly operators will replace existing AC infrastructure rather than use sidecars or other staged approaches.

Grid connection delays and equipment queues may further shape adoption. The source cites IEA estimates that transformers can take two to three years to arrive and grid connections five to ten years in many locations. Berkeley Lab data put US interconnection requests at about 2,300 gigawatts at the end of 2024, while only 13% of earlier requests were ultimately built. Those figures describe broad constraints, not the delivery outlook for either named project.

Project Timelines and Grid Plans

The next concrete signals will be project-specific: when Vantage’s transformer capacity becomes equipment deliveries, and how NiSource and Microsoft schedule the generation and battery work under their long-term agreement. For NVIDIA’s design, the company has said full-scale production is planned with Kyber systems in 2027; deployments and independently documented operating results will show whether the projected efficiency and material savings hold in practice.

Utilities, regulators and data-center customers will also need to clarify how new power plants, batteries and grid connections are funded and operated. Further information on contracts, construction milestones, local approvals and rate treatment will help show whether these arrangements add capacity without shifting costs unfairly to other electricity customers.

Key Questions

What happened in the Vantage and NiSource deals?

Vantage reserved more than one gigawatt of factory capacity for solid-state transformers, according to the source report. NiSource signed a 20-year agreement with Microsoft involving about 1,100 megawatts of generation and a 600-megawatt battery.

Why are data centers considering 800-volt DC power?

AI racks are expected to draw far more power than older server racks. Higher-voltage DC distribution could reduce conversion steps and copper requirements, but savings described by NVIDIA are vendor estimates rather than independently verified results.

Will batteries power AI data centers continuously?

The reported projects do not establish that. Batteries can provide backup, help manage rapid changes in demand, or support grid operations, depending on their design and contracts. The source does not specify the operating plan for NiSource’s battery.

Do the deals mean household electricity bills will rise?

They do not establish that outcome. The source reports rising US residential and PJM wholesale prices, but does not show that these particular agreements caused the increases. The effect on customers depends in part on how new generation and grid costs are allocated.

Source: ThorstenMeyerAI.com

You May Also Like

Top 10 Mirrorless Cameras In 2026 That Leverage AI Innovation

Discover the 10 best mirrorless cameras of 2026 that incorporate advanced AI features, transforming photography and videography for all users.

Measuring Benchmark Optimization In Speech Recognition

Hugging Face researchers introduce tests revealing how some speech models overfit benchmarks, impacting real-world transcription reliability.

10 Best Mobile Workstations For AI-Focused Work In 2027

A source-based comparison of 10 mobile workstation configurations, with attention to GPU, memory, portability and limits in the available specifications.

ByteDance GenAI R&D’s Recent Departures Put Team Changes In Focus

36Kr reports ByteDance’s GenAI R&D head is leaving for embodied intelligence and says three core members departed in four months; details remain limited.