VoltaGrid Contracted to Deploy 2.3 GW of Modular Gas Generation

Texas-based on-site power specialist VoltaGrid announced on October 15, 2025 that it will deploy 2,300 MW of modular natural gas generation to support Oracle Cloud Infrastructure’s artificial intelligence data centers across Texas, with firm fuel supply from Energy Transfer’s 140,000-mile pipeline and storage network. The agreement makes VoltaGrid’s total contracted behind-the-meter capacity more than 4,350 MW.

Technology Platform and Architecture

VoltaGrid’s modular platform uses [INNIO](/?p=4923) Jenbacher reciprocating engines paired with ABB automation and control systems. The company describes the configuration as capable of “grid-grade performance with ultra-fast response, zero reliance on battery storage, and near-zero criteria air emissions.” Jenbacher gas engines offer fast-start capability of under 5 minutes from cold start, allowing rapid response to the variable power demands of AI training and inference workloads. Oracle EVP Mahesh Thiagarajan noted that “AI workloads create swings in demand” and that VoltaGrid’s platform is “engineered to dampen these swings, making AI power usage more stable, predictable, and grid-friendly.”

The 2.3 GW deployment follows VoltaGrid’s February 2025 agreement with Vantage Data Centers to deploy more than 1 GW of behind-the-meter gas generation across Vantage’s North American campuses. VoltaGrid also announced in January 2025 a technology partnership with INNIO Jenbacher to co-develop the QPac modular power solution, which can generate up to 200 MW per facility from a standardized containerized platform.

Grid Interconnection Drivers

The Oracle-VoltaGrid deal reflects the behind-the-meter generation market shift driven by multi-year utility interconnection delays. Data center developers report utility grid connection timelines of 2 to 7 years in congested service territories, while modular gas platforms can be operational in 12 to 18 months. ERCOT’s interconnection queue carries more than 225 GW of large load requests, and data center operators seeking power in Texas face both queue delays and transmission upgrade costs that can reach hundreds of millions of dollars per project.

The Oracle-VoltaGrid project in Shackelford County, Texas, will power a facility connected to Project Stargate, the $500 billion AI infrastructure consortium formed by Oracle, OpenAI, and SoftBank in January 2025. The project will operate entirely off the public grid, using VoltaGrid’s modular gas fleet as primary power source. Energy Transfer’s Oasis Pipeline provides firm natural gas, securing fuel supply without exposure to spot market price volatility.

Market Scale and Industrial Context

VoltaGrid projected nearly $1 billion in revenue for 2025, backed by investors including Canada Pension Plan Investment Board and Pilot Co., and a $150 million revolving credit facility. The company’s manufacturing expansion in the Dallas and Houston regions is expected to create more than 400 manufacturing positions and 350 operational roles in Texas. Industry analysts estimate between 2 GW and 5 GW of behind-the-meter generation for data centers could be installed globally by end of 2025, with AI data center power procurement driving the majority of new capacity announcements in 2024 and 2025 combined exceeding all prior years.

Why It Matters

This agreement between VoltaGrid and Oracle Cloud Infrastructure is significant for utility planners, grid operators, and data center developers as it highlights a structural shift in how large power users are approaching interconnection. The deployment of 2,300 MW of modular natural gas generation for Oracle’s AI data centers in Texas demonstrates a growing trend among major power consumers to opt for behind-the-meter generation as a workaround for the multi-year grid queue times and high transmission upgrade costs they face.

For ERCOT grid operators, this development has implications in terms of managing an “invisible load” – power that is generated and consumed on-site without passing through the traditional grid infrastructure. This can complicate load forecasting and grid stability, as behind-the-meter generation reduces the visibility of power consumption patterns. Moreover, the fast-start capability of under 5 minutes from cold start offered by the INNIO Jenbacher reciprocating engines paired with ABB automation and control systems in VoltaGrid’s platform underscores the growing importance of flexible and responsive power generation solutions to support the dynamic power demands of modern data centers.

Critical Perspective

VoltaGrid’s description of ‘near-zero criteria air emissions’ from INNIO Jenbacher reciprocating engines is a vendor characterization, not a regulatory classification; Texas Commission on Environmental Quality New Source Review thresholds for NOx, CO, and VOC from reciprocating internal combustion engines apply to installations of this size, and no permit details or emissions rate figures are disclosed. The 225 GW figure in ERCOT’s interconnection queue includes all project types, not exclusively data center demand, providing incomplete context for the behind-the-meter rationale. Off-grid operation removes Oracle’s facilities from ERCOT’s visibility entirely, creating a single-vendor reliability dependency where a VoltaGrid operational failure has no utility grid backup. The $500 billion Stargate investment figure cited is a planned investment announcement, not contracted or deployed capital as of the October 2025 agreement date.

Related Coverage

• VoltaGrid Orders 1.5 GW of Jenbacher Gas Engines From INNIO: VoltaGrid orders more Jenbacher gas engines from INNIO
• ABB Deploys HiPerGuard Medium-Voltage UPS at 700 MW AI Data: ABB deploys UPS for AI data centers, a key VoltaGrid partner
• DG Matrix Ships First Commercially Available Multiport Solid-State Transformer Aligned With NVIDIA 800VDC AI Factory Spec: Solid-state transformers align with AI factory specs, relevant to data centers
• Amperesand Raises $80 Million Series A Led by Walden Catalyst and Temasek for AI Data Center Solid-State Transformers: Amperesand raises funds for AI data center solid-state transformers