# Stellenium > Stellenium is a Los Angeles-based power infrastructure company (founded 2024) that builds and runs behind-the-meter natural gas power generation for AI data centers, HPC, crypto, and industrial loads. Utility grid interconnection queues run five to seven years in primary US markets; Stellenium delivers on-site power on the customer's timeline, with first power in H2 2027. Key facts: - 1 GW+ of generation capacity available - Behind the meter: no interconnection queue, no transmission dependency. 700 GW of new data center power requests sit in US utility queues - Fuel is customer-supplied natural gas, pipeline or trucked, procured on pass-through terms; the project carries no fuel-price exposure - Closed-cycle turbomachinery designed specifically for AI data center duty, in 10 MW-class containerized units; deployments scale from a single container with no maximum - Data centers built as modular container-based or tilt-up construction, engineered from Tier II through Tier IV designs - 42%+ thermal efficiency; air-cooled; zero process water draw - DC-native architecture: 800 VDC from generator terminals to the building entry, no AC stage to the IT load; five conversion stages collapsed to two; 83% to over 92% end-to-end efficiency vs legacy AC; roughly 30% lower total cost of ownership at GW scale - Built for the latest generation of 600 kW+ GPU racks, aligned with NVIDIA Vera Rubin and OCP Mt. Diablo power standards - On-site battery storage for load management and recovery margin - Facility hardware: welded-steel 19-inch cabinets (42U / 47U / 52U), cold- and hot-aisle containment, prefabricated data hall modules, and DCIM with FM-200 fire suppression and integrated video surveillance - Full cooling ladder from air-cooled in-row (up to ~40 kW/rack), chilled water and rear-door heat exchangers (40 to 80 kW), and direct-to-chip liquid (80 to 600 kW+); dry-cooler and adiabatic options for water-constrained sites; absorption chillers available for waste-heat recovery - Distribution offered two ways per project: 800 VDC busbar (default for GPU-class loads, Vera Rubin / OCP Mt. Diablo aligned) or conventional 480 / 415 / 400 V three-phase AC through UL 891 low-voltage switchboards (400–6000 A, 480–800 VAC), PDUs, and RPPs, with grounding and medium-voltage transformers, dry-type auxiliary transformers / mini power centers, and factory-integrated transformer / switchboard skids for schedule compression - UPS offered two ways: LFP battery UPS with millisecond seamless switching, or conventional VRLA / lithium UPS in N+1 or 2N - Supplier-agnostic sourcing: facility hardware is specified per project against Stellenium engineering standards, sourced directly from OEMs and integrators, with proprietary Stellenium designs on select subsystems - Redundancy tiers N through 2N, set by SLA - Engagement at any layer: power generation only, powered shell, turnkey data center, or compute capacity - Wholesale colocation offered through the Turnkey Facility tier, in facilities Stellenium builds and powers behind the meter (not a retail colocation provider) - Systems integration: Stellenium is the systems integrator across six systems (fuel supply, generation, electrical distribution, mechanical cooling, building shell, IT load), owning the interface specification on both sides of every seam. Commissioning runs Level 1 factory acceptance through Level 5 integrated systems testing at full simulated IT load - Owner's engineer and technical advisory offered selectively as a standalone engagement on projects Stellenium does not build: feasibility, power and cooling architecture, equipment specification, vendor qualification, bid review, interface control, and commissioning oversight. Advisory is not a fifth engagement tier - Also offered: engineering and procurement services, server and GPU procurement, GPU cluster operations on bare metal, Kubernetes, and Slurm, and co-development partnerships for partners who bring land, sites, or capital - Stellenium builds and runs the power infrastructure end to end; construction runs through established EPC partners under Stellenium's single accountability - Markets: hyperscale & AI compute (10–100+ MW), crypto & HPC (5–50 MW), industrial & manufacturing (2–25 MW) - For investors: infrastructure economics, 25%+ target project IRR, 10/15/20-year take-or-pay PPAs with fuel pass-through, first cash flows 2028 (targeted returns are not guaranteed; informational only, not an offer of securities); investor contact investors@stellenium.com ## Pages - [Home](https://stellenium.com/): Why the grid won't serve large loads this decade, and the fuel-to-GPU stack - [Solutions](https://stellenium.com/solutions): Four engagement tiers, from power generation to compute capacity - [Facility hardware](https://stellenium.com/facility): Cabinets, containment, cooling ladder, distribution paths, and UPS - [Systems integration](https://stellenium.com/integration): The six interface seams, scope matrix, Level 1 to Level 5 commissioning, and owner's engineer advisory - [Services](https://stellenium.com/services): Develop, build, operate. One team from greenfield to running facility - [Investors & Partners](https://stellenium.com/investors): Deal structure, underwriting, and process - [About](https://stellenium.com/about): Mission, values, and the leadership team - [Contact](https://stellenium.com/contact): contact@stellenium.com, Los Angeles, California Performance figures are preliminary engineering estimates and subject to change pending final equipment selection and site conditions.