Power colocation · ERCOT · Texas

Your hardware, energized in weeks.

Standardized battery and compute modules, placed behind existing commercial meters. You bring the GPUs. You keep the GPU-hour.

Reserve capacity See the modules
4-7 yr
typical wait for new grid power
Weeks
from signature to energized racks
$/kW·mo
one take-or-pay contract for power and place
5 hr
battery at full continuous compute load
What Vulcore is

We deploy the physical layer for everyone outside the integrated clouds.

Integrated clouds build sites for their own compute. Vulcore deploys a standardized module with battery, cooling, racks, and fiber behind an existing commercial meter.

The battery earns on the Texas grid and carries part of the facility cost, which is why the rate works. Your hardware stays yours.

From signature to energized racks.
A four-step deployment.

01

Site

A commercial property with spare electrical capacity. Vulcore screens it, contracts it, and shares the energy economics with the host.

02

Module

One standardized unit arrives with battery, cooling, racks, and fiber. Installed in one to two weeks behind a single lockable tie-in.

03

Energy stack

The battery charges off-peak, rides through grid events, and earns on the ERCOT market. That revenue is why the rate works.

04

Your racks

You install your hardware, or we receive it for you. Your customers, your software, your GPU-hour pricing. We bill power and place.

One product family · three sizes

The same manufactured design,
at three densities.

Module family at common scale V-30 3.05 M · AIR · 30 kW V-60 6.06 M · LIQUID · 60 kW VULCORE V-100 12.19 M · DIRECT LIQUID · 100 kW

V-30

The compact module. Edge inference and lighter density.
  • Compute envelope30 kW
  • Battery150 kWh LFP
  • CoolingAir
  • Racks1 × 42U
  • Form factor10 ft skid

V-60

The mid-tier module. Mainstream inference density.
  • Compute envelope60 kW
  • Battery300 kWh LFP
  • CoolingLiquid, pre-plumbed
  • Racks2 × 42U
  • Form factor20 ft ISO

V-100

The full-capacity module. Neocloud colocation density.
  • Compute envelope100 kW
  • Battery500 kWh LFP
  • CoolingDirect liquid
  • Racks3-4 × 42U
  • Form factor40 ft ISO

Every module is the same manufactured design. Sites scale by adding units, and capacity aggregates across sites under one contract and one telemetry view.

V-100RA ruggedized variant of the same platform serves federal and defense sites. Defense →

V-100 · full-capacity module

Forty feet of energized compute.
Built once, deployed everywhere.

V-100 module LIQUID COOLING PLANT 500 kWh LFP STORAGE 3-4 × 42U RACK BAY VULCORE V-100 12.19 M · 40 FT ISO HIGH CUBE
100 kW
continuous compute envelope
500 kWh
LFP storage · 5 hr at full load
30%
thermal headroom above spec
DLC
direct liquid cooling for current-generation rack densities
Bring your own hardware

If it fits the envelope, it racks.

Vulcore modules are power-defined. Fit is a power and cooling question: if your nameplate load lands inside a module envelope, your hardware deploys. Current NVIDIA data-center platforms are supported across SXM and PCIe form factors. AMD platforms on request.

H100H200L40SB200B300CUDA 12.x - 13.x

Rack-scale NVL72 systems exceed the module envelope and are out of scope. Full compatibility reference available on request.

Built for inference

Each module is a complete
serving cluster.

Model cache, on site

An NVMe storage tier in every module. Weights pull once over fiber and serve from local storage, so distributed siting carries no bandwidth penalty for serving.

Fiber that rides the battery

Dual-carrier uplinks with automatic failover. Switching and networking run on the same five-hour storage as the racks, so serving holds through grid events.

Self-contained clusters

GPU fabric inside each node, a module-level spine across racks. Each module is a complete serving cluster with its own power, cooling, and uplink.

Scope: Production inference is the volume workload, and it scales out: add replicas across modules and sites, no cross-site fabric required. Fine-tunes and batch jobs run inside a module. The one exclusion is tightly coupled multi-site training: the trade that sets the rate.
Partner console · Interface illustrative
Security and separation

We energize your hardware.
We never touch its software.

Physical

Anchored steel enclosure, access-controlled and monitored, with a single lockable utility tie-in. Sites are screened for security posture before they are contracted.

Operational

Vulcore operates power, cooling, and connectivity. The compute plane stays under your control: your hardware, your images, your keys. Facility operations never require access to workloads or data.

Telemetry, partitioned

You see what you need to run: power, thermal, uptime, and the carbon-free share of site energy for your Scope 2 reporting. Site economics stay between Vulcore and the host.

Supply chain: module components are screened against federal FEOC restrictions, control-system firmware carries documented provenance, and remote access into the unit is limited to documented, customer-visible paths. Your servers arrive sealed through your own supply chain and move into your racks under recorded chain of custody. We never source them, open them, or touch their software.
Status

First deployments underway in ERCOT Central.

We are reserving capacity for early compute partners now. Early partners get first position on sites, input on configuration, and pricing that reflects being first.

Reserve capacity
Designed and built in Texas
Property hosts

Own commercial property?

A Vulcore module earns your site monthly revenue at zero cost to you. We finance, install, insure, and operate. Your facility gains backup power.

Talk to us about your site

Reserve capacity.
Tell us what you need energized.

Capacity is limited on the first sites in ERCOT Central. Send your load, timeline, and the hardware you have on hand. We will reply with a position and a rate.

Received

Received. We reply within two business days about capacity and timeline.