Frequently Asked Questions
How fast GR1D builds AI data centers, what GPU colocation and compute cost, and how our Texas sites secure power as low as 5.5¢/kWh.
Once power is in place, our modular builds take 6–12 months, versus 18–36 months for a traditional build. Securing power is the long pole: our South Houston 20 MW site already has live power (it was powered on in under 30 days), and the rest of the pipeline is scheduled to go live in 2027 (East Austin 60 MW), Q1 2028 (South Houston 150 MW) and 2029 (North Texas 370 MW). The speed comes from owned land with pre-secured transmission access, on-site substations, and modular construction designed for GPU workloads.
From a single 20 MW facility to 1 GW+ campus-scale deployments. Our current Texas pipeline totals 600 MW across South Houston (20 MW live + 150 MW), East Austin (60 MW) and North Texas (370 MW).
Site acquisition and due diligence, power interconnection and substation construction, modular facility construction (liquid-cooling infrastructure, N+1 redundancy, scalable rack layouts), and full commissioning and handover. Customers can move in their own hardware or use GR1D's direct OEM procurement relationships for NVIDIA GPUs and networking equipment.
Hyperscalers and cloud providers needing 50–500+ MW, sovereign AI programs requiring dedicated secure infrastructure on owned land, AI companies scaling from cloud to owned infrastructure, and strategic partners such as landowners and utilities.
As low as 5.5¢/kWh. We secure long-term power contracts and build dedicated substations on-site, insulating customers from volatile energy markets. Traditional builds typically pay 6–12¢/kWh.
Our Texas-first strategy targets land with pre-secured transmission access in the deregulated ERCOT market, and we construct our own substations. That shortens the multi-year interconnection wait that stalls most U.S. data center projects.
All in Texas on the ERCOT grid: South Houston (20 MW live power plus a 150 MW development site), East Austin (60 MW in development) and North Texas (370 MW, site acquisition in progress).
Yes. Unlike the traditional model of leasing land and power, GR1D owns the land, the power contracts and the facilities.
Bespoke liquid-cooled colocation starts at $180/kW/month at our South Houston site, with high-density racks supporting 50–100+ kW and N+1 redundancy. Power is live; data center modules deliver about 6 months from order.
Power is live at the South Houston 20 MW site today and the site is AI build ready. Liquid-cooled colocation modules are built to your spec and deliver about 6 months from order, so reserve rack space now for GB300 and Vera Rubin deliveries. Reservations are prioritized in the order received.
GR1D takes requests for dedicated GPU servers and clusters for AI inference and training, from NVIDIA H100 and H200 to Blackwell-class B200, GB200 and GB300. Tell us the GPU, count, term and location you need, and we'll source capacity and come back with availability, terms and pricing.
Host Servers is colocation: you bring your own GPU hardware and we provide space, power and liquid cooling from $180/kW/month. Rent Compute is for teams that want GPUs without buying them: tell us what you need, and GR1D sources dedicated capacity and quotes it to you.
By GPU model, server or cluster size, interconnect, term and location. Reserved terms usually cost less per GPU-hour than on-demand, and data-transfer and support terms change the total. GR1D quotes each request in writing before you commit.
Our team responds within 24 hours, and custom build proposals are delivered within 48 hours. You can also book a 30-minute call with our infrastructure team directly from the site, or email info@gr1d.us.
Yes. GR1D works with infrastructure investors and strategic partners across its 600 MW Texas pipeline. Select 'Investment Opportunity' or 'Strategic Partnership' on the contact form and our team will follow up.
A data center takes 18 to 30 months to build once it has power, permits and a design; JLL's 2026 average for a 50 MW facility is 18 months. Power is the longer job: the average wait for a grid connection in primary US markets now exceeds four years, and a large power transformer averages 128 weeks. Realistic data center timelines →
About $17.6 million per MW in 2026, all-in and before GPUs, inside a range of $8.9 million to $23.3 million. Shell and core alone runs about $11.3 million per MW, and Texas builds come in between $9 million and $16 million per MW. Data center cost per MW, line by line →
In twelve phases: site selection and control, power interconnection, entitlements and permits, design, long-lead procurement, sitework, shell and core, MEP fit-out, five levels of commissioning, utility energization, IT installation and operations. The building takes 18 to 30 months; power sets the schedule. The 12 phases, step by step →
Somewhere between 20 and 40 kW per rack. Today's AI racks sit far beyond that line: a GB200 NVL72 draws 125–135 kW and sends more than 90% of its heat into liquid, and Vera Rubin NVL72 is specified to 250 kW on 45 °C coolant. Liquid cooling for GPU data centers →
An 800 VDC data center converts utility AC to 800 volts DC once, at the row or the facility, and distributes DC to racks that step it down to GPU voltages. Sidecar hardware ships in 2026 and native 800 VDC buildings follow in 2027–2029, and most of what they need is poured, hung or trenched when the shell is built. 800 VDC and Vera Rubin, explained →
As of October 14, 2026, a data center of 75 MW or more is a large load in ERCOT: it files with the utility, pays a $100,000 study fee, posts $50,000 per MW of security before ERCOT studies it, and can't energize without ERCOT's approval. Below 75 MW, the utility handles it. ERCOT large-load interconnection, gate by gate →
Talk to our infrastructure team — we respond within 24 hours and deliver custom proposals within 48.
Or email info@gr1d.us
AI data center infrastructure — from raw land to GPU-ready racks. Build, host, or rent compute at scale.
South Houston, TX
© 2026 GR1D. All rights reserved.
AI Data Center Infrastructure — Build, Host, Compute