Building the OTC marketplace for compute.
Compute changes hands in a slow, offline market: deals pass through a chain of middlemen, terms are non-standard, and there's fraud on every side. We're building the marketplace that fixes it — standardizing contracts, running diligence end to end, and bringing liquidity so supply and demand can actually transact.
We work on larger deals — 64 nodes and up, a year or longer — bridging demand (labs, enterprises, compute-heavy startups, and VC funds) to upcoming and in-flight US datacenters and wholesale compute.
Data Centers
Qualified offtakers for your capacity
The compute market is noisy — datacenters get irrelevant inbound that misses the details that actually close a deal. We bring you qualified offtake demand instead: buyers with the scale, timing, and intent to transact, from 1MW to 100MW and above. Most of the capacity we place comes online in 2026, and we also work with 2027 build-outs.
Qualified offtake demand
Buyers are filtered before they reach you — by power, GPU type, term, density, and commercial intent — so the ones landing at your facility are a real fit.
Requirement-led matching
We match around what actually decides the deal: power, RFS, GPU generation, cooling, network, contract term, and how the capacity is structured.
Colocation (BYO-GPU)
Tenants bring their own GPUs. You provide secured power, high-density racks, liquid-cooling readiness, network, redundancy, and operations.
GPU-as-a-Service
Reserved access to deployed GPU capacity — H100 and H200 today through Blackwell-class — for buyers who want compute, not their own hardware to place.
We diligence: available MW · RFS · power topology & redundancy · PUE · DLC / liquid cooling · rack density · network & carriers · tenant isolation · term · pricing.
Resellers & supply
Hardware and rental supply
We partner with authorized resellers to move inventory to vetted demand. We prefer working directly with supply wherever we can.
What we move
Bulk GPUs — H100, H200, B200, B300, GB200, GB300, AMD Instinct — sold outright, rent-to-own, or on offtake.
How deals are structured
Hardware sales and fixed-term rental contracts, on a standardized RFQ, with KYC and a standing reputation record on every counterparty.
Demand
Need compute?
Need capacity at a datacenter, or to buy or rent compute outright? We match you to supply on your terms. We specialize in 64 nodes and up for a year or longer, and bundle smaller demand together for better pricing.
Foundation model labs
Placing large forward reservations — Blackwell, GB200/GB300 — at multi-thousand-GPU scale for training runs.
Enterprises
Internal training and inference at scale, often with security, tenant-isolation, and data-residency requirements.
Compute-heavy startups
Video, agents, biotech, and robotics teams placing forward reservations for training and inference.
Neoclouds & compute platforms
Operators building dedicated capacity and regions, or reserving compute to serve customers downstream.
We qualify: GPU type & generation · node count & MW scale · RFS · term · pricing · region & data residency · decision path.
For VC funds
Guarantee capacity for your portfolio
We partner directly with funds as the offtaker — aggregating portfolio demand to reserve capacity and pass through rates your companies couldn't reach negotiating one by one. One agreement, one counterparty, allocation guaranteed when they need it.
Working with us
How a deal comes together
Reach us , by text at 415-823-5520 or kyle@epilogue.inc.
Reach out
Text, email, or the form — we reply within 24 hours.
Scope + KYC
A short call to understand the deal. We work from a standardized RFQ and run KYC on both sides.
Paperwork
Mutual NDA and a partnership agreement before anything moves.
Close
We run the deal to close, with escrow available to hold funds.
Inquiring on something specific? Browse example live inventory at computedesk.live and reference a listing when you reach out.