Marin.
AI capacity behind the meter.
The Problem
- Demand for AI inference is exploding
- Datacenter construction is gated by power, politics, and siting, not by chips
- Gigawatt campuses take years and land where the grid allows
- 97% occupancy, 77% pre-leased - new supply is spoken for
- Buyers are increasingly latency-sensitive about where inference runs
The next unit of AI supply must be distributed in nature.
“I don’t have warm shells to plug into.”
What is Marin?
- Modular data centers deployed behind the meter in commercial buildings
- An orchestration fabric makes many small sites one schedulable network
- We buy the GPUs, fund the install, pay the power, run the site
- Reserved capacity contracted to labs, clouds, and enterprises
- Idle capacity sold into burst markets
We turn underused commercial power into AI infrastructure.
The Hardware
| GB300 NVL72 72 gpus · 48u | 8 × B300 8 gpus · 8u | |
|---|---|---|
| Power | 132–142 kW | 14 kW |
| Compute | 1,080 PFLOPS | 112 PFLOPS |
| Capital | $3.7–4.0M | $400–500K |
| At 75% | $270/hr | $30/hr |
| At 100% | $360/hr | $40/hr |
| Cooling | Liquid | Air |
Capacity is sized to the building, not the other way round.
how buildings become one product
The Fabric
- Replicas, not shards. Every site serves whole models, nothing splits across buildings
- Routing. Requests reach the nearest warm replica in milliseconds
- Placement. A slower loop moves models by demand, price, and SLA
- Failover. If a site drops, traffic reroutes rather than fails
- One endpoint. Buyers get an API, not forty buildings
Reliability is a property of the network, not the building.
The Host Offer
- The host provides space and access to power it already holds
- Marin funds hardware, install, electricity, and operations
- A guaranteed $50/kW per month, or 5% of post-power revenue if greater
- About $125K a year from a rack in 200 square feet, or $14K from a node in 20
- Recurring AI revenue, no GPU exposure, no datacenter staff
A new NOI line. No capex, no utilization risk, just like cell towers.
GTM
- Launch at 5 pilot sites in Los Angeles (already committed, est. 1.25MW)
- Negotiate anchor LOIs before hardware ships, finalize once we prove performance
- Install 6 racks and 20 nodes; prove latency, uptime, thermals, ops
- Earn aggregator revenue from day one, not just on idle hours
- Repeat city by city
Revenue on day one. Contracts once the record exists.
how we fill the fleet
Demand
- Routers. Listed on OpenRouter from day one. Traffic arrives based on price and latency at zero acquisition cost
- Platforms. Fireworks, Together and Baseten buy regional capacity wholesale - one contract, large volume
- Marketplaces. SF Compute and Vast.ai set the floor beneath backfill
- Anchors. Two or three per metro, founder-led: latency-bound and regulated workloads
- Proof. A head-to-head latency benchmark from inside their metro, not a deck
Supply is scarce. Utilization is a credibility problem, not a demand problem.
per gb300 nvl72 rack · annual
Illustrative Economics
Sell capacity like a utility. Finance it like real estate.
Why now?
- Power, not silicon, is the binding constraint on AI capacity
- Roughly 97 GW of new data-center capacity forecast by 2030
- Data-center electricity use nearly doubles by 2030, to ~950 TWh
- Half of 2025 projects slipped; lead times average 33 weeks
- Modular and micro data centers: roughly $48B a year by 2030
Not replacing hyperscale. Adding a faster layer beside it.
Competitive Landscape
| Segment | Scale | Speed | Local | Private | SLA |
|---|---|---|---|---|---|
| Hyperscalersaws · azure · gcp | ✓ | — | — | ~ | ✓ |
| GPU cloudscoreweave · lambda | ✓ | ~ | — | ~ | ✓ |
| DePIN computeakash · io.net | ~ | ✓ | ✓ | — | — |
| Marinbehind the meter | ~ | ✓ | ✓ | ✓ | ✓ |
Founding Team
-
Amir Haleemceo
Founded Helium and built the world’s largest decentralized wireless network, with more than a million distributed nodes, and millions of daily cellphone users.
-
Frank Mongcro
Founding member and COO of Helium, running sales and business development. Before that, 25 years in security leadership at Palo Alto Networks, Hortonworks, and HP.
-
Charles Kimcfo
Founding partner of Orion Capital. Previously 20 years as EVP at Wells Fargo Commercial Capital as Head of Strategy, Head of Capital Markets, Head of Syndication. Sourced, syndicated, and purchased nearly $50 billion of senior bank debt.
-
Mark Phillipscdo
EVP of business development at Helium, where he drove the AT&T, T-Mobile, and Telefónica carrier partnerships and built the off-grid disaster-relief hotspot program.
Our team built the largest decentralized physical network in the world. Marin solves the same problem, except supply constrained.
Ask
- Raise $20M seed to certify the first commercial unit
- Secure three anchor LOIs and three to five host partners
- Deploy 6 racks and 20 nodes across Los Angeles, equity plus equipment finance
- Live aggregator revenue first, reserved scheduling layered on top
- Prove uptime, cross-metro routing, margin, and refinanceability
A repeatable, financeable metro playbook in 12–18 months.
- Amir Haleem · founder & ceo
- amir.haleem@gmail.com