Global compute demand outpaces supply. Energy and capacity are the constraints. We solve both.
Compute demand outpaces supply globally, particularly in energy-constrained markets. We have access to grid and other energy in those markets. Unlike traditional colos, we are able to reduce our energy cost in any market we operate in. We are able to get capacity in places where capacity is constrained. Depending on the type of site (hotel, hospital, brewery, wastewater treatment plant), we build specifically for the application: brick and mortar, modular, or edge cabinet.
Per-site capex of $80 to 95k installed. Anchor revenue of $8 to 11k per month at 60 to 85 percent utilisation. Per-site cash-on-cash of 38 to 55 percent annualised. Payback inside 9 to 18 months across the modelled cases, project-dependent.
Eleven sites in active development against a market that scales with energy access, not real estate.
We compete on landed cost per kW, on the revenue side via anchor take-or-pay, and on a pathway no incumbent owns.
$12 to 15M per MW landed. Locked to grid pricing in saturated metros.
$80 to 95k per site, roughly $8 to 9.5M per MW. Host-supplied energy reduces marginal cost in any market we operate in.
Wholesale rack lease. Long lead time, capital-intensive presales.
Anchor take-or-pay per site. Cash-positive on installation. MSA framework drafted with one anchor in advanced negotiation.
No formal NCP route. Direct sales path requires hyperscale scale.
Active NCP introduction via Hydra Host. Application-specific spec aligns with the AI Grid roadmap.
Land plus permits plus grid connection. Replicable by any well-capitalised entrant.
Host relationships, application-specific spec, locked-in energy access. Compounds with each signed site.
Forty percent committed against a $5.0M target. $1.6M remains open at this allocation.
The round funds Phase 1 site origination, FEED engineering, host integration design, EPA thermal discharge and local environmental permits, vendor deposits, and long lead procurement. A development fee on close, plus recurring management and tech licence fees, funds team build, BD, legal, and IP work.
This capital does not fund construction equipment, those flow from non recourse project debt at the next raise. It does not fund the lender required debt service reserve account, US rollout sites, host side capex, or HoldCo opex beyond the SPV fee mechanism.
Months 0 to 12 deliver signed Newtown Creek host agreement, FEED and thermal validation, EPA approval, and the next raise trigger of anchor LOIs covering 60 to 70 percent of capacity. EPC contract and lender diligence package close inside Q1 2027.
Conservative, base, and bull case modelled to 2030. MOIC stated against this round entry.
| Scenario | Path | 2028 ARR | 2029 EV | 2030 MOIC |
|---|---|---|---|---|
| Conservative | One anchor only, NCP path slipsSix sites operational by 2028. Edge cabinets defer to 2029. International replication held. | $24M | $180M | 3.6x |
| Base | Plan executes as modelledTwo anchors signed. Eleven sites operational by 2028. NCP partner selected Q4 2026. | $58M | $540M | 10.8x |
| Bull | NVIDIA partnership acceleratesNCP designation 2026. Anchor expansion past initial sites. International replication starts 2028. | $112M | $1.2B | 24x |
Five risks tracked. Each mapped to an active mitigation owner.
| Severity | Risk | Mitigation |
|---|---|---|
| High | Anchor MSA closureBase case depends on a signed anchor. Without one, model reverts to the conservative case. | Two anchors in advanced diligence. Target signature by end of Q3. |
| High | Permitted power timelineSan José anchor needs permitted power by December 2026. Bloom on PPA is the only viable path. | Bloom conversations advanced. Chicago Stickney WRP backup if timeline slips. |
| Medium | NVIDIA partner selectionNCP designation depends on AI Grid partner choice. Introduction in motion, not closed. | Hydra Host introduction live. Plan B via Rafay partner of partner. |
| Medium | First site executionFirst commercial deployment is visible. Any execution miss compounds across the pipeline. | Jacobs as EPC partner under evaluation. AB InBev pilot site (1 MW) de-risks the spec. |
| Low | Capital structure complexityMulti jurisdiction stack adds legal overhead and time to close. | Stripe Atlas interim path. Meridian Stack memo locked. Capital advisor Deniz in seat. |
MicroLink is the physical, thermally coupled, sub megawatt industrial host tier that completes the NVIDIA AI Grid where named orchestration vendors do not reach. Inception is the entry; the three tracks below are the route to NCP.
A hardware first, thermally symbiotic industrial edge node operator that plugs into NVIDIA blessed control planes. We complete the AI Grid at sub megawatt industrial host sites, paired with sovereign and municipal positioning, optimised for PUE alongside ERE.
Not a hyperscaler, not a parallel control plane, not an NCP today. Not a software wedge competing with Armada, Rafay, or Spectro Cloud. Not a distributed inference fabric or a training at scale play.
Primary track owning Telco, Sovereign AI, and the AI Grid initiative. Entry path is a warm introduction positioning MicroLink as the industrial host tier that completes the NVIDIA telco grid in tier 2 and tier 3 markets. Milestone, anchor sponsorship inside the AI Grid programme.
Technical track gating NCP certification, NCX integration, and ISV validation. Entry path is a working session translating a live pod specification, with PUE alongside ERE targets, into the onboarding playbook. Milestone, formal NCP evaluation and designation filing.
Ecosystem track routing MicroLink to ISV partner status for Kubernetes and multi tenant orchestration, with Mirantis k0rdent as the working hypothesis. Entry path is partner enablement intake. Milestone, NVIDIA endorsement that de risks the stack for hosts and capital partners.
Post-funding decisions
This tab is for core team and board level discussions. It documents open questions, decisions in progress, and areas where we need alignment. This section will be removed once the deck and investment structure are finalised.
What liquidation preference structure should we lock in at Series A close?
1x non participating preferred is the baseline. Investor takes 1x back or converts to common, whichever is higher. If the anchor refuses this structure, the round may be mispriced.
What option pool size do we authorise at close?
10 to 15 percent of fully diluted shares for employees, with 12 percent as the working standard. Sized to cover team build through Raise 2 without re expansion that triggers investor consent.
What board composition do we adopt at Series A?
Start with a 3 person board, 2 founders plus 1 lead investor seat. Move to 4 person with a mutually selected independent director before Series B for governance credibility.
Which decisions require investor class consent versus board only?
Investor class consent gates new financing rounds, M&A, debt above $1M, and option pool expansion above 20 percent. Budget, hiring, and operations remain management only.
What contract structure governs the Newtown Creek host relationship?
Long term lease plus service agreement. 20 to 25 year term to support debt amortisation. MicroLink owns and removes all compute and thermal equipment at term end.
How is thermal discharge liability allocated between MicroLink and the WWTP host?
Liability boundaries written explicitly into the host contract. Joint EPA thermal discharge permit signed by both parties. MicroLink carries pollution liability insurance at $5M for thermal injection sites.
What revenue model applies under the host agreement?
Fixed fee per kW plus variable participation in energy savings delivered to the host. The host receives a bankable cost reduction; MicroLink retains tenant revenue and operates all systems.
How do we structure compliance for the WWTP frontier site versus lower risk sites?
Dedicated compliance officer for Newtown Creek given EPA frontier risk and 5 plus site scaling, budgeted at $100k to $200k annually. Shared host model acceptable for brewery and hotel deployments.
Do we adopt JV co ownership for university and strategic partner sites?
Open · pending board.
What insurance floor do we require across the deployed portfolio?
General liability $3M to $5M plus pollution liability $5M at thermal sites plus cyber $1M to $2M. Insurance is belt and suspenders, never standalone, paired with active compliance officer oversight.
Equinix and Digital Realty operate large network dense colocation portfolios with limited heat recovery, mostly in Nordic sites.
A data center company built around thermal integration with industrial hosts, not generic interconnection real estate.
CoreWeave (post Core Scientific) and Lambda vertically integrate AI compute and power on conventional grid sites.
We host AI capacity on host supplied energy with measured thermal recovery, not on bulk grid procurement.
Cloud&Heat and adjacent vendors push toward distributed edge data centers for low latency inference.
We build purpose sited multi megawatt facilities for anchor workloads, not many small distributed nodes.
Heata, Deep Green, Leafcloud, and Qarnot deploy single server or small modular units into homes, pools, and basements.
An industrial scale operator integrated into wastewater and process hosts, not a sub megawatt module vendor.
Meta Odense, AWS Tallaght, Microsoft Fortum Espoo, and Nebius Mäntsälä export heat into established Northern European district networks.
We integrate behind the meter at US wastewater digesters where no district network and no operator currently exists.
Adjacent compute platforms operate behind the meter on host owned generation, with no formal anchor framework.
A data center company taking host supplied energy under anchor take or pay terms, treating the host as partner.
Carrier led AI compute overlays positioned on tower estates, central offices, and edge points of presence.
An application specific industrial build tied to wastewater thermal sinks, not a carrier distributed compute overlay.
Leafcloud markets sovereign data storage with onsite heat reuse; Keppel and Digital Core REIT serve Asia Pacific colocation.
We anchor on US industrial sites with take or pay capacity, not jurisdictional data residency as the primary product.
Acquire land and facilities, build data centers on owned property, collect revenue end to end.
Host owns the facility. MicroLink signs a long term operating contract and owns the compute equipment.
Register an Infrastructure Investment Trust. Investors buy units. MicroLink acts as Investment Manager.
Years 0 to 4 raise as a C Corp. Draft contracts as fund transferable. Year 5 plus sponsor an InvIT or list publicly.
Delaware C Corp HoldCo, Delaware LLC OpCo, per site LLCs owning real property.
Delaware C Corp HoldCo, management OpCo, per site operator entities.
Registered InvIT plus manager corp, independent trustee, per site SPVs.
Delaware C Corp HoldCo, Delaware LLC OpCo, templated SPVs with fund ready language.
Equity at HoldCo plus non recourse debt per site secured by real property.
Equity at HoldCo plus project debt secured by contract cash flows, not property.
Unit issuance, $50M to $100M minimum AUM threshold, mandatory ninety percent cash distribution.
Traditional equity rounds early, project debt per site, pivot to fund structure at exit.
MicroLink owns adjacent land or the building itself.
Land ownership is undesirable and a clean opex model is preferred.
Goal is a dedicated infrastructure fund with a Year 7 acquisition plus IPO exit.
You want simplicity early, flexibility per site, and a preserved acquisition plus IPO pathway at Year 7.
| Site type | Traditional Developer | Contract Holder | InvIT (Fund) | Hybrid |
|---|---|---|---|---|
| WWTP | Acceptable, only if land adjacent | Strong fit, host owns plant | N/A, fund level vehicle | Strong fit, preserves both paths |
| Brewery | Strong fit, often co located land | Strong fit, clean opex inside host | N/A, fund level vehicle | Strong fit, flexible per site |
| Hotels | Weak fit, no land ownership | Strong fit, host owns the building | N/A, fund level vehicle | Strong fit, contract based scaling |
| Universities | Acceptable, depends on campus deal | Strong fit, long contract horizon | N/A, fund level vehicle | Strong fit, public friendly optionality |
Hybrid Model. Start as Delaware C-Corp. Preserve fund optionality at exit.
San José Santa Clara Regional Wastewater Facility, the lead anchor partner. A city owned plant 11 km (7 mi) from NVIDIA HQ. Phased compute heat integration into the treatment process.
Pacific Northwest municipal wastewater partner under active Gate B negotiation. Reject heat integration into local thermal loops at 65 °C (149 °F). Sovereign regional AI footprint with a publicly accountable host.
Ten further partner sites under evaluation across the United States, spanning municipal wastewater facilities, district energy systems, brewing operations, and hospitality estates. Aggregate identified capacity sits in the low tens of megawatts.
San José (First Light site) is our anchor deployment. A 10 MW facility on an 11 km (7 mile) radius from NVIDIA headquarters, integrated into the TPAD digester thermal loop at the San José Regional Wastewater Facility. We worked directly with NVIDIA's Metropolis and Smart Spaces teams to validate the AI Grid reference design for industrial edge compute. The partnership proved the model: host facility thermal integration, modular liquid cooled deployment, and native NVIDIA orchestration stack alignment. This is how we want to build every site.
MicroLink is a Delaware C Corp HoldCo with a Delaware LLC OpCo. Delaware gives institutional investors a familiar venue, clean governance, and friction free Series A mechanics, while keeping per site SPVs templated for later fund use.
Capital stacks at HoldCo through equity rounds, with project debt drawn per site against long term host contract cash flows. Tax equity sits on the table where federal credits apply, layering on top without disturbing the corporate cap table.
From Year 5, MicroLink can sponsor an InvIT, sell to an infrastructure fund, or pursue a public listing. Contracts are drafted fund transferable from day one, so the exit form follows market conditions rather than constraining structure.
Four tiers, four power rate scenarios. Project dependent. Hosts are partners.
Per site capex of $12.0M per MW. Stabilised EBITDA of $24.0M per year, 70.8 percent margin. Anchor MOIC 2.3x on $25M Raise 1.
| Line item | Q3 2026 | Q4 2027 | Q4 2028 |
|---|---|---|---|
| Colocation revenue | $3.96M | $15.05M | $15.05M |
| AI ready revenue | $4.20M | $15.96M | $15.96M |
| Power margin | $0.21M | $0.84M | $0.84M |
| Biogas CHP credit | $0.51M | $2.05M | $2.05M |
| Gross revenue | $8.88M | $33.90M | $33.90M |
| Site OPEX | ($0.88M) | ($3.50M) | ($3.50M) |
| Net power cost | ($1.14M) | ($4.55M) | ($4.55M) |
| EBITDA | $6.39M | $24.00M | $24.00M |
| DSCR (covenant 1.20x) | 0.33x | 1.25x | 1.25x |
Capital is risk staged correctly. Equity first, debt unlocked by contracted partner revenue, successive raises gated on prior phase delivery.
Base case $0.07 per kWh power cost paired with $220 per kW per month colo rate at 95 percent stabilised occupancy. Year 7 HoldCo enterprise value $5.47B at 12x base.
| Line item | 2028 | 2030 | 2032 plus |
|---|---|---|---|
| NYC Newtown Creek | $24.0M | $24.0M | $24.0M |
| Chicago Stickney | $12.0M | $63.4M | $63.4M |
| Seattle West Point | pre online | $10.5M | $14.5M |
| University network | pre online | $3.0M | $7.1M |
| US portfolio EBITDA | $36.0M | $100.9M | $109.0M |
| EMEA | pre online | $45.0M | $120.0M |
| Asia Pacific | pre online | $20.0M | $100.0M |
| Latin America | pre online | $10.0M | $60.0M |
| IP Co royalty | ~$1.0M | $10.0M | $17.0M |
| Global consolidated | $39.0M | $185.9M | $406.0M |
Lead with base case. Frame bull as comp aligned. Founders retain $2.76B at compressed multiples; the NYC anchor still earns 2.3x.
Non recourse project debt at 9.5 percent fixed for 7 years. NYC stabilised debt service ratio computed against $24M EBITDA and $19.19M annual debt service.
| Line item | Y2 2027 | Y3 2028 | Y4 2029 |
|---|---|---|---|
| Closing debt balance | $73.70M | $61.50M | $48.15M |
| EBITDA | $24.0M | $24.0M | $24.0M |
| DSCR | 1.25x | 1.25x | 1.25x |
| Refinance LTV vs $288M EV | 25.6% | 21.4% | 16.7% |
| Anchor cash distribution / yr | nil | $0.96M | $0.96M |
| HoldCo equity distribution / yr | nil | $2.45M | $2.45M |
Address tax and tenant concentration gaps illustratively for Raise 1. Fully model all six gaps before Raise 3 institutional diligence.
Annual thermal value at NYC of $2.89M per year. Strip out thermal and the EBITDA margin compresses from 70.8 percent to about 62 percent. The 8.6 point uplift is the moat.
| Line item | Industry average | MicroLink target | Measured today |
|---|---|---|---|
| PUE alongside ERE | 1.55 | under 1.20 | pre commissioning |
| Thermal recovery rate | 30 to 40% air cooled | 80 to 90% liquid cooled | pre measurement |
| Stabilised utilisation | 85 to 95% | 95% (Y1 ramp 25%) | model target only |
| Capex per MW | $10 to 15M brownfield | $12.0M / MW | tracked in model |
| Revenue per MW | $2.5 to 3.5M | $3.39M / MW NYC | mixed, NYC modelled |
| EBITDA per MW | $1.8 to 2.5M | $2.40M / MW NYC | mixed, NYC modelled |
| Energy cost per kWh | variable | $0.063 / kWh net at NYC | from model |
| Uptime SLA | 99.99% standard | target tba | pre measurement |
Host pipeline detail and host avoided cost are the Raise 1 questions. PUE alongside ERE and thermal recovery validation justify the moat.
Research library indexed in project knowledge base. Available on request.
Industry terms used in this document defined in supporting glossary. Available on request.