Asset class 13 · Applications
How mixed-use podium, transit-oriented development, special economic zones, brownfield redevelopment and sovereign programs design ROSE compute in from the start.
Where the building has not been designed yet, the load can be designed for. The film walks the surfaces, the electrical constraint, an exploded view of the panel assembly, and how the income line works for the property.
Press and analysts are welcome to embed this film. Please credit ROSE and link to this page.
Every other category in this taxonomy is about discovering what a standing building can carry. This one is about designing for it before the building exists, which changes the structural answer completely and raises the achievable density on the same footprint. It is also where the international and sovereign programs sit. Special economic zones, industrial city programs and national development initiatives are buying infrastructure at a scale where compute capacity is a policy instrument, not a line item.
The constraint is electrical service, not roof area. Electrical service is the gate, not roof area. Capacity is sized to measured service headroom, which is the service rating less the measured building peak less a design margin. Area is never the binding constraint. A very large roof on a building with a small service hosts very little. Where there is no headroom today, that is a different design rather than a decline.
These are the property types we screen inside mixed-use, land and programmatic. Every one of them gets the same four gate screen: electrical service, structure, host counterparty, then network and thermal.
Designed to be safe for the people in the building, the animals around it, the environment it sits in, and the property itself.
A ROSE panel is digital computing hardware inside a grounded, conductive enclosure. It is designed to the same emissions and immunity standards that govern the servers, switches and building automation already inside occupied space, and the enclosure itself acts as a shield. It is not a transmitter. It broadcasts nothing, it steers no beam, and it produces no ionizing radiation of any kind. Where a site carries radio equipment for backhaul, that equipment is a separate, separately certified device and it is sited and labeled to the rules that govern radio, not folded into the panel.
Every surface a person can reach is designed to be touch safe at normal operating temperature. Energized components sit behind tool access barriers with rated disconnects. Walkable installations are engineered to the walking surface live load, with slip resistance and trip edge treatment. Roof work follows fall protection design, not a promise about behavior.
Enclosures are sealed against nesting and intrusion. There are no exposed hot surfaces, no open blades and no exposed conductors at grade or at roof level. Canopy installations are designed to avoid the reflective glare and the thermal plumes that make some rooftop equipment a hazard to birds.
The thermal loop is sealed and closed, rejecting heat dry. Using liquid inside a loop is not the same as consuming water, and near zero operational water consumption is a design requirement rather than an aspiration. There is no combustion in the panel, no stack, and no onsite process emission.
Panels are designed to the electrical code path that governs the building they attach to, with rated overcurrent protection, listed disconnects, and a documented shutdown that a first responder can find and operate. Fire service access, roof pathways and clearance are part of the layout, not an afterthought.
Mechanically attached mounting is the default assumption on existing buildings, engineered by a licensed structural engineer of record against the actual roof, not a catalog assumption. Where a membrane warranty matters, the attachment path is coordinated with the roofing manufacturer before anything is fastened. Wind uplift, seismic and snow are designed to the governing code edition for the site.
There is no engine and no combustion. Heat rejection is designed to the sound limit the property has to live with, including residential setback ordinances, and the target is that a neighbor cannot tell the difference.
Responsible sourcing and governance are designed in at the bill of materials, not bolted on in a report.
The program is built to know where its materials come from. Suppliers are screened for conflict minerals reporting on tin, tantalum, tungsten and gold, and for cobalt and mica due diligence, using the established OECD framework rather than a self declaration. Preference goes to suppliers who can produce a real chain of custody.
Enclosure, structure and heat rejection are specified in aluminum and steel, which have mature recycling streams and high recovered content availability. Panels are designed for service access and component level replacement so a refresh swaps the compute cassette rather than landfilling the assembly.
Hardware retirement runs through certified electronics recyclers with documented downstream accountability, with data bearing devices sanitized to a recognized media sanitization standard before they leave the site.
The environmental case is the siting itself. Putting compute on buildings that are already built, already served and already connected avoids the greenfield land take, the new substation and the new transmission that a conventional campus requires. Dry heat rejection keeps operational water near zero. Waste heat recovery is designed for on properties that can use it.
Installation, commissioning and maintenance are local trades work by definition, because the sites are wherever the buildings are. Electricians, roofers, ironworkers, controls technicians and network technicians in the community that hosts the site. Apprenticeship and local hire commitments are structured into deployment programs where the host wants them.
Every design decision, number and commercial term the program publishes is recorded in an append only, hash chained ledger inside our own systems, so any figure in any document can be traced to what it came from and when it changed. That is how a diligence reader gets an answer instead of an assurance.
Every panel sits on a revenue metered service inside somebody's franchise territory, which makes the serving utility a counterparty on every single site. The program opens that relationship ahead of deployment rather than treating it as a permit to clear afterwards, and the destination is a standing framework per territory so that any owner served by that utility has a known path instead of a bespoke negotiation per building. Where measured headroom will not carry a design, the second design is non export behind the meter generation paired with storage, sized to daily energy rather than to peak, raised with the utility as a way to let a site go ahead that they could not otherwise fully serve. The customer stays their customer, on their meter.
Because the compute load is curtailable, a field of ROSE sites is dispatchable capacity a utility can lean on during a peak event. Building the control and metering architecture now is what makes that flexibility real later.
The organizations building frontier AI need capacity faster than conventional campuses can be permitted, interconnected and built. The queue for a large single point interconnection is measured in years. A distributed field of sites sized into the electrical service those buildings already have is not in that queue. ROSE is designed to aggregate that distributed capacity into contracted compute offtake for large model developers, cloud platforms, enterprise AI teams and sovereign programs, and to place workloads where latency, jurisdiction or data residency require them to be.
Sovereign and national programs are among the most serious compute buyers in the world, and they want capacity inside their own borders under their own governance. Designing it into what they are already building is the fastest path there is.
Site screening, structural and electrical intake, proposal generation, contracts, deployment scheduling, field service, energy and compute telemetry, revenue share accounting and partner reporting run in one operating system. It is built to sit alongside the systems a large property owner, a utility or a lender already runs rather than to replace them, and it integrates through standard interfaces to the accounting, asset management, CRM, building automation and identity platforms those organizations already own. An owner sees the same numbers we do, on the same day, from the same source.
A surface that produces contracted income is a different asset than a surface that produces nothing. The program is designed so that income line is documented, metered and auditable, which is what a lender needs before it will underwrite against it. The intent is a wide bench of financial partners: banks, credit unions, insurance capital, infrastructure funds and the property owners themselves, so that the same infrastructure that improves a property's economics also expands what a lender can responsibly lend against. Better collateral performance is how small, medium and large businesses get credit.
The United States has an enormous stock of aging commercial, institutional and public buildings, and a national policy priority to rebuild that infrastructure alongside a national priority to lead in artificial intelligence. Those two goals point at the same asset. Compute demand can pay to upgrade roofs, electrical services, controls and envelopes on buildings that already exist, in the communities that already depend on them, instead of consuming new land and new transmission. That is the thesis.
A distributed program creates work in the places it deploys and cannot offshore it. Site surveys, structural and electrical engineering, roofing, steel, electrical, controls, fiber, commissioning, monitoring and ongoing field service, on every site, in the county where the building is. Manufacturing, integration and logistics sit behind that. The design intent is a program that trains and hires from the community that hosts it.
Compute is becoming the input to economic participation, and the places that cannot get it are the places that most need it. Building compute onto buildings that already exist is the only approach that scales into markets that cannot wait on a new substation and a new campus. Sovereign and industrial city programs, university systems, ports and public institutions on every continent have the surfaces and the service. The mission is to make compute capacity something a community can host and benefit from rather than something that gets built somewhere else and sold back to them.
Tell us the property and we will send you a short overview built for it: what surfaces it has, what the electrical service means for capacity, and what the income line could fund. No NDA required.