The Civic Compute Covenant
A covenanted percentage of the facility's compute — at cost, to the host community, for the life of the facility, running with the land. Public compute allocation already works. Cities already have the uses. The covenant joins them.
Three rungs from proven to proposed.
The covenant is a proposal — but not a leap. Each rung below is an established fact; the covenant simply joins them.
- PILOT
Public compute allocation already works.
Governments are already standing up shared compute and handing access to researchers, students, and civic projects. Demand vastly outstrips supply — which is the point: the appetite for public-interest compute is real and proven.
- NAIRR Pilot (US National AI Research Resource)~3.77 exaFLOPS-class pilot supporting ~600 projects and ~6,000 students; demand ran 4x+ over supply, with only 35 of 150+ first-round proposals funded.
- EuroHPC AI FactoriesFree public access to shared supercomputing; ~50% of the JUPITER system managed for public allocation.
- OPERATING
Cities already have compute-hungry civic uses.
Urban digital twins are operating today — running solar planning, flood and heat modeling, emergency routing, and open-data services. The civic demand for compute is not hypothetical; the bottleneck is access to it.
- Virtual SingaporeOperating city-scale digital twin used for solar planning, flood modeling, emergency routing, and care-desert analysis.
- Helsinki 3D+Operating city digital twin for solar, flood, noise, and heat-island analysis, published as open data.
- PROPOSAL
The Civic Compute Covenant — a proposal built on that precedent.
A dedicated percentage of the facility's compute, provided at cost to public entities, for the life of the facility, running with the land. Rungs 1 and 2 prove both halves — public allocation works, and cities have the uses. The covenant joins them. It has no operating precedent as of mid-2026; only adjacent reference proposals exist.
PROPOSAL — no operating precedent as of mid-2026
- FAS / Brookings (reference proposal)Proposed 0.5%-of-gross-revenue community benefit fund tied to large compute facilities.
- Schiavo (reference proposal)Proposed ~1%-of-compute allocation as resident/land credits — characterized as "almost a rounding error" against total facility capacity.
What could your city do with 1%?
Move the slider to your facility's size and pick a civic use. Every number below is an illustrative teaching figure — a feel for scale, not an engineering estimate or an offer.
At 100 MW, a 1% civic covenant could fund roughly 3 × city government AI capacity for a year.
One implementation: LIFE AI City Government ↗ — the covenant itself is vendor-neutral.
Illustrative ExampleLIFE AI models this compute could build
The covenanted compute is the substrate. These are the public-benefit models a city can stand up on it with LIFE AI ↗ — one implementation. The covenant is vendor-neutral: any qualified provider can operate these uses, and the term sheet binds capacity to the city, not to a vendor.
- LIFE AI City Government ↗The city's own operating intelligence — permitting, planning, budgeting, and services on models the city governs.
- VLAS — Verified Living Asset Standard ↗Verification compute for recognizing measured improvement in the living systems of the place.
- LIFE AI Regenerative Development ↗Story-of-Place, Five Capitals accounting, and whole-place design models.
- LIFE AI Learn ↗Compute credits and tutoring models for local students and educators.
- LIFE AI City Twin ↗A living digital twin of the city — solar, flood, and emergency-routing models on one shared model of place.
- LIFE AI Climate Adaptation ↗High-resolution flood and heat-island models that target adaptation where the place needs it most.
- LIFE AI Public Health ↗Population-scale, privacy-preserving analytics for care-desert mapping and outbreak response.
- LIFE AI Civic Assistant ↗Multilingual assistants that make every city service faster to reach.
- LIFE AI Emergency Response ↗On-demand wildfire, flood, and evacuation models for emergency managers.
- LIFE AI Data Trust ↗Governed, community-owned models over data held in public trust.
A term-sheet concept — six elements.
A starting point for a council conversation, not a binding instrument. Six elements define the covenant.
A fixed percentage of the facility's compute capacity — a covenanted share (illustratively ~1%), measured against installed capacity and reserved for public entities. The share is defined up front, not discretionary.
Public entities draw the covenanted compute at cost — the facility recovers direct operating expense, no margin. Access is scheduled through a public queue with priority tiers for emergency and public-health uses.
A citizen governance board (residents, local institutions, and the steward) sets priorities, approves the annual civic-use plan, and represents the community's interest independent of the operator.
A public dashboard reports covenanted capacity, utilization, which civic projects ran, and cost recovered — so the community can see the covenant working (or not) in near real time.
If the covenanted compute is not made available as agreed, defined remedies apply — restored access, community credits, or financial make-good — enforceable by the citizen board.
The covenant lasts for the life of the facility and runs with the land: it survives changes of ownership or operator, binding successors for as long as the facility operates.
Published openly — the six elements above are the complete concept. No operating precedent as of mid-2026.