Six flagship campuses. A national foundation for American AI.

Build at the scale America’s AI future demands

American Super Intelligence proposes an initial network of six flagship AI campuses distributed across the United States, each intended to grow into a hyperscale computing center. The ambition is infrastructure at world-leading scale for frontier model development, scientific research, public AI services and separately controlled government missions. Six is the initial flagship target; additional regional service campuses would extend the network as demand requires.

The delivery strategy prioritizes concurrent development across all six locations, early operational capacity and rolling commissioning of substantial capacity additions. Each campus would be planned for sustained expansion and future generations of AI technology. Site-specific scale, workload mix and commissioning schedules may differ while all six retain the flagship ambition.

Geographic distribution would support national reach, operational resilience, access to different energy resources and participation by American communities and workforces. Locations, funding and delivery commitments remain to be established.

Demand shapes capacity and operations

American Super Intelligence must support four distinct sources of demand: LibertyGPT service for the public, Liberty Frontier development and serving, Liberty OSS research and release work, and separately controlled government missions. Each workload has different requirements for computing, storage, response time, availability and security. This planning sits within a much larger national load: Lawrence Berkeley National Laboratory’s 2025 update estimates that all U.S. datacenters could use 9.5% to 15.3% of U.S. electricity in 2030. That forecast does not predict American Super Intelligence demand; it shows why measured workloads and power planning must precede construction.

The capacity plan will measure expected users, peak activity, task length, model requirements and service targets before committing to a build. Public and protected environments will retain separate controls. The six-campus target establishes the ambition; engineering and supplier work will define actual locations, platforms, capacity and national costs.

Two kinds of computing, one continuing responsibility

Model training creates and improves models. It requires coordinated accelerators, fast networks, storage, data pipelines and systems that let teams recover and reproduce experiments.

Inference runs the models people use. LibertyGPT will combine Frontier capability with smaller models for suitable tasks. Its capacity plan must cover simultaneous users, request lengths, reasoning workloads, response-time targets and service availability at national scale.

Frontier’s 2+ trillion total-parameter target requires a distinct architecture and capacity study. Active parameters, precision, model-training methods, memory, networking, power and measured demand will determine the design and costs. Enterprise model-training and serving platforms remain to be selected.

American Super Intelligence will work with American accelerator providers, hyperscalers and infrastructure specialists to define the platforms. NVIDIA and Intel are examples of the intended supplier ecosystem; no supplier has been appointed.

Develop in parallel. Bring capacity online in successive waves.

The proposed strategy prepares multiple flagship sites at once and coordinates power, construction, equipment delivery and staffing. Common designs where practical and repeatable deployment methods would support an aggressive expansion cadence across the network. Completed capacity would enter service while subsequent phases continue.

Early services can use qualified leased capacity or suitable federal facilities while flagship campuses are developed. Time to first useful capacity and time to full campus buildout are separate milestones. An opening deployment is the first phase of a flagship campus, with expansion planned from the outset. Additional regional service campuses would follow demand and service requirements.

Each proposed location must show how computing, power, cooling, water, fiber, construction and staffing work together. Geography and resilience will shape the national portfolio. Public services, Liberty Gov and Liberty Sentinel will retain their own access and data boundaries.

No locations have been selected. Regional proposals must include local engineering evidence and community participation.

Solar, nuclear SMRs and dependable power

Energy planning will include solar alongside nuclear small modular reactors, grid supply and storage. The mix will be chosen for each site’s operating needs and feasible delivery schedule.

• Solar: evaluate rooftops, parking canopies, ground-mounted arrays and off-site supply against local land, weather and connection capacity.
• Storage and grid supply: plan for nighttime demand, changes in generation, maintenance and outages.
• Nuclear SMRs: examine longer-term supply where licensing, fuel, construction and operating arrangements support a credible schedule.

Engineering studies will compare hourly supply and demand, resilience, full costs and community effects. Solar capacity or annual energy totals alone do not establish continuous service. DOE explains the role of solar and storage. The Nuclear Regulatory Commission’s advanced-reactor program tracks designs and licensing activity; American Super Intelligence therefore treats SMRs as a site-specific option to evaluate, not capacity available on demand or on a guaranteed schedule.

Direct work, sustained operations and American supply chains

The infrastructure program would create several kinds of work:

• Construction and commissioning: electricians, pipefitters, equipment installers, controls specialists, civil crews, safety teams and project managers who build and bring sites online.
• Sustained local operations: electrical and mechanical technicians, facilities operators, network and reliability engineers, security teams and service staff who keep each site running.
• National technical teams: model scientists, data engineers, software developers, evaluators, accessibility specialists, educators and user-support teams.
• American suppliers: computing, electrical, cooling, construction, communications and maintenance businesses supporting the full operating life.

Apprenticeships, community colleges and technical education can prepare people for those roles. Site studies will distinguish construction job-years, permanent local positions, shared national teams and indirect supplier activity before publishing employment estimates.

Host communities will help examine utility upgrades, water use, housing, traffic and environmental effects alongside local economic opportunities.

Plan for the full operating life

Facilities and models need maintenance, upgrades and qualified people over time. The infrastructure plan will include hardware refresh, recurring energy costs, incident response, accessibility and public support. Building toward super intelligence means preparing American teams and facilities to support new generations of research and public service as the evidence justifies expansion.

The next delivery step is a coordinated demand and engineering study for the six-campus objective, covering workloads, measured platform performance, site alternatives, staffing, utilities, costs and schedules. That evidence will define opening phases, parallel development opportunities, successive capacity additions and the employment each site can credibly support.

Explore public and regional benefits →