Flight Deck -- Open Source AI Infrastructure Economic Model
Flight Deck is a bank-grade, scenario-driven economic model for data centers and Energy AI infrastructure that is 100% free and open source. It connects power strategy, chip fleet composition, pricing, and financing so you can see -- in one place -- how engineering choices translate into cash flows, asset returns/equity returns, and lender-ready coverage metrics.
What it’s for
- Building and comparing unit economics for GPU clusters across training and inference use cases
- Stress-testing deals before you commit to power, hardware, or contracts
- Aligning developers, operators, lenders, and offtakers on a single source of economic truth
What it models
- Power: behind-the-meter, on-grid, hybrid, or fully off-grid; configurable tariffs, fuel curves, capacity factors, curtailment, and ancillary costs
- Hardware: from A100-class through next-gen accelerators and LPUs; mixed fleets and phased buildouts; utilization profiles by workload
- Revenue: on-demand pricing, 1–3 year (or custom) terms, prepayment structures, take-or-pay and minimums, and blended go-to-market mixes
- Capital: equity-only, blended debt/equity, and max-leverage variants; cost of debt, tenor, fees, reserves, covenants, and coverage thresholds
- Operations: site opex, networking, cooling, maintenance, spares, downtime and derate assumptions, ramp schedules, and overhead
What you get out
- Transparent cash waterfalls (project and equity), IRR/NPV, payback, and lender-facing ratio packs
- Per-MW, per-rack, and per-GPU unit economics; margin bridges from chip-level ROIC to asset-level returns
- Sensitivity and scenario views for power price, utilization, pricing, capex, and financing terms
- A concise summary that explains the “Watt-Bit Spread”: why great chip returns do not always become strong asset or equity returns -- and how to close the gap
Why Flight Deck
- Bank-grade structure with clear audit trails and assumption isolation
- Built for operator speed: change a few cells, see the whole deal re-underwrite
- Open and extensible: use your own price curves, fleet roadmaps, and offtake structures; integrate with OVERCLOCK physics outputs to ground economics in real operations
- 100% Free (CC0 license)
What’s included
- A structured Excel workbook with tabs for assumptions, power, fleet, revenue, opex, capex, financing, outputs, scenarios, and sensitivities
- Starter templates for common architectures (grid-tied, BTM gas, hybrid, and staged expansions)
- In-sheet documentation and flags to prevent common modeling mistakes
- Proprietary chip pricing data feeding price decline curves
Requirements
- Microsoft Excel (Microsoft 365 or recent desktop versions)
- No specialized plugins; stays within widely supported functions
Use Flight Deck to turn design ideas into lender-ready numbers -- and to ship better AI infrastructure deals, faster.
"Yes, the Watt-Bit Spread is wide -- but great chip returns don’t always translate to positive asset or equity returns. Why? Hop on the Flight Deck and see for yourself."
FAQs
What exactly is Flight Deck?
A scenario-driven Excel model that links power, hardware, pricing, and financing. It produces transparent cash flows, coverage ratios, and returns suitable for internal ICs and lender reviews.
Who is it for?
Developers and operators of AI infrastructure, corporate and project finance teams, lenders, and offtakers who need a shared, auditable economic view.
How is this different from OVERCLOCK?
OVERCLOCK simulates the physics and operations of AI infrastructure. Flight Deck turns those operational configurations into economics—unit costs, cash flows, and returns. Many users export utilization and performance outputs from OVERCLOCK and reference them as inputs in Flight Deck.
Which hardware can I model?
Anything from A100-class to current-gen GPUs and LPUs. You can mix accelerators, set phased procurement, and assign utilization per workload (training vs inference).
Can I model different power strategies?
Yes. Model on-grid tariffs, behind-the-meter generation, fully off-grid, or hybrids (e.g., BTM plus grid import). Add capacity factors, curtailment, fuel curves, start costs, and ancillary charges.
How do I represent grid constraints or curtailment?
Use capacity factors, availability windows, or explicit caps on deliverable MWh. You can run sensitivities on curtailment to understand revenue and coverage impacts.
Does it include nodal price forecasting or dispatch optimization?
No. Flight Deck consumes your price assumptions; it does not forecast nodal prices or co-optimize dispatch. For that, plug in your own forward curves or outputs from a market model.
How are revenue models handled?
Choose from on-demand pricing, fixed-term contracts (e.g., 1–3 years or custom), and offtaker prepayments. You can blend models and add take-or-pay minimums or floors.
Can it model colocation versus owner-operator economics?
Yes. You can model the entire value chain from “sell compute” to “sell power/space”. The structure supports either posture.
Can I reflect offtaker credit or prepayment structures?
Yes. Prepayments can be applied to reduce upfront equity or to fund reserves; timing and recognition are configurable. You can add credit haircuts via discounts or minimum availability penalties.
What financing features are supported?
Equity-only, blended, and max-leverage cases. Configure debt tenor, interest, fees, sculpting rules, and reserve accounts Sensitivity tables let you test cost of debt and leverage headroom.
How do I bring my own price curves (chips, power, HPC)?
Paste or reference your forward curves into the assumptions tabs. The model separates inputs from logic so you can refresh curves without breaking formulas.
Can I phase construction and procurement?
Yes. You can stage capex, commission in separate projects, and ramp utilization by cohort. This supports brownfield repowers and stepwise expansions.
What outputs are lender-ready?
Cash waterfalls (project and equity), IRR/NPV, payback, and covenant headroom. A summary page rolls up key assumptions, sensitivities, and flags.
How do I run sensitivities?
Use built-in sensitivity tables for key variables (power price, utilization, pricing, capex, leverage). Save named scenarios and compare them side-by-side.
Does the workbook use macros?
The core model is designed to run on standard Excel functions. If you add macros for your workflow, keep logic separate from assumptions.
What does Flight Deck not do?
It does not forecast nodal prices, perform power-flow studies, or replace site engineering and interconnection analysis. It is an economic layer that depends on your technical inputs.
Can I integrate with other tools?
Yes. Most users link CSV exports from operational simulators (e.g., OVERCLOCK) and market models directly into the assumptions tabs.
What’s the best way to start?
Pick a starter template (grid-tied, BTM gas, hybrid), paste your fleet and power curves, load an initial pricing strategy, and run three scenarios: Base, Downside, and Lender Case.
What license applies and can I use it commercially?
Flight Deck is CC0 and can be modified and used commercially without attribution. Have at 'er!
What about data privacy?
All calculations are local in Excel. Your assumptions and price curves stay in your environment.