Power is now the constraint. Model your options before you commit.
Grid queues stretch past your build schedule, hourly CFE commitments are getting real teeth, and every megawatt of redundancy costs money. Levelised lets you lay out grid, on-site generation and hybrid configurations on one canvas and solve them hour by hour across a full year, so time-to-power, reliability and cost per MWh are numbers, not assertions.
Questions this answers
- What does a 90% hourly carbon-free supply actually cost for a 100 MW campus, and how steeply does the last 10% climb?
- Grid connection in three years, or on-site gas plus solar and storage now: which gets the site energised sooner, and at what cost per MWh over the asset life?
- If the grid connection is capped below the IT load, how much battery and firm on-site capacity closes the gap without breaking the reliability standard?
Data centre power decisions fail in the averages. An annual renewable percentage says nothing about the 3 am winter lull when the batteries are flat; a levelised cost that ignores curtailment flatters the solar case. Levelised solves dispatch and capacity for every hour of the year, so the awkward hours are in the answer, not hidden behind it.
Build the whole strategy in one model: grid import with a connection limit, on-site generation, storage, backup, and the load profile you actually expect. Compare configurations by total cost, cost per MWh delivered, and how each one behaves in its worst weeks. When the board asks what the CFE commitment costs, you can show the curve.
It runs on the open-source PyPSA framework with the HiGHS solver (the same class of optimisation used for national grid studies), and everything you build can be downloaded as an open PyPSA file for your own team to interrogate, with a runnable Python script export on the paid tier.
Run the numbers yourself
Sketch your campus, its grid connection and the on-site options, and solve it in the browser. Free to use; create a free account only when you want to save your work.
Or start from the Data Centre example : a complete, solvable model that opens straight in the app, no account needed.