Europe · grid-independent compute
Build the datacenter
before the grid arrives.
A grid connection in Europe takes seven to ten years. We simulate seventeen candidate sites hour by hour across a full meteorological year and answer one question: can wind, solar and storage alone hold the load?
- 17
- candidate sites
- 8 760
- hourly points / site
- 3
- IT loads simulated
- 1 %
- viability threshold
Loss-of-load probability is the only honest metric.
Interconnection is the bottleneck
Compute demand moves in eighteen-month cycles. Transmission moves in decades. Off-grid siting decouples the two.
Weather decides, not nameplate
A 300 MWp array in Seville and one in Kiruna are different machines. Only hourly reanalysis exposes the difference.
One number, three verdicts
LOLP under 1 % supports AI training. Under 0.1 % approaches continuous service. Above 10 % the site cannot power itself.
Seventeen locations, ranked by the hardest test we run.
Each row shows two numbers. LOLP · ref build is the share of the year the site cannot power a 50 MW compute load, even with the largest sizing we simulate. CAPEX ≤ 1 % is the cheapest build that keeps outages under 1 % of the year — a dash means no sizing in the grid gets there.
- Loading site index…
Open the instrument.
Rotate the globe, hover a site, dial turbines, solar and storage — and watch the loss-of-load probability move in real time.