Separate the layers, price the risk.
An AI campus is not one asset but a stack of three, each with a different risk profile and a different natural owner. At the bottom: land, interconnected power and the built shell — slow to create, durable for decades, and secured by long-term contracts. In the middle: the IT layer — GPUs and network gear that depreciate in three to five years and belong on the balance sheet of whoever runs the compute. On top: the model and product layer, which is genuine venture risk.
Why should a tenant care? Because when one balance sheet carries all three layers, the building your GPUs depend on is entangled with venture risk it has nothing to do with. The discipline that separates the layers is old and unglamorous — it is the discipline of institutional real estate — and it is what keeps the facility beneath your compute stable for the life of your deployment.
Why triple-net fits AI.
Under a triple-net lease, the tenant pays power, operating costs, and maintenance directly, and rent is a clean, contracted figure. Applied to AI campuses, the structure does three things for the tenant at once:
- It gives you security of tenure. A 15–20 year lease with defined escalators mirrors the useful life of the shell and its electrical backbone — your site, your power position, and your expansion path are contracted for the life of your platform, not the refresh cycle of the chips inside it.
- It puts you in control of the biggest cost line. You choose the workloads and you carry the energy bill directly — no blended tariff, no landlord mark-up on your power. The most consequential line item stays in your hands.
- It makes cost transparent. Rent, escalators, and pass-through costs are defined at signing. There are no hidden service layers to be repriced mid-term — what you pay is what the contract says, for fifteen to twenty years.
Under a 15–20 year triple-net lease, your data centre behaves like core infrastructure — not like a venture bet.
Why your landlord's financing matters to you.
Lenders increasingly treat data-centre leases like other long-term contracted infrastructure: they assess the lease and the landlord, not the resale value of GPUs. For a tenant, that is quietly good news. A landlord whose financing rests on long-term leases has every incentive to deliver on time and keep the facility to specification — and a solid financing base is what turns your expansion options from a slide into a schedule.
The land-and-power layer compounds this: as interconnection queues lengthen, a secured power position becomes ever harder to replicate. For tenants, that is the layer to scrutinise before signing anywhere — it decides whether the next phase of your capacity actually exists.
What this means at Castellan.
Castellan applies exactly this logic. As a Swiss stock corporation, we develop, build and own our campuses — normally on our own balance sheet — and lease them on 15–20 year triple-net terms: as a powered shell for your own fit-out, or as a turnkey data centre where you only bring your GPUs. Our pipeline reaches at least 800 MW by 2032 in four phases, with access to more than 5 GW of power through our network. For you as a tenant, that means defined lease terms, transparent costs and capacity delivered on a contracted schedule.