Why a Signed Lease Is the Asset

A signed lease is a financial instrument before it is a real estate document. Once that is accepted, the rest of the data center capital market becomes legible.

Consider what a lender or investor sees. A developer has a site, a utility commitment and a signed lease with a cloud provider for a set number of megawatts, a set monthly rate per kilowatt, a fixed term and an agreed annual rent escalation. Before a single wall goes up, the developer already knows, within a narrow band, what the building will earn for the next decade or more. That knowledge has a price. Capital that would never fund a speculative office tower will fund a pre-leased data center, because the risk it is taking is mostly construction risk and tenant credit risk, both of which can be measured.

The terms of the lease therefore drive the value. Four matter most. Term is the length of the commitment. Escalation is the annual increase in rent, often a fixed step written into the contract. Renewal options determine whether the tenant can extend, and on what terms. Termination rights, if any, determine whether the tenant can leave early. A fifteen-year lease with fixed escalators and no early exit is close to a bond. A five-year lease with a termination option is closer to a bet.

Retail colocation has a different contract profile, and investors read it differently. Individual contracts are short, but the customer base is wide, and customers are slow to move once their equipment is installed and connected to partners. Moving means downtime, new cabling and new contracts with every counterparty. The industry calls the resulting behavior low churn. Equinix reports churn, the share of recurring revenue lost to departing customers, in its quarterly filings, and investors watch the figure closely because the retail model depends on it staying low. Here the protection is not the length of any one contract. It is the cost of leaving.

Now add the credit dimension. The value of contracted revenue depends on who is paying. A lease with a tenant that holds an investment-grade rating from agencies such as S and P Global or Moody's can support more debt, at a lower interest rate, than the same lease with a start-up. This is why hyperscale developers compete so hard for the largest cloud and artificial intelligence tenants. The tenant's balance sheet is, in effect, borrowed by the landlord.

There is a newer twist. Some of the fastest-growing tenants are not the long-established cloud providers but specialist compute companies. CoreWeave, which rents access to graphics processing units, or GPUs, disclosed in the registration statement it filed with the Securities and Exchange Commission, or SEC, before its 2025 listing that Microsoft accounted for about sixty-two percent of its 2024 revenue. A data center leased to a company like that is exposed, one step removed, to the concentration of its tenant's own customer base. Credit analysis now has to look through the tenant to the tenant's customers.

The last element is cost. Revenue is contracted; operating costs are partly passed through. Electricity is commonly billed back to the tenant, which insulates the landlord from energy price swings. What the landlord keeps is the spread between rent and the costs it cannot pass on: staff, maintenance, insurance, property tax and the capital required to keep the building current. The ratio of operating profit to revenue in a well-leased hyperscale building tends to be high and stable, which is again the shape of infrastructure.

Put it together and the valuation logic is simple to state. Capital is buying a stream of contracted payments, discounted for tenant credit, for the risk of non-renewal at lease end, and for the capital the owner must keep spending. The building is collateral. The lease is the asset.

Two consequences follow for anyone reading a deal. First, a headline price means little without the lease book behind it: the weighted average remaining term, the mix of tenants, and how much of the capacity is signed rather than merely planned. Second, the same market can support very different valuations for buildings that look physically identical, because the contracts inside them differ. Faraday's read is that the most common analytical error in this sector is to price the building and ignore the paper.

Later modules turn to who owns these contracts and how they are financed. The question to carry forward is narrow and useful. When someone says a data center platform is worth a certain sum, ask how much of that sum is signed.