The Four Risks and How They Combine

Faraday's read is that four risks account for most of what goes wrong in data center investments: tenant concentration, power availability, technological obsolescence and interest rates. They are not independent. The same event can trigger several at once, and the interaction is where losses tend to come from.

Tenant concentration is the most visible. A hyperscale building leased to one tenant has one source of revenue. If that tenant consolidates its footprint, renegotiates or fails, the building's income can fall sharply. The risk compounds up the chain. CoreWeave, whose registration statement showed Microsoft as about sixty-two percent of its 2024 revenue, is the example from the first module. A landlord whose tenant is CoreWeave therefore holds indirect exposure to Microsoft's purchasing decisions. Concentration at one level becomes concentration at the next.

The mitigants are known. Longer leases, investment-grade tenants, parent-company guarantees and residual value guarantees of the kind Meta provided in the Hyperion joint venture all reduce the damage if a tenant leaves. Diversification across tenants and markets reduces the probability that one departure matters. Each mitigant has a price, usually in lower rent or a lower valuation, and the investor must decide how much protection is worth paying for.

Power availability is the second risk and, in the current market, often the binding one. A data center without sufficient power is an empty building. The risk appears in two forms. In development, a promised utility connection may arrive late, as Dominion Energy's 2022 notice to developers in eastern Loudoun County demonstrated. In operation, the price and reliability of power can change with grid conditions and regulation. Power scarcity cuts both ways for investors: it raises the value of existing powered capacity and delays new projects. An asset whose value depends on scarcity is exposed if the scarcity ends, and a project that depends on new power is exposed if it does not.

Obsolescence is the risk most specific to this asset class. A data center is designed around assumptions about how much power each rack will draw and how heat will be removed. The artificial intelligence accelerators now being deployed at scale, such as Nvidia's GB200 NVL72 rack-scale system, are designed for liquid cooling and draw far more power per rack than the equipment most existing halls were built for. A facility built for air-cooled servers may need expensive retrofits, or may simply be less attractive to the tenants paying the highest rents. The building does not fall down. It falls behind. Adjusted funds from operations, the measure that deducts recurring capital spending, exists partly to capture this cost, and investors who ignore it overstate returns.

Interest rates are the fourth risk, and they operate through several channels at once. Higher rates raise the cost of construction loans and refinancing. They raise the discount rate applied to future rents, lowering valuations. And they raise the yield on competing investments, lowering the relative appeal of real estate dividends. The Federal Reserve's increases beginning in March 2022 tested every one of these channels. Developers that had locked in long-term fixed-rate debt were insulated; those relying on floating-rate construction loans or near-term refinancing were not.

Now the interactions. Consider a single-tenant building, designed for older equipment, financed with floating-rate debt, in a market where power is abundantly available. If the tenant declines to renew, the owner faces a vacancy, a retrofit bill and a refinancing at higher rates, while competing sites with fresh power offer the tenant an alternative. Each risk alone is manageable. Together they can wipe out the equity. Faraday's read is that data center due diligence is mostly the work of finding which of these risks share a trigger in the specific asset under review.

A practical checklist follows, expressed as questions rather than rules. Who are the tenants, and whose credit ultimately stands behind them? When do the leases expire, and what happens if the largest tenant leaves? Is the power delivered, contracted or merely hoped for? What density and cooling does the building support, and what would an upgrade cost? How is the debt structured, and when must it be refinanced?

None of these questions requires special access. Much of the evidence sits in public filings, utility announcements and deal releases. The skill is in connecting them, and in noticing which answer, if it changed, would change all the others.