A data center is silicon wrapped in a financing structure, and the Federal Reserve just raised the price of the wrapper. The central bank hiked its benchmark rate against a backdrop of sticky inflation and faster growth, and the layer of the AI stack that feels it first is not the chip designer — it is the operator borrowing billions to pour concrete, run power, and rack GPUs on a multi-year payback.
The Development
The rate decision arrived with Treasury yields already in focus, and Wall Street enters the week weighing Fed signals alongside oil, semiconductor trade policy, and Trump-Xi talks. For technology specifically, the mechanism is narrow and direct. AI infrastructure is financed with debt and long-lived leases. Every basis point on the curve raises the hurdle rate a colocation build must clear before it earns its cost of capital.
Into that, Rothschild Redburn initiated coverage on two of the largest specialized data center REITs. Digital Realty drew a buy rating with a $227 price target, roughly 25% above where the stock trades. Equinix drew a buy with a $1,261 target, about 23% upside. Those are not defensive initiations in a rising-rate week.
The Revenue Bridge
Rate-sensitive assets normally derate when the discount rate rises. The counterargument in colocation is that pricing power has changed shape. AI training and inference tenants are competing for power-ready capacity, not square footage, and power-ready capacity is the scarce good. When supply is constrained and leases reprice upward, the operator can pass financing costs through to the tenant.
That is the actual test embedded in a 25% price target during a hiking cycle. Either releasing spreads outrun the rise in borrowing costs, or they don't. If they do, higher rates are a margin story rather than a valuation problem. If they don't, the AI capex cycle gets slower and more selective, and the marginal project simply doesn't pencil.
Two initiations in one session, on two names in one subsector, both with targets above 20% upside — issued in the same week the Fed tightened. That is a specific view, not a sector sweep.
Across the Stack
Follow the propagation. At the silicon layer, higher financing costs slow the pace of new deployments but barely touch order books already committed. Broadcom's positioning against its semiconductor peers remains driven by custom accelerator and networking content per rack — content economics, largely insulated from the discount rate over a single quarter.
The squeeze concentrates one layer up, in infrastructure. Operators with existing power interconnects and low-cost balance sheets gain; developers relying on new construction debt lose. That is the split the initiations are implicitly playing.
The consensus beneficiary of any AI-spending story is the chip vendor, and that view is thoroughly priced. The less-crowded read is software that monetizes AI deployments already funded. Morgan Stanley raised its Palo Alto Networks target to $410 from $394, tied to security demand rising as enterprises adopt AI. Securing a workload is an operating expense on infrastructure someone already paid for — it doesn't need new capital to grow.
One correction owed to readers: earlier this month I argued energy prices would become the explicit variable in cloud margins, with calls on Brent in both directions. Neither resolved, and the framing was wrong. The input repricing cloud economics right now is the cost of money.
The Leading Indicator
The argument lives or dies on whether data center pricing power absorbs higher rates. Digital Realty is where that gets settled. A close above $190 within 30 days says leasing economics are winning and the $227 target is credible. A close below $172 in the same window says the discount rate is winning, and by extension that the AI buildout's pace assumptions need trimming.
What would change my mind: a Fed signal that this hike was terminal, or a pullback in Treasury yields. Either removes the pressure and makes the REIT derate a non-event.
Watch releasing spreads on renewed data center leases over the next quarter — the single number that shows whether operators can hand financing costs to tenants.