Texas has stopped treating every proposed data centre as guaranteed electricity demand. More than 470 GW of data-centre projects had applied for grid connections in the state, more than five times Texas peak electricity demand. State authorities responded by reviewing proposed projects, their financing and their demands on electricity and water infrastructure.

The problem starts before construction. Developers apply for grid capacity years before some projects receive final financing. Utilities then have to decide how much generation, transmission and substation capacity to build around requests whose completion rates remain uncertain.

Data centres can arrive faster than the grid

A new data centre often takes one to three years to develop. Large power infrastructure takes far longer. Utilities face a timing mismatch. Approving too little capacity delays real projects. Building around every application leaves electricity customers financing equipment for sites never completed.

Developers now face stronger proof requirements. Utilities increasingly ask for deposits, financing evidence and firmer commitments before reserving large amounts of power. Several grid operators have already reduced projected demand after removing speculative projects from connection queues.

Electricity is only part of the bill

A data-centre developer pays for servers, buildings and private infrastructure. The surrounding power system still needs transmission lines, substations and generation.

Governments then decide who carries those costs. Electricity customers can absorb part through tariffs. Taxpayers can subsidise infrastructure. Developers can pay more of the connection and grid-upgrade bill themselves. Texas has moved towards the third option.

Higher infrastructure charges change site economics. Cheap land means less when a developer also funds grid upgrades or accepts limits on power consumption during periods of tight supply.

Cities assessing a proposed data centre now need two numbers: the investment promised on the site and the public infrastructure required to serve it.

A connection queue full of projects provides little information if many never break ground. Completed projects relative to requested capacity offer a better test of real demand.