Small caps AI stocks

Most of the small-cap AI trade is sentiment. The names that move have a story about AI but no revenue tied to it. I want the opposite: small and mid caps whose reported segment revenue or backlog is already contracted to the datacenter buildout.

The mechanism is that hyperscaler capex is committed twelve to twenty-four months ahead through orders and backlog. That spending lands first on rack-level physical infrastructure — optical interconnect, power distribution, thermal management, and the electrical contractors who install it. These suppliers book the work as backlog before it prints as revenue, so backlog growth leads the reported number rather than following it.

The margin is captured by suppliers and installers that are capacity-constrained and therefore have pricing power. It is not captured by the software names whose AI exposure is narrative, and not by regulated utilities, whose allowed returns are capped no matter how much load arrives.

Horizon is twelve to twenty-four months, set by equipment lead times and construction schedules, not by sentiment.

Exclude any company whose AI exposure is narrative only — no disclosed segment revenue or backlog tied to datacenter or AI infrastructure. Exclude market caps under one billion dollars and average daily volume under ten million dollars, because the original basket filled up with names too small and too thin to be investable.

I am wrong if hyperscaler capex guidance is revised down for two consecutive quarters, or if equipment lead times shorten materially from current levels.