Infrastructure economics, exposed

Model the AI multiplier.

Trace one dollar of AI infrastructure spend into supplier revenue, then apply a sales multiple. The headline is simple. The assumptions are where the story lives.

Spend× capture× sales multiple

One clean equation three debatable inputs.

The post's headline case

At 100% revenue capture and a 10× sales multiple, every $1B of infrastructure spend maps to $10B of implied supplier value.

10×

Capture is not automatic

Power, land, construction, networking, chips, and cooling split the spend. Lower capture to model only the revenue that reaches the supplier group you care about.

Multiples move

Price-to-sales is a market assumption, not an accounting identity. Growth, margins, rates, and risk can compress or expand it without changing current revenue.

ChipsAccelerators and memory often carry different margins, supply constraints, and replacement cycles. Data centersBuildings, networking, cooling, and operations turn silicon into usable compute capacity. PowerGrid access and generation can become the binding constraint before demand for compute does.

Stress-test the bridge, not just the headline.

Headline case

$10B spend, 100% capture, 10× sales. This reproduces the post's clean $100B implied valuation.

Shared wallet

Model a supplier group that captures 65% of spend and trades at 7× sales. The implied value falls to $45.5B.

Scaled buildout

Increase spend to $25B, hold 80% capture, and apply 12× sales. The model produces $240B.

“The arithmetic is easy. Deciding whose revenue the spend becomes is the real model.”

“A multiple amplifies a revenue assumption; it does not make the assumption more certain.”

“Use the export as a record of the case you tested, not as a forecast.”

Keep the math. Keep the caveat.

Export the exact inputs, formulas, and outputs behind your scenario. This model illustrates a valuation mechanism; it is not a market forecast or investment advice.