Hyperscaler / capex

Why it matters

The ~$620–630B midpoint is the demand-side denominator that the supply-side stories (Dominion 51 GW contracted, Saline $16B closing, Joule 4 GW Utah, AEP's 765 kV Ohio buildout) ladder up against. Two new things vs yesterday's lead: (1) memory/component cost inflation is now an explicit budget driver, which means the per-MW cost of *equipping* a data center is rising fast enough that hyperscalers are choosing to lock in compute via leasing (yesterday's OpenAI/Stargate retreat to bilateral leasing) rather than build-and-own — that durably reinforces the site diligence-platform-for-developers positioning. (2) The Meta Indiana 1 GW announcement adds another rust-belt site to the Ohio/Illinois cluster Hunterbrook flagged as the secretly-Meta $1B Ohio facility — i.e., Meta is concentrating heavily in MISO/PJM-Central, and that's exactly where the Ohio ballot-initiative ban would hit. The capex denominator is durable; the question is which of the 8 functional layers in strategy/competitive-landscape-by-layer.md gets share of that denominator. Cliff's wedge is that ~5–7% of capex (the site diligence fraction) is currently going to hourly consultants with no productized ingest of regulatory text — that's the $30–45B/year addressable layer the regulatory knowledge graph is positioned against.