What We Are Watching
Energy · Infrastructure · AI Demand
The Strategic Move That Won the AI Race Was Not Technological. It Was Perceptual.
A major oil company signed a 20-year power supply agreement with a hyperscaler for an AI data center — repositioning existing capability as the input to a contracted revenue stream in a market that barely existed five years ago. The question is not how to become an AI company. It is what you already control that AI infrastructure requires.
5/20/2026
What We Are Watching is Vaxa's signal intelligence column. We identify markets, technologies, and structural shifts already in motion but may not yet reached the corporate strategy conversation.
highlights
01
A major oil company signed a 20-year agreement to supply 2.67 gigawatts of power directly to a hyperscaler's AI data center
02
The strategic move was perceptual — recognizing that an existing capability had acquired new value before that value was priced into the market
03
AI infrastructure requires power, water, land, transmission, fiber, construction, and maintenance — each is a structural adjacency for companies in relevant industries
04
The window to enter on favorable terms is not indefinite — as constraints become widely recognized, pricing adjusts and negotiating leverage shifts
05
The question is not how to become an AI company — it is what you already control that AI infrastructure requires
Signal
A major oil company has signed a 20-year agreement to supply electricity to one of the world's largest technology companies for an AI data center. The facility — a natural gas-fired power complex co-located directly on the tech company's campus, bypassing the public grid — will deliver up to 2.67 gigawatts of capacity when fully operational.
Read as a technology story, this is straightforward. Read as a strategy story, something more interesting is happening.
The question is not how do we become an AI company. It is what do we already have that AI infrastructure requires — and are we positioned to supply it on terms that reflect what it is now worth.
Pattern
The oil company in this deal is not a technology company. It identified a structural demand signal — AI infrastructure power — and repositioned a core operational capability as the input to a 20-year contracted revenue stream in a market that barely existed five years ago. The deal does not require the company to become something it is not. It requires seeing that what it already is has become newly valuable.
The company did not have to become a technology company. It just had to notice that what it already owned had become infrastructure someone else desperately needed.
Mechanism
Power is the most visible constraint on AI infrastructure growth, but it is not the only one. Data centers require water for cooling. They require land in specific geographies with specific grid access characteristics.
They require transmission infrastructure, fiber connectivity, specialized construction, and ongoing maintenance workforces.
Each of these requirements represents a structural adjacency for a company that currently operates in the relevant industry.
Every one of these adjacencies is a physical asset with a multi-year lead time — which is exactly why the company that moves first tends to be the only one who gets to.
Implication
Software can be built fast. Power plants cannot. Transmission lines cannot. Water rights cannot. The scarcity is real, and the hyperscalers building AI infrastructure have both the capital and the urgency to pay for certainty.
The companies that will look back at this period with regret are the ones that had the relevant assets, recognized the demand signal late, and watched someone else sign the 20-year contract first.
Software scales in weeks. The physical assets AI infrastructure actually needs take years — that gap is the opportunity.
Question
The AI infrastructure buildout is creating a second tier of winners that has nothing to do with building better models — companies whose existing physical assets, land positions, and utility relationships have quietly become some of the most valuable inputs in the technology industry. Recognizing that position before the contract gets signed elsewhere is the actual strategic question.
The company with the land, the water rights, or the power contract already has the hard part done.
The question to ask.
What do you already control that AI infrastructure requires — and are you positioned to supply it before someone else locks in the contract?