Infra Mews
Start with the daily mews, then scan the board: landlords, hyperscalers, utilities, liquid cooling, and the network fabric underneath the AI buildout.
Today's board
Start with the live board. If you are new to infra, the quick map sits right below.
The physical layer behind AI
A short orientation for people who know the software story but want the infrastructure map underneath it.
AI does not float in the cloud. It sits in racks, pulls power, throws heat, and depends on data centers, networking, substations, and cooling systems that have to be built in the real world.
Models are the brain. GPUs are the muscle. Networks are the nerves. Power and cooling are the circulatory system. Data centers are the body.
- Why this matters now
AI infrastructure is more power-hungry, heat-dense, and capital-intensive than classic cloud, so the bottlenecks are physical first. - What changed
The AI boom is also a boom in rack density, utility demand, liquid cooling, land, and interconnect planning. - What this page tracks
The companies, constraints, and system shifts that decide who can actually scale AI.
Keep the rest of the board warm
Short reads for the operators, suppliers, and bottlenecks worth keeping in working memory.
Names that move the room
The names to keep hot while the sector keeps re-pricing around AI demand.
Equinix + Digital Realty
Where colocation, enterprise adjacency, and premium capacity economics stay easiest to read.
Microsoft + Google + Meta + Amazon
Hyperscaler capex still sets the temperature for the rest of the physical stack.
Vertiv + Schneider + liquid stack
Cooling vendors are becoming central constraints, not background facilities suppliers.
NVIDIA + AMD + cluster suppliers
Chip roadmaps now dictate rack density, thermal assumptions, and deployment windows.
Utilities + onsite power builders
The grid is part of the product, and time-to-power is becoming the gating metric.
Fiber + interconnect + optics
Data movement is now eating enough power and latency budget to deserve top billing.
Keep these open
Trade pulse for daily movement, primary sources for hard signal, and a few strategic reads for structure.
Data Center Dynamics
Power, cooling, campus expansion, supplier moves, and the day-to-day operating pulse.
Digital Realty Investor
One of the clearest windows into landlord posture, global expansion, and capacity logic.
Equinix Blog
Interconnection, AI-ready data center architecture, and how networks shape deployment.
NVIDIA Newsroom
The compute roadmap that keeps changing thermal, rack, and fabric assumptions downstream.
Bloom Energy
Useful for behind-the-meter posture, time-to-power, and the tension between grid and onsite generation.
Bessemer’s AI stack roadmap
A clean strategy frame for reading the physical AI buildout as a layered system.