Intelligence DisruptMining Disrupt Buy ticketsTickets

AI infrastructure glossary

The words of AI infrastructure, in plain English: 20 terms on power, cooling, capacity and the companies that build it. Each has its own page.

Behind-the-meter power
Behind-the-meter power is electricity a data center takes directly from generation on or beside its site, without passing through the utility's meter.
CDU (coolant distribution unit)
A coolant distribution unit is the equipment that circulates coolant to liquid-cooled racks and passes their heat to the building's water loop.
Colocation
Colocation is renting space, power and cooling in someone else's data center for equipment you own.
Data center tier rating
A tier rating is the Uptime Institute's four-level classification of how much redundancy a data center has, from Tier I to Tier IV.
Direct-to-chip cooling
Direct-to-chip cooling runs coolant through cold plates mounted on the processors and GPUs, taking heat off at the source.
GPU cluster
A GPU cluster is a group of servers with graphics processors, connected by a very fast network so they can work on one job together.
HPC (high-performance computing)
High-performance computing is solving very large computing problems by running many processors together, traditionally for science, engineering and weather.
Hyperscaler
A hyperscaler is one of the few companies that run computing at global scale and build or lease data centers by the gigawatt, such as the largest cloud providers.
Immersion cooling
Immersion cooling is running servers submerged in tanks of non-conductive fluid that absorbs their heat.
Interconnection queue
The interconnection queue is the list of projects waiting for a grid operator or utility to study and approve their connection to the grid.
Large load
A large load is a single customer whose demand is big enough for the grid operator to study and manage it individually. Data centers and Bitcoin mines are the usual examples.
Liquid cooling
Liquid cooling is removing heat from servers with a liquid instead of air: through cold plates on the chips, a heat exchanger on the rack door, or by submerging the hardware.
Megawatt of data center capacity
Data center size is quoted in megawatts of power, and the figure that matters is critical IT load: the power available to the computing equipment itself.
Neocloud
A neocloud is a cloud provider built around renting GPU capacity for AI, as distinct from the general-purpose clouds.
Powered shell
A powered shell is a data center building delivered with its structure, power connection and network access, but without the cooling and electrical fit-out inside.
PUE (power usage effectiveness)
PUE is a data center's total energy use divided by the energy that reaches its IT equipment. A PUE of 1.0 would mean nothing is spent on cooling or losses.
Rack density
Rack density is the power a single rack draws, in kilowatts. It measures how much computing, and how much heat, is packed into one footprint.
Rear-door heat exchanger
A rear-door heat exchanger is a water-cooled coil that replaces the back door of a server rack and cools the hot air as it leaves.
Training and inference
Training is building an AI model from data. Inference is using the finished model to answer requests. They load a data center differently.
WUE (water usage effectiveness)
WUE is the water a data center consumes on site divided by the energy its IT equipment uses, in liters per kilowatt-hour.

Intelligence Disrupt 2027 tickets and passesIntelligence Disrupt 2027 agendaIntelligence Disrupt 2027 FAQ

The power is on this floor. Be in the room.

March 22–24, 2027 · Irving Convention Center · Dallas

Intelligence Disrupt 2027Dallas · Mar 22–24Passes from $199Get tickets