Inside a 120 kW AI rack: power and cooling
High-density AI hardware is not simply more computing in less space. It changes the cables, switchgear, cooling loops, backup systems, and utility connection around the cabinet.
Read the full explainerAI's power, chips and data centers, explained in plain English
THE EXPLAINER LIBRARY
Explore the systems, engineering and consequences behind AI infrastructure.
High-density AI hardware is not simply more computing in less space. It changes the cables, switchgear, cooling loops, backup systems, and utility connection around the cabinet.
Read the full explainerCoolant can circulate in a sealed loop without being consumed. Water use depends on how the facility ultimately rejects heat—and whether evaporation is part of that design.
Read the full explainerA new power plant does not help a data center unless the transmission network, substations, transformers, and local circuits can deliver that electricity to the exact site.
Read the full explainerThe overlooked connections that turn a room full of processors into a coordinated machine.
Read the full explainer →Keeping an AI service available means protecting more than the electricity entering its servers.
Read the full explainer →The saved progress that can prevent an interrupted training job from losing all its work—and the infrastructure needed to preserve it.
Read the full explainer →IT load, total site power and actual consumption measure different things. Learn how to read the numbers behind an AI campus announcement.
Read the full explainer →How grid studies, utility planning and generation queues shape the wait for a data-center connection.
Read the full explainer →How utility tariffs, transmission planning and customer contracts divide data-center grid costs.
Read the full explainer →Grid supply, gas, nuclear, renewables, storage and power purchase agreements explained in plain English.
Read the full explainer →Why custom transformers, specialized materials and difficult delivery routes can become the hidden clock behind an AI campus.
Read the full explainer →How onsite solid-oxide fuel cells make electricity, why natural gas still brings emissions, and which grid constraints remain.
Read the full explainer →Understand learning, serving answers and the factors that change AI electricity use—without treating every query as the same task.
Read the full explainer →Understand on-site generation, grid connections and microgrids—and the questions a private power plant still has to answer.
Read the full explainer →Follow electricity supply, utility investment and rate rules to understand when data-center growth may raise—or help spread—household costs.
Read the full explainer →Trace heat from an AI chip through cold plates, coolant loops and heat exchangers to understand what liquid cooling actually changes.
Read the full explainer →Convert MW and GW into an explicitly dated household-energy comparison while avoiding the common confusion between power, energy and capacity.
Read the full explainer →Separate existing reactors, restarts and proposed small modular reactors—and check what a nuclear deal actually promises a data center.
Read the full explainer →Learn to read power usage effectiveness, check its measurement boundary and separate building efficiency from useful computing and total energy.
Read the full explainer →Follow an AI calculation through a GPU, its memory and its network—and learn why buying more chips is only part of the job.
Read the full explainer →Power, fiber, customers and local rules help explain data-center clusters—and why a successful hub can become harder to expand.
Read the full explainer →Taiwan holds 92% of the world's manufacturing capacity for the most advanced logic chips (below 10 nm, 2019 data) — most of it one company, TSMC. Here's how that happened, and what is being built to change it.
Read the full explainer →Space offers endless solar power and no land or water use — but no air to carry heat away. What has actually flown, what is only announced, and why cooling sets the limit.
Read the full explainer →Microsoft sank a data center off Scotland and its servers failed one-eighth as often as on land. China now runs them commercially. How underwater data centers work.
Read the full explainer →On 28 April 2025, Spain and Portugal lost power in about ninety seconds. It was not renewables, not inertia and not a cyberattack — it was a voltage collapse. Here is what happened, and what stopped it happening again.
Read the full explainer →From ENIAC's 30 tons of vacuum tubes to buildings full of GPUs: how AI went from a 1956 workshop idea to an industry.
Read the full explainer →Cooling fans, chillers, transformers and backup generators make data centers hum around the clock. Here is how loud they are, why the legal limit does not always match what neighbours hear, and what actually quiets one down.
Read the full explainer →Air is an insulator — until the voltage gets high enough. Why sparks jump gaps, why the grid runs at high voltage, and what arc flash is.
Read the full explainer →How stacked memory feeds AI processors, why capacity and bandwidth are different, and when faster arithmetic still has to wait for data.
Read the full explainer →Glossary — plain definitions for words used here.