Part 2 opens after the part 1 check

Finish the check at the end of part 1 to open this lesson.

Back to the course

Lesson 6 of 9 · Part 2 · Check yourself · 2 min Draft

Check yourself

Four quick questions on chips, memory, power and datacenters.

Four quick questions on Part 2. Look back at the lessons whenever you like.

Check yourself

Get every answer right to open part 3.

1. About how big is the die at the heart of an H100 chip? About the size of a dinner plate About the size of a postage stamp About the size of a grain of rice

The H100 die is about 28.5 mm on a side, close to the largest rectangle a chipmaking machine can print in one go.

2. Why does an AI chip sit next to stacks of fast HBM memory? To store your chat history safely To keep the chip cool Each new token means reading all the weights To hold spare transistors

Writing each token means reading every weight once, so memory speed sets how fast the reply appears.

3. What does Microsoft's "tokens per megawatt" figure for Maia 300 measure? How much text a system writes for its electricity How many chips fit in a rack How fast one chip can read memory

Tokens per megawatt is text produced for a given amount of power, so it links chip design to running costs.

4. By 2030, about what share of world electricity does the study expect AI datacenters at six large firms to use? About 10% About 25% About 0.01% About 1%

The study projects 239 to 295 terawatt-hours a year for six large firms by 2030, about 1% of global demand.

← Where the power goes Smaller models, same answer →

Books

The Model Race
All books
Errata & corrections

Learn

Facts & News
Courses

Build

Agents
Robotics

Lab

About

Company

Contact