CPU, RAM, Storage, Motherboard, and why AI needs ALL the GPUs.

How your message travels through switches, routers, and undersea cables to reach ChatGPT.

- **[Crash Course - Binary](/dt7/Crash%20Course%20-%20Binary)** *(way2.cc/ca8)* — Carrie Anne Philbin walks you from "binary is just two states" to ASCII, RGB, and why a kilobyte is 1024 bytes.

- **[Vox - Undersea Cables](/dt7/Vox%20-%20Undersea%20Cables)** *(way2.cc/ca9)* — 99% of the world's internet runs through cables thinner than a garden hose, sitting on the ocean floor.

- **[Raspberry Pi Foundation - What Is A Raspberry Pi](/dt7/Raspberry%20Pi%20Foundation%20-%20What%20Is%20A%20Raspberry%20Pi)** *(way2.cc/caa)* — The official 4-minute tour: Pi family, the $35 charity backstory, real-world projects from weather stations to the ISS.

Type a number and see it in binary — or type binary and see the decimal:

- [RapidTables Binary Converter](https://www.rapidtables.com/convert/number/decimal-to-binary.html) — quick conversion both ways

- [CS Unplugged Binary Cards](https://www.csfieldguide.org.nz/en/interactives/binary-cards/) — the dot card activity from class, but online

- [Firia Labs Network Simulator](https://netsim.erinn.io/) — build a virtual network with switches and routers

- [Subnet Game](https://www.subnetting.net/) — test your networking knowledge (extension challenge)

Modern CPUs use 3-nanometre transistors. A human hair is about 80,000 nanometres wide. You could fit roughly 26,000 transistors across a single hair. The Apple M4 chip has 28 billion of them.

Because transistors are switches — they're either ON (1) or OFF (0). It's easier to build reliable hardware with two states than ten. Everything else (text, images, video, AI) is built on top of these two states.

About 200,000 km/s — roughly two-thirds the speed of light. A message from Melbourne to London (17,000 km) takes about 85 milliseconds through undersea cables. That's faster than you can blink.

Thousands of server racks, each holding dozens of computers. Google has over 30 data centres worldwide. A single data centre can use as much electricity as a small city. Most of that power goes to cooling — the chips generate enormous heat.

AI training needs thousands of GPUs running together for weeks or months. Companies like OpenAI, Google, and Meta are buying every GPU they can get. NVIDIA's revenue tripled because of AI demand. The same chip architecture (CUDA cores) that makes games look good also makes AI calculations fast.

|--------|---------|

| `00001101` | 13 |

| `01010101` | 85 |

| `11000011` | 195 |

| `10000001` | 129 |

|---------|--------|

| 25 | `00011001` |

| 100 | `01100100` |

| 42 | `00101010` |

| 200 | `11001000` |

**Check your working:** For each 1 in the binary number, add up the place value. For decimal to binary, find the biggest power of 2 that fits, subtract, repeat.

---

Wondering what your child is learning? Here's a quick summary:

- Students are exploring **how computers work** — opening old laptops, identifying components, understanding what each part does

- They're learning **binary** — the number system computers use (just 0s and 1s)

- They're building **networks** — connecting devices with real cables and switches

- Everything connects to **AI** — helping students understand that AI runs on the same hardware they're studying

This aligns with the [Victorian Curriculum: Digital Technologies](https://victoriancurriculum.vcaa.vic.edu.au/technologies/digital-technologies/introduction/rationale-and-aims) for Years 7-8.

---