Terabyte

A terabyte is a unit of measurement for amounts of data, just as the kilogram is a unit of measurement for weight. The smallest unit is the byte. One byte is roughly enough for a single letter in a text. A thousand bytes make a kilobyte, a thousand kilobytes make a megabyte, a thousand megabytes make a gigabyte. And a thousand gigabytes make a terabyte, in other words roughly one trillion letters. The abbreviation is TB.

This becomes concrete with examples. A photo from a phone takes up about 3 megabytes. A cinema-quality movie is 5 to 15 gigabytes. So a terabyte can hold roughly 300,000 photos or about a hundred movies. A normal hard drive in a laptop today usually has 0.5 to 2 terabytes.

The unit in which AI training data is measured

Twenty years ago, a terabyte was an amount that only interested data centers. Today it has become the standard size for everyday devices. This exact shift explains why the unit constantly shows up in tech news. What was once huge has become the normal reference point.

This is especially important in artificial intelligence. A large language model is trained on text from the internet, from books, and from program code. These amounts of text run into several terabytes. For comparison: the entire German-language text of Wikipedia fits into a few gigabytes. A training dataset is therefore often a thousand times larger than a complete encyclopedia.

The unit also carries economic weight. Cloud providers bill storage space and transferred data volumes per gigabyte or terabyte. When a company permanently stores several hundred terabytes, this results in ongoing costs of tens of thousands of dollars per year. That’s why the figure appears in business reports and not just in technical spec sheets.

Why 1 TB on a hard drive is only 931 GB

Computers don’t calculate in base ten but with powers of two. That’s why the unit historically exists in two variants. Hard drive manufacturers mean exactly one trillion bytes by a terabyte, i.e. 1000 times 1000 times 1000 times 1000. Operating systems like Windows, on the other hand, often calculate with 1024 instead of 1000 per step.

This creates a well-known source of confusion. You buy a hard drive with 1 terabyte, and the computer only shows about 931 gigabytes. Nothing is missing; these are simply two different ways of calculating. The correct term for the power-of-two variant is tebibyte, abbreviated TiB. In everyday use, however, hardly anyone uses this term.

Another pitfall is confusing byte and bit. A byte consists of eight bits. Storage is given in bytes, while internet speed is given in bits per second. A connection with 100 megabits per second therefore transfers only about 12.5 megabytes per second. To transfer a terabyte over it, you would need about an entire day.

Terabytes in price lists, phone plans, and data centers

You most often encounter the unit when buying devices. Smartphones are sold with 128 or 512 gigabytes, top models with one terabyte. External hard drives come in 2, 4, or 8 terabytes. Cloud storage such as Google Drive or iCloud is also sold in this unit, usually as a monthly subscription.

The term appears in the news in connection with data leaks. When attackers hack a company, it is often reported how many terabytes they copied. This figure is a rough indicator of the extent of the damage, but it says nothing about the sensitivity of the content. A terabyte of video material is more harmless than a gigabyte of customer data containing passwords.

For very large systems, the unit is no longer sufficient. Above it lie the petabyte, a thousand terabytes, and the exabyte, a million terabytes. Data centers that train AI models have long been operating at these scales. Nevertheless, the terabyte remains the unit you encounter most often in everyday life.