Bits, Bytes and Binary: How Computers Measure Data (KB, MB, GB, TB)

Every cloud storage contract, broadband quote, laptop specification and backup plan is priced in the same small set of units: bits, bytes, megabytes, gigabytes and terabytes. Most people use these words daily without being sure what they measure, why a new hard drive never shows its full advertised size, or why a “1 gigabit” office connection still takes minutes to upload a large video. The industry itself uses two counting systems with nearly identical abbreviations.

This guide starts with binary, the two-symbol language all digital hardware speaks, then shows how bits become bytes, characters and files. It walks up the unit ladder from kilobyte to petabyte, resolves the 1,000 versus 1,024 problem and the bits-versus-bytes confusion in internet speeds, and ends with what it all means for storage, bandwidth and cloud spending.

What is binary and why do computers use it?

Binary is a number system with only two digits, 0 and 1. We count in decimal (ten digits) because humans have ten fingers; computers count in binary because electronic circuits are most reliable when they only need to tell two states apart. A transistor in a processor is either conducting current or not. A region of a hard disk is magnetized one way or the other. A fiber-optic cable carries a pulse of light or it does not.

Two states make hardware cheap, fast and tolerant of noise. If a circuit had to distinguish ten voltage levels, a small electrical disturbance could turn a 6 into a 7. With only “on” and “off”, the gap is wide enough that minor interference rarely causes an error. Everything a computer does reduces to billions of such on/off decisions. Our guide on how computers work shows how those switches become a working machine.

What is the difference between a bit and a byte?

A bit (“binary digit”) is the smallest unit of data: one 0 or one 1. On its own it answers a single yes/no question. Bits become useful in groups, because each extra bit doubles the number of possible combinations: two bits give four, three bits give eight.

A byte is a group of 8 bits, giving 2⁸ = 256 combinations, enough for every English letter in both cases, the digits, punctuation and some control symbols. The byte therefore became the standard unit for storage and memory. In “16 GB of RAM” or “a 2 TB SSD”, the B means bytes. Our explainer on RAM vs storage covers how those two kinds of capacity differ.

The capitalization rule that matters

• Lowercase b = bits. Used for network and transmission speeds: Mbps, Gbps.
• Uppercase B = bytes. Used for file sizes, storage and memory: MB, GB, TB.

So 8 Mb (megabits) and 1 MB (megabyte) are the same amount of data.

How do you count in binary?

Binary works like decimal with fewer digits. In decimal, each position is worth ten times the one to its right (ones, tens, hundreds). In binary, each position is worth twice the one to its right: 1, 2, 4, 8, 16, 32, 64, 128. To read a binary number, add the values of the positions that hold a 1.

Worked example: reading binary

The binary number 1101 has 1s in the 8, 4 and 1 positions, so it equals 8 + 4 + 1 = 13. A full byte uses eight positions: 01000001 has 1s in the 64 and 1 positions, so it equals 65. The largest single-byte value, 11111111, is 128 + 64 + 32 + 16 + 8 + 4 + 2 + 1 = 255; with zero, that makes 256 values.

Worked example: decimal to binary

To convert 200, subtract the largest position values that fit. 200 − 128 = 72 (128 position is 1); 72 − 64 = 8 (64 position is 1); 32 and 16 do not fit (0, 0); 8 − 8 = 0 (8 position is 1); the 4, 2 and 1 positions are 0. The result is 11001000.

This explains why so many technical limits are powers of two (256, 1,024, 65,536), and why one extra bit in an encryption key doubles the possibilities.

How do computers turn bits into text?

Text needs an agreed table mapping each character to a number, called a character encoding. When you type “A”, the computer stores the number assigned to “A”, and any system using the same table displays “A” again.

ASCII

ASCII (American Standard Code for Information Interchange), standardized in the 1960s, uses 7 bits to define 128 characters: English letters, digits, common punctuation and control codes such as “new line”. “A” is 65 (the 01000001 above) and lowercase “a” is 97. ASCII had no room for accented letters, the Turkish “ş”, Greek, Arabic, Chinese or thousands of other characters.

Unicode and UTF-8

Unicode gives every character in every writing system, including emoji, a unique number called a code point; the standard covers well over 100,000 characters. UTF-8, the dominant encoding on the web, decides how those numbers are stored as bytes, using a variable length:

• 1 byte for ASCII characters, so plain English text is as compact as before.
• 2 bytes for most accented Latin letters and scripts such as Greek, Cyrillic and Arabic (“é”, “ş”).
• 3 bytes for most Chinese, Japanese and Korean characters and symbols such as “€”.
• 4 bytes for emoji and rarer characters.

Encoding mismatches cause garbled customer names after CSV exports; standardizing on UTF-8 prevents most of them.

How big are KB, MB, GB, TB and PB?

Larger quantities of bytes use prefixes from the metric system: kilo, mega, giga, tera, peta. In the decimal (SI) system, each step is 1,000 times larger than the last. In the binary system, used for memory and by some operating systems, each step is 1,024 times larger, because 1,024 (2¹⁰) is the power of two closest to 1,000.

To separate the two systems, the International Electrotechnical Commission introduced binary prefixes in 1998: kibibyte (KiB), mebibyte (MiB), gibibyte (GiB), tebibyte (TiB) and pebibyte (PiB). Technical documentation, Linux tools and many cloud providers use them; consumer software and packaging often do not.

The gap between the systems is small at kilobyte level and grows with each step, reaching about 10 percent at terabyte scale. Above the petabyte sit the exabyte, zettabyte and yottabyte.

Why does a 1 TB drive show only 931 GB?

This is the most common unit complaint in IT support, and nothing is missing. Drive manufacturers use decimal units, correctly by the SI definition: a 1 TB drive holds 1,000,000,000,000 bytes. Windows divides by 1,024 at each step but still labels the result “GB”:

1. 1,000,000,000,000 bytes ÷ 1,024 = 976,562,500 KiB
2. 976,562,500 ÷ 1,024 ≈ 953,674 MiB
3. 953,674 ÷ 1,024 ≈ 931.3 GiB, displayed as “931 GB”

Formatting, recovery partitions and system files reduce free space further. Apple’s macOS switched to decimal units in 2009, so the same drive shows as about 1 TB on a Mac. Memory goes the other way: RAM is built in powers of two, so a “16 GB” module really holds 16 GiB (about 17.2 billion bytes).

Why are internet speeds measured in bits, not bytes?

Networks move data one bit at a time down a wire, fiber or radio channel, so telecom engineers measure transmission in bits per second. Internet plans, Wi-Fi and mobile networks are quoted in Mbps or Gbps, always with decimal prefixes: 1 Mbps is exactly 1,000,000 bits per second. Files are measured in bytes, so divide the bit rate by 8 to compare.

Worked example: how long will this upload take?

Your office has 100 Mbps of upload and you need to send a 5 GB video to a client.

1. Speed in bytes: 100 Mbps ÷ 8 = 12.5 MB per second.
2. File size: 5 GB = 5,000 MB.
3. Time: 5,000 ÷ 12.5 = 400 seconds, about 6 minutes 40 seconds.

Expect somewhat longer in practice. Protocol overhead, shared lines and slower servers all cost speed. On a good wired link, 80 to 90 percent of the advertised rate is realistic; busy Wi-Fi delivers less. That is why a “1 Gbps” connection typically shows downloads of roughly 110 to 118 MB/s rather than 125.

Also check whether a plan is symmetrical. Many business broadband packages have fast downloads but much slower uploads, and uploads drive cloud backup, video calls and file sharing. Our guides on how the internet works and Wi-Fi vs 5G vs Ethernet explain where bottlenecks usually sit.

How big are typical files?

Sizes vary with quality settings and formats; treat these 2026 ranges as rough planning guides.

• Email: a few KB for plain text; formatted HTML emails with logos often 50 to 200 KB. Attachments dominate everything else.
• Office documents: a text-only Word file is often under 100 KB; an image-heavy slide deck commonly runs 5 to 50 MB.
• Smartphone photo: roughly 2 to 5 MB as JPEG or HEIC; professional RAW photos 25 to 60 MB each.
• Music: about 1 MB per minute as standard-quality MP3.
• Video: one hour of streamed 1080p HD is on the order of 3 GB; 4K roughly twice that or more. Raw professional camera footage can reach tens or hundreds of GB per hour.

Text barely registers; images and video dominate storage growth.

How does compression make files smaller?

Compression represents the same information with fewer bits. It is why a phone photo takes a few MB rather than tens of MB of raw pixels. There are two families.

Lossless compression

Lossless methods find redundancy and describe it more efficiently, so the original can be rebuilt exactly, bit for bit. ZIP, PNG, FLAC and the compression inside many databases and backup tools are lossless. Text, spreadsheets and logs often shrink to a fraction of their size. Files that are already compressed, such as JPEGs or MP4 videos, barely shrink further, worth remembering when a backup vendor quotes an optimistic ratio.

Lossy compression

Lossy methods discard detail people are unlikely to notice, such as subtle color variations or sounds masked by louder ones. JPEG, HEIC, MP3, AAC and video codecs such as H.264, H.265 and AV1 are lossy. Savings are dramatic, but discarded detail is gone for good and repeated re-saving degrades quality. For contracts, engineering drawings or original marketing footage, keep a lossless master and share lossy copies.

What does this mean for your business?

Data units stop being trivia the moment they appear on an invoice. Three areas matter most.

Storage planning

Estimate how much data you hold, how fast it grows and how many copies you keep. The common 3-2-1 backup rule (three copies, two media types, one offsite) triples raw capacity before version history. When comparing quotes, confirm whether figures are decimal or binary, and raw or usable after redundancy, since RAID or erasure coding can consume 20 to 50 percent of raw space. Our comparisons of cloud storage for business and backup and recovery software show how vendors price capacity.

Bandwidth planning

Plan for peaks: simultaneous video calls, a nightly cloud backup and a remote office syncing design files can collide. Convert to one unit first. A 200 GB nightly backup in an 8-hour window needs 200,000 MB ÷ 28,800 seconds ≈ 6.9 MB/s, or about 56 Mbps of sustained upload before overhead. For customer-facing downloads, a content delivery network moves heavy files closer to users; see our CDN comparison.

Cloud billing per GB

Public clouds charge per GB (sometimes per GiB) stored per month, plus fees for data transferred out, called egress. As of 2026, standard object storage is commonly priced in the low US cents per GB per month, archive tiers less, and internet egress typically several cents per GB. That looks tiny until you remember 100 TB is 100,000 GB. Check retrieval fees on archive tiers and model the cost of moving data out before committing. Our overview of cloud computing covers the wider pricing model.

Questions to ask vendors and IT

• Are capacities in decimal (GB/TB) or binary (GiB/TiB) units, raw or usable?
• Is the internet speed guaranteed or “up to”, and what is the upload speed?
• What compression and deduplication ratio does the storage quote assume, and is it realistic for our file types?
• What are the egress, API request and retrieval charges beyond storage?
• How fast is our data growing, and when will we hit current limits?

Frequently Asked Questions