- The evolution of storage has progressed from punched cards and magnetic tapes to hard drives, optical drives, and increasingly compact and cheaper flash memory.
- The combination of high-speed networks and large data centers has boosted the cloud, data lakes, and distributed massive storage for Big Data.
- Current challenges focus on durability, security, and managing exponential data growth, exploring technologies such as holographic memory and DNA storage.
Throughout history, humanity has devised all sorts of tricks and devices to store information : from clay tablets and punched cardboard cards to data clouds spread across the globe. The volume of data we generate today is so staggering that our brains, however evolved, fall short and need technological assistance.
Understanding how we have gone from a few kilobytes in a metal cabinet to terabytes in our pockets and zettabytes in distributed data centers is not just a historical curiosity: it explains why we work the way we do, how we consume digital entertainment, and what risks we face of losing memories, documents, or knowledge if we choose the wrong storage medium.
From human memory to the first data storage machines
Before discussing disks, tapes, or clouds, it's worth remembering that the first storage device is the human brain, with its limited and imperfect biological memory . Evolution increased its size and capacity: species like Australopithecus afarensis had brains of around 400-500 cm³, Homo habilis reached 600-700 cm³, Homo erectus reached 800-1100 cm³, and Neanderthals were around 1200-1600 cm³ . Modern humans, Homo sapiens, have brains of around 1230 cm³.
This leap allowed us to manufacture tools, create cultures, rituals, and complex languages , but memory remains fragile: we forget, distort memories, and are unable to handle the current avalanche of data. Hence the need to develop external devices that store information stably , without depending on our neurons.
Punched cards: the language of holes
The first major leap toward digital storage was mechanical: punched cards . In 1725, the Frenchman Basile Bouchon devised a system of perforated cards to control looms. It wasn't computing in the modern sense, but it was a method of encoding and reusing instructions.
In 1837, the British mathematician Charles Babbage designed his famous Analytical Engine , a calculating machine with moving parts that used punched cards for both commands and results. The holes acted as on/off switches , anticipating binary code.
Later, the American Herman Hollerith put this idea into practice with his Analytical Engine for the Census: the punched holes represented instructions and also data that the machine could read automatically. For decades, punched cards were the standard storage medium for programs and data in many computers.
From the 60s onwards they began to be superseded by magnetic technologies, although they continued to be used well into the 80s for tasks such as standardized tests or electronic voting on punched paper.
The beginnings of magnetism: tapes, drums, and cores
In the 50s, the world of computing embraced magnetism for storing information. The basic idea is simple: coat a medium with magnetizable material and organize it into tiny domains (dipoles) that represent bits.
In 1928, German engineer Fritz Pfleumer patented magnetic tape , inspired by Valdemar Poulsen's magnetic cable. In 1965, Mohawk Data Sciences introduced a magnetic tape encoder designed to gradually replace punched cards . Tapes offered greater capacity at a lower price, although access remained sequential.
In 1932, the Austrian Gustav Taushek invented the magnetic drum , a cylinder coated with magnetizable material. These drums could store around ten thousand words and were the direct predecessors of modern hard drives, offering faster access than tape.
Around the same time, technologies for main memory emerged. Between the late 40s and 50s, magnetic core memory , also called toroid memory, was developed: small rings of ferromagnetic material traversed by conductive wires. Each core stored one bit , and its magnetic state could be read and written quickly.
In 1953, MIT acquired the patent and built the Whirlwind computer , a pioneer in using magnetic core memory as RAM. It was much faster and more reliable than punched cards, but its manufacture was complex and expensive , requiring very precise and laborious assembly.
Electronic memories: from the Williams tube to the semiconductor
Before transistors revolutionized everything, there were intermediate electronic storage solutions. In 1946, Professor Frederick C. Williams and Tom Kilburn developed the so-called Williams tube at the University of Manchester. It was a modified cathode ray tube that allowed binary data to be stored as charge patterns on the screen . It was used as RAM in early stored-program computers.
In 1948, another unusual technology emerged: the Selectron , a thermionic valve capable of functioning as random access memory . It was developed by Jan A. Rajchman and his team at the Radio Corporation of America (RCA). Although technically interesting, its complexity and cost significantly limited its adoption.
In 1949, delay-line memory was introduced , based on taking advantage of the time it takes a signal to propagate through a physical medium (for example, mercury or quartz). Bits were encoded as pulses that circulated continuously through the medium; it was an ingenious solution, but not very flexible compared to later RAM.
In the late 50s, magnetic core memory became the dominant RAM, until semiconductors began to emerge in the 60s and 70s . In 1966, the newly formed Intel began selling 2000-bit semiconductor memory chips , where each cell contained miniaturized transistors or capacitors.
In 1966, DRAM (Dynamic Random Access Memory) also appeared . In DRAM, each bit is stored as an electrical charge in a capacitor . It is volatile memory: if the power is cut off, the information is lost. However, it is very dense and inexpensive, so it became the standard main memory in personal computers and servers.
In contrast, ROM (Read-Only Memory) is programmed permanently and retains data even when the device is turned off, very useful for storing firmware and boot routines.
Advanced magnetic innovations: Twistor and Bubble Memory
In 1957, researcher Andrew Bobeck invented Twistor memory at Bell Labs . This technology wrapped magnetic tape around a conductive wire , instead of using toroidal cores, which reduced weight, power consumption, and production cost.
Bell Labs presented Twistor as a superior alternative to magnetic cores: cheaper, lighter, and easier to manufacture . However, its commercial window was brief, because more compact and scalable semiconductor RAM chips soon arrived.
Based on this research, Bobeck developed the famous Bubble Memory in 1980. This technology used a very thin film of magnetic material in which small magnetized areas, the "bubbles," were formed , each representing a bit. It was non-volatile and quite robust , although difficult to manufacture on a large scale.
Magnetic disks and tapes: from the IBM 350 to the floppy disk
The big leap to what we are familiar with today came with magnetic disks . In 1951, the UNIVAC I already incorporated tape drives capable of storing up to 128 words per inch. But the king of the modern party was the hard disk drive.
In 1956, IBM launched the IBM 350 , considered the first modern hard drive. It had a capacity of about 4,4 MB , consisted of fifty 24-inch platters spinning at 1200 rpm, and was about the size of a refrigerator, weighing over a ton. Even so, it offered very fast random access compared to tapes.
Over time, improvements in magnetic recording technology led to hard drives gaining capacity, decreasing in price, and becoming miniaturized , eventually leading to today's desktop and laptop HDDs, with several terabytes in 3,5 or 2,5-inch formats, a decision that is now compared to SSD vs external HDD.
Alongside the large HDDs, IBM also promoted floppy disks . In 1971-1972, 8-inch floppy disks appeared , with an initial capacity of about 80 KB. They consisted of a disk of flexible material coated with magnetic oxide, inside a protective sleeve.
In 1975, Allan Shugart developed the 5,25-inch floppy disk , which was more manageable for personal computers. These disks had a capacity of around 110 KB in their early versions and were faster and cheaper than 8-inch disks. By 1978, there were already about ten manufacturers producing 5,25-inch drives.
In the early 80s, the classic 3,5-inch floppy disk burst onto the scene , with its rigid casing and a metal tab protecting the magnetic surface. It was more durable, more compact, and had a greater capacity than its predecessors. It became so popular that even today, in many programs, the "Save" icon is still a floppy disk.
Despite being practically obsolete in the domestic sphere, some critical infrastructures, such as certain military or nuclear systems, have maintained floppy disk drives for surprisingly many years, precisely because of their simplicity and isolation, a practice related to how to store your old PCs.
Magnetic tape in computing: from backup to longevity
Although the sequential access of magnetic tape makes it slower than disks, its low cost/capacity ratio and durability have kept it as a star medium for massive backups.
In the 80s, formats such as the audio tape cassette adapted for data appeared, widely used in home microcomputers, and later specific backup tapes such as DAT (Digital Audio Tape) , developed by Sony in 1987. It was a redesigned cassette, with 4 mm tape in a compact cartridge.
In 1989, Sony and Hewlett-Packard launched the DDS (Digital Data Storage) standard , an evolution of DAT focused on computer data storage. It offered increasing capacities in relatively small formats , ideal for businesses.
Optical discs: CD, DVD and Blu-ray
While magnetic recording dominated, another revolutionary approach was taking shape: optical storage using lasers . In the 60s, inventor James T. Russell worked on the idea of using light to record and play back music. For years it was considered little more than a novelty.
In 1975, Sony and Philips invested heavily in his project and financed its development. The result was the compact disc (CD) , commercially launched in 1980. For audio, it could store around 700 MB of data in computer format, an enormous amount at the time compared to floppy disks.
In 1984, the CD-ROM (Compact Disc Read-Only Memory) appeared , which used the same physical format as the audio CD but encoded computer data. Micro-pits and flat areas (lands) are etched onto the plastic surface of the disc , which the laser interprets as bits.
Later, recordable variants emerged: the write-only CD-R and the rewritable CD-RW , introduced in 1995. This allowed any user to record and erase data multiple times on the same disc.
In 1995, the DVD (Digital Versatile Disc) arrived , substantially increasing data storage capacity. A single-layer DVD held 4,7 GB , and dual-layer DVDs held twice that. It became the standard for home video and large software distribution.
As an attempted rival, the HD-DVD format appeared in 2005 , promoted by Toshiba, NEC, and Sanyo. It offered high definition, but the commercial battle was ultimately won by Blu-ray , launched in 2003 and based on a shorter wavelength blue-violet laser , which allows more information to be stored in the same physical space.
Blu-ray established itself as a medium for high-definition video and high-density storage , with capacities of 25 GB per layer and multi-layer variants reaching tens of gigabytes per disc.
Magneto-optical media and hybrid formats
Between pure magnetism and optical discs, a hybrid family emerged: magneto-optical discs . Introduced around 1990, they combined magnetic and optical techniques to store and read data.
These discs, often 3,5 or 5,25 inches in cartridges, used a laser to locally heat the surface and a magnetic field to orient the platters. Reading was based on variations in the polarization of the reflected light (Kerr effect) . They offered good durability and rewritable capabilities , although they were more expensive and complex than CDs or DVDs.
The 90s also saw some short-lived but interesting formats. In 1992, Sony released the MiniDisc , designed to replace the audio cassette and also store data (in MD Data versions), with a capacity of around 140 MB. In 1994, Iomega launched Zip and later Jaz , high-capacity removable disks for the time (100 MB to 1 GB), which attempted to fill the gap between the floppy disk and the first external hard drives.
Flash memory and USB drives: the great portable leap
In the late 90s, a silent revolution arrived: flash memory , a type of non-volatile electronic storage. Although it was initially designed for cameras and portable devices , it soon spread to everything.
In 1993, CompactFlash (CF) was introduced , a card that integrated flash memory into a robust format. It was widely used as internal storage in digital cameras and some embedded computers . Shortly afterward came SmartMedia (Toshiba, 1995) and the Multimedia Card (MMC) from Siemens and SanDisk in 1997.
In 1999, IBM released the Microdrive , a mini hard drive the size of a compact card, and in 2000 , Secure Digital (SD) cards became popular , with built-in encryption and standard dimensions of 32 x 32 x 2,1 mm. SD cards became the de facto standard for mobile phones, cameras, and a multitude of portable devices.
The major turning point for the average user was the USB flash drive . Around the year 2000, the Singaporean company Trek 2000 International introduced the ThumbDrive , considered the first commercially successful USB flash drive. It used NAND flash memory and connected directly to a USB port , without the need for external power or complicated controllers, although it's sometimes useful to be aware of potential issues with USB hubs . These drives, also called pen drives, USB sticks, or flash drives , offered capacities starting in just a few megabytes and soon grew to tens of gigabytes. They had no moving parts, could withstand thousands of write cycles, and were highly resistant to shocks and electromagnetic interference . With this, they replaced floppy disks and began to overshadow CDs and DVDs as a storage medium.
Solid-state drives (SSDs) and SMRs: Accelerate and squeeze capacity
The same flash memory technology gave rise to solid-state drives (SSDs) . These devices function like hard drives from the operating system's perspective, but internally they are interconnected sets of flash chips , without platters or read/write heads; if you're interested in the difference with other technologies like NVMe storage , it's quite revealing.
The first commercially successful SSDs came from companies like SanDisk and quickly spread to laptops and desktops, displacing HDDs where access speed and reduced power consumption are key. Their chips differ from those typically found in USB flash drives, offering higher performance, durability, and wear resistance , which translates into a higher price per gigabyte.
Meanwhile, traditional hard drives have continued to evolve. One recent technology is Shingled Magnetic Recording (SMR) . Instead of writing separate tracks, they are partially overlapped like roof tiles , cutting tracks without losing relevant content.
This allows for increased capacity in the same physical space , while maintaining low costs and leveraging the existing HDD infrastructure. SMR is already present in many modern hard drives, especially in high-capacity ranges designed for mass storage and archiving.
From physical warehouse to data center: silos, lakes and Big Data
The explosive growth of digital information led organizations to create large logical storage systems. This gave rise to data silos , which are sets of information stored for a specific department or system, incompatible or difficult to integrate with the rest of the company.
Initially, these silos were seen as strange islands of data; over time, they became a valuable source of information for Big Data , enabling the combination and analysis of large volumes of historical data. Later, the concept of a Data Lake emerged.
A data lake stores information in its original, unprocessed format , typically managed using NoSQL databases. These systems accept structured, semi-structured, and unstructured data, allowing analysts and AI algorithms to process the information on demand.
According to initiatives like BBVA OpenMind, data lakes employ flat architectures , without rigid hierarchical structures, to facilitate flexible access and keep costs down. They have become essential components for Big Data and artificial intelligence projects.
The cloud and online storage: data everywhere
The next major shift wasn't so much a new physical medium as a change in model thanks to high-speed networks . Improved bandwidth and cheaper disk space made it possible to build enormous data centers accessible via the internet.
Thus, cloud computing was born , and with it, cloud storage . From the user's point of view, the cloud offers virtually unlimited capacity, accessible from any device and location , in exchange for a fee or even for free if limitations are accepted.
In practice, the "cloud" is a massive collection of servers, disk arrays, SSDs, tapes, and internal networks . It's used for both backups and primary storage of documents, photos, and videos . If you use email, social media, or streaming services, you're already using cloud storage daily.
Providers have had to strengthen security through encryption, authentication, and access controls because a shared architecture poses confidentiality challenges. Sectors like banking are subject to strict regulations to mitigate these risks.
In addition to general-purpose online storage, there are specific wireless home storage solutions . A historical example is the Apple AirPort Time Capsule , a Wi-Fi device that integrated a router and a hard drive of up to 3 TB, designed for automatic backups and wireless data access from Apple devices.
New frontiers: holograms, DNA and ultra-long-lasting materials
Faced with the tsunami of data expected in the coming decades, storage research is exploring paths very different from disks or flash drives. One of these is holographic memory , which stores digital data within the volume of a material such as crystals or photopolymers , rather than being limited to its surface.
The great advantage of holographic memory is that it can utilize the thickness of the storage medium to store information in 3D (Bragg volume), achieving enormous densities. Although prototypes exist, it is not yet a widely used technology, but it is emerging as one of the possible long-term solutions for very high-density archiving.
Another fascinating field is DNA storage . DNA, the molecule of life, can encode a staggering amount of information: an estimated 2,2 petabytes per gram . In theory, all the data generated by humanity could fit into a spoonful of DNA.
Furthermore, DNA is an extremely long-lasting medium , capable of preserving information for thousands of years if stored correctly. The current problem is cost and speed: encoding less than 100 KB of data can cost around $1500, and synthesis and sequencing processes remain slow.
The use of artificial DNA or modified biomolecules is being investigated to reduce the cost and speed of these operations, but it is still in the experimental stage. The idea, however, is clear: to move from disks and chips to truly molecular storage.
In parallel, nanotechnologies are opening doors to applications such as the use of carbon isotopes (for example, carbon-12 atoms for “0” and carbon-13 atoms for “1”) or particularly stable mineral materials. In 2023, the company Cerabyte announced a system that uses a laser to engrave three-dimensional arrays of data similar to QR codes onto a mineral substrate , resistant to extreme temperatures, fire, floods, and power surges, with an estimated lifespan of over 5000 years.
Reliability, bit rot, and copies: the less glamorous side of storage
Regardless of the storage medium, they all share one problem: they have a limited lifespan and are susceptible to failure . Hard drives can be damaged by impacts, tapes can become demagnetized, optical discs can degrade, and flash memory can lose its charge over time.
A frequently mentioned phenomenon is bit rot , the silent degradation in which some bits change value without being immediately noticeable . This especially affects long-term storage media such as old hard drives, tapes, or flash drives that go many years without being read.
Therefore, it is recommended to periodically access storage media (for example, reading a USB drive every so often) and replicate the information to new media every few years. Systems such as RAID with error-correcting codes allow for the reconstruction of damaged data by introducing redundancy, but they never completely eliminate the risk.
Regarding classification, we typically distinguish between primary storage (RAM, cache), which is ultra-fast but volatile and directly accessible by the processor, and secondary storage (disks, SSDs, tapes, cloud storage), which is slower but non-volatile. Within secondary storage, we further differentiate between sequential access (tapes) and random access (disks, SSDs, flash storage), depending on whether we can jump directly to the data we want or have to scan the entire data from the beginning.
Regardless of the technology, the only real defense against data loss remains redundancy and well-planned backups . And, when dealing with sensitive information, cryptography comes into play , both public-key (RSA, etc.) and symmetric-key (AES, DES), which is especially relevant in cloud storage contexts.
Overall, the history of storage has taken us from enormous machines with a few kilobytes of capacity to minuscule chips capable of handling terabytes and distributed systems managing zettabytes; and yet, we continue to search for new, denser, more durable, and more secure storage media , from holograms to DNA or exotic mineral materials. Everything suggests that, just as has been the case so far, the next leaps will combine physical innovations with increasingly intelligent network architectures to continue enabling us to store and retrieve, almost without realizing it, the vast amount of data we produce every day.
