In the 1950s, two IBM engineers built a new way to store computer data on spinning disks; it held 5 million characters and recalled a record in under 1 second

Somewhere between "please don't lose my file" and "why is my phone begging me to buy more cloud storage" was the rather unglamorous fact that a computer "remembering" something instantaneously, without a person sifting through a filing cabinet, was barely a human lifespan old when it first appeared. It happened with a creation that was the size of two refrigerators, a ton and a half heavy, and held as much information as a single low-resolution image.

According to the National Inventors Hall of Fame, engineers William Goddard and John Lynott, with help from Louis Stevens and a wider IBM team up in San Jose during the 1950s, developed a magnetic disk storage device that could store five million characters and recall any individual record in under a second. In 1956, that was pretty astonishing.

The problem (literally) was a filing cabinet


Before then, businesses used punch cards, and if a warehouse needed to update a single record, they would have to have someone go dig through the filing cabinet, find the individual card with the information they wanted, update it, and write it back in place. The Computer History's Museum account of the project notes that IBM put together a team of people, led by researcher Reynold Johnson, to try and come up with something faster than punch cards or magnetic tape for the storage and management of accounting and inventory data, a genuinely mundane logistics headache that ended up birthing an entire industry.

The answer was a spinning stack and a clever form of "floating"

The creation that followed, the IBM 350 disk storage unit, a component of the IBM 305 RAMAC system, was a fifty-disk stack spinning on a central axis, every twenty-four inches across. According to the original 1957 technical paper titled “The Random-Access Memory Accounting Machine—II. The Magnetic-Disk, Random-Access Memory,” by T. Noyes and W. E. Dickinson, published in the IBM Journal of Research and Development, the system's overall capacity was five million alphanumeric characters, with read/write heads held up to the spinning disks by jets of air.

The air-cushioned heads were specifically the innovation credited to Goddard and Lynott, and it's arguably an even more fascinating engineering story than the five-million-character headline alone. As the National Inventors Hall of Fame describes it, the read/write heads "floated" on a cushion of air just above the surface of the spinning disks, not the disks floating above one another, which the original draft's wording confusingly implied. That air-bearing technique is still used, in far more refined form, in hard drives today.

A ton of hardware for less storage than a text message thread

Here's where the story gets humbling for the person who carries a phone with 256 gigabytes of storage in their pocket. Five million characters is equal to about 3.75 megabytes, or less than the space needed to store a single good-quality photo. However, according to historyofinformation.com, the unit, which weighed over a ton, needed to be shipped around with forklifts and cargo planes. Companies could lease the unit for $3,200 a month, a steep price at the time, and that was the cost of storing five million characters.

IBM's engineers reportedly could have expanded the system's storage capacity well beyond five million characters relatively easily. But EDN has reported speculation that IBM's marketing department wasn't sure how to position that much added capacity as something customers would pay more for, even though the inventory and accounting backlogs the RAMAC was built to solve could plainly have used it. It's a frustratingly familiar complaint for anyone who's seen a modern technology company hold back capability a product could already deliver, until a "premium" tier finally packages it for sale.

Why it still matters

The IBM 350 wasn't just an early hard drive. It was the first proof that a computer could reach inside and pull out a single record out of a mass of stored records, on demand, rather than slogging through the entire thing. Every time an application today instantly recalls a single file rather than scanning an entire library, it owes something to the work two engineers and a broader IBM team, using the vacuum-tube electronics of the era, proved out in San Jose.