Fujifilm will invest roughly 10 billion yen, about $63 million, to increase its production capacity for magnetic tape fivefold, in an announcement released on Thursday that credits the growth in archiving demand to generative AI.
The company said the spread of generative AI is producing “exponential growth” in data needs worldwide. The fivefold increase will be measured against its production levels in the 2025 financial year.
The money goes to the Odawara site of Fujifilm’s Kanagawa Factory, in what the company calls “the largest investment Fujifilm has made to date in dedicated LTO tape facilities”. The project runs from the 2026 through the 2028 financial years and funds manufacturing expansion and additional personnel, primarily to produce the latest generation of tape, FUJIFILM LTO Ultrium10.
Mainstream consumers abandoned tape decades ago. The enterprise world never did, and cloud giants rely on it to back up vast quantities of data. Fujifilm’s work on magnetic tape as a storage medium dates back to the 1950s; it released its first LTO Ultrium tape in 2000, became in 2011 the first company to commercially produce tape using barium ferrite magnetic particles, and in 2025 debuted the Ultrium10, the first tape to use what it calls “fine hybrid magnetic particles”, combining barium ferrite with strontium ferrite. Each generation has brought meaningful increases in storage limits.
According to Fujifilm, “major cloud service and data center operators” have been expanding their use of magnetic tape because it stores large volumes of data at relatively low cost.
The second reason is the one the company leads with. Tape cartridges can be easily disconnected and stored in a drawer, a vault, or a separate facility. “Given the sophistication of cyberattacks, magnetic tape is attracting increasing attention as a means of protecting data from threats such as ransomware,” Fujifilm said. Ars Technica, which reported the announcement, summarised the appeal as low costs and offline security.
Tape has lost on almost every metric a consumer would apply to it. It is slow to search, physically bulky, and has to be handled by somebody. What it retains is cost per terabyte and the ability to be removed from the network entirely.
This is where the industry’s reasoning becomes worth following to the end. The most advanced computing in history requires an archive, and the archive’s principal security feature is that nobody can reach it.
The property being purchased, in other words, is inaccessibility. A cartridge in a drawer is safe from ransomware because it is safe from everything, including its owner. Security and usefulness are being traded openly against one another, and security is winning by five to one.
Our analysts have followed the gradient to where it leads. If disconnection protects data, the optimum is total disconnection, and the most secure archive is one that cannot be read at all. A tape in a vault in another building comes close. A tape that was never written comes closer.
The Department of Unnecessary Research accordingly proposes a Generation 11 that dispenses with the magnetic layer altogether. It would be cheap, immune to every known class of attack, and fully compatible with the existing drawer.
Fujifilm began making this material in the 1950s, when the fastest way to move a large quantity of data was to pick it up and carry it. Three generations of networking later, that is once again the recommended procedure. The tape has not changed. The threat model has come round to meet it.