The company most people know for cameras and instant film is making its biggest bet yet on a storage medium plenty of people assumed was finished. Fujifilm will spend about 10 billion yen, roughly $63 million, at its Odawara site in Kanagawa to raise data tape shipment capacity to about five times its fiscal 2025 level. The work runs from fiscal 2026 through fiscal 2028 and covers new manufacturing equipment and more staff. Fujifilm calls it the largest investment it has ever made in dedicated LTO tape facilities.
Most of the money goes into FUJIFILM LTO Ultrium10, the current generation of Linear Tape-Open cartridges, which launched in 2025.
Why tape, why now
AI training and inference create huge amounts of data that has to be kept and is rarely read again: raw training sets, checkpoints, logs, old model versions, the archives behind retrieval systems. Keeping all of it on hard drives or flash costs too much, and both are in short supply. The memory squeeze that pushed SK Hynix to record margins and lifted Sandisk’s revenue almost fivefold in a year has spread to disks as well. Tape is the one storage tier where adding capacity is still relatively cheap.
Fujifilm names the buyers: big cloud providers and data center operators. Tape costs a fraction of disk per terabyte stored, draws no power while it sits on a shelf, and keeps data for decades. It can also be pulled off the network entirely. With ransomware gangs going after backups as a matter of routine, an offline copy on a cartridge is one of the few defenses nobody can reach remotely.
A film company’s chemistry
Fujifilm began researching magnetic tape in the 1950s and moved through broadcast videotape and consumer HD tape before launching LTO Ultrium1 in 2000. The core skill is the same one behind its photographic film: coating very fine particles evenly onto a thin base at speed. In 2011 it became the first company to commercialize barium ferrite particles for tape, which made the later jumps in capacity possible. Ultrium10 uses what Fujifilm calls fine hybrid magnetic particles, a blend of barium ferrite and strontium ferrite designs that stay stable at smaller sizes and pack more data into the same length of tape.
Fujifilm says it holds the top share of the tape market by production. LTO media is a very small club, so a fivefold capacity jump from the leader is a supply decision for the whole industry, and a signal of how much its biggest customers expect to archive by 2028.
What to watch
The plan runs to fiscal 2028, so most of the new output arrives in 2027 and 2028. Drives and tape libraries come from other makers, IBM and Quantum among them, and their order books will show whether hyperscalers are buying at the pace Fujifilm is planning for.
If they are, the AI storage boom has reached its cheapest and slowest tier. That usually means the expensive tiers are full.