Magnetic tape sounds like something out of a museum, yet it is where a growing share of the world's data ends up, and Fujifilm is betting big that AI will send far more of it that way. The company will invest about 10 billion yen, roughly $63 million, to raise its tape output fivefold against fiscal 2025 levels, its largest investment yet in dedicated LTO tape facilities. The reminder for anyone following the AI build-out is that it reaches well past GPUs: the training data and model outputs have to be stored somewhere cheap.
Fujifilm is blunt about the cause. It credits generative AI for what it calls exponential growth in data needs worldwide, and says cloud and data-centre operators are turning to tape because it is cheap at scale and, crucially, can be kept offline. An air-gapped tape sitting on a shelf cannot be encrypted by ransomware reaching across a network, which is a plain security argument that does not depend on any AI hype. The investment runs across fiscal years 2026 to 2028 at the company's Kanagawa factory in Odawara, and funds both the expansion and extra staff.
The product behind the bet is LTO Ultrium 10, the latest generation, which Fujifilm launched in 2025 using what it calls fine hybrid magnetic particles that combine barium ferrite and strontium ferrite. The company has made tape since the 1950s and shipped its first LTO Ultrium in 2000, so this is a decades-old business rather than a sudden pivot, and each generation has lifted how much a single cartridge holds.
The context explains why cloud giants never abandoned a format consumers forgot decades ago. Tape is slow to read, so it is useless for anything interactive, but for archives that must be kept for years and rarely touched, it is far cheaper per terabyte than disk and uses no power while idle. That is exactly the shape of the data an AI boom produces: vast, mostly cold, and worth keeping.
The honest caveat is what the fivefold figure is and is not. It is Fujifilm's planned capacity measured against last year, not confirmed orders, so it is the company's bet on where demand is heading rather than proof that demand has arrived. If the AI data surge slows, that capacity could outrun its market. For now it is a useful signal that the physical footprint of AI runs from the chip all the way down to a reel of tape in a dark room.