Fujifilm Bets $63M On 100TB Magnetic Tape Storage As AI Data Swells
by
Aaron Leong
—
Thursday, October 29, 2026, 11:48 AM EDT
Magnetic tape storage is making a major comeback than ever in the age of AI - Image: Fujifilm
Fujifilm is launching a significant ¥10 billion (around $63.3 million) investment to scale up its magnetic tape production fivefold by fiscal 2028, backed by a surprising resurgence for archival storage driven by the artificial intelligence boom.
The funding, channeled into the company’s Odawara manufacturing facility, happens to be Fujifilm’s largest injection of capital into magnetic tape production in 20 years. While consumer tech still trend towards solid state drives and dense enterprise flash arrays, the staggering data requirements of modern generative AI are forcing hyperscalers and data center operators to look backward to move forward. Training massive foundational models and retaining petabytes of inference data require long-term, highly secure archiving solutions that do not incur the higher energy footprints or costs associated with keeping disks or flash memory running 24/7. Enter magnetic tape drives, which aren't as archaic as you might remember anymore.
LTO-8 tape drives developed by IBM and Fujifilm that can store up to 580 TB of data in a single cartridge - Image: IBM/Fujifilm
Current Linear Tape-Open (LTO) and barium ferrite media allow single cartridges to store an immense amounts of data, with new development pushing capacities toward 100TB and industry roadmaps charting a course all the way to 365TB per cartridge. When stacked against enterprise solid state options, which can cost 2-3 times more per terabyte, tape offers an economically unbeatable tier for cold data storage while being disconnected from online networks. Some studies have also shown that magnetic storage are also less polluting than other mainstream mediums, perhaps helping decrease the ongoing stigma against data centers in some instances.
Fuji's sudden interest comes after a brief dip in global tape shipments, followed by a sharp rebound as enterprise giants realized that discarding or prematurely deleting historical datasets was no longer an option, especially with the prices of flash and silicon storage as they are now and for the foreseeable future. After all, AI development relies heavily on cumulative data accumulation, meaning every scrap of training text, synthetic output, and historical log must be preserved indefinitely. For now, the humble magnetic tape has successfully transformed from a relic of early computing into a surprising cornerstone component powering the generative AI revolution.