Fujifilm to boost tape production 5x as AI fuels archiving demand
Fujifilm’s plan to quintuple magnetic tape production is seen as a response to exploding long-term storage needs from AI, big science, and ever-growing “never delete” data habits, rather than just traditional backups. Commenters weigh tape’s extremely low cost per terabyte and air-gapped resilience against its high drive prices, fragility, and operational complexity, comparing it with alternatives like cloud cold storage, hard drives, and Blu-ray M-Discs. Many conclude that while tape remains unmatched for large-scale archiving, there is still no affordable, robust archival solution for ordinary users.
Drivers of Tape Demand (AI, Security, “Never Delete”)
- Some see AI and “harvest-now-train-later” as a major driver: AI firms may be saving everything for future training, regardless of user consent.
- Others point to security: air‑gapped tape as defense against ransomware and “harvest-now-decrypt-later,” plus resilience to datacenter loss in wars.
- HPC and large experiments rarely delete data; hierarchical storage (NVMe → SSD → HDD → tape) makes long‑term tape attractive.
- Debate whether a 5× demand jump means there is 5× more truly valuable data, or just less incentive to curate and delete.
Backup Reality vs Theory
- Many organizations discover during attacks that backups are incomplete or unrecoverable; disaster‑recovery drills are rare compared to having “a backup process.”
- Common failure modes: unreadable incremental backups, lack of capacity to restore, corruption, misaligned drives, untested restore paths.
Tape Economics and Hardware
- Media is cheap per TB (e.g., LTO‑9 ~18 TB at mid‑single‑digit $/TB), but drives are very expensive (tens of thousands of dollars new).
- Libraries/robots are long‑lived capex; drives are niche, complex electromechanical devices, largely made by a single vendor, keeping prices high.
- Backward compatibility of LTO has worsened: older generations could read two versions back; newer ones dropped to one and LTO‑10 reportedly none.
Alternatives: Blu‑ray, HDD, Cloud, New Media
- M‑Disc/Blu‑ray praised for consumer/prosumer use and environmental robustness, but capacity is tiny vs tape and $/TB much higher for bulk storage.
- HDDs remain the main prosumer backup medium (often multiple copies), but require periodic migration and redundancy.
- Cloud backup (e.g., S3 Glacier Deep Archive) is seen as a tape‑backed, pay‑as‑you‑go option; concerns include long‑term cost, restore pricing, and reliance on providers.
- Experimental media like glass storage are viewed as promising but not yet practical or commercial.
Operational Practices and Reliability
- Experiences with reliability vary widely: some report almost no tape failures under controlled conditions; others see many issues at scale.
- Environment (temperature, humidity, dust, vibration) and maintenance (cleaning heads, cycling media) are emphasized as critical.
- Best practices: verify by actual restore, write multiple independent copies, avoid vendor‑specific compression/encryption, index contents, and use robotic libraries when possible.
Prosumer Gap and Social Effects
- Several lament the lack of a reasonably priced, robust tape solution for individuals/small businesses.
- Growing “never delete” culture and data hoarding are seen as environmentally and economically costly, but potentially a boon for future historians.