AI News

Fujifilm expands LTO tape production: AI cold data archiving demand drives tape market growth

cnBetaUpdated Nov 8 02:10

AI summary

Fujifilm is making its largest investment in nearly 20 years in facilities dedicated to LTO tape, mainly due to the massive data archiving demand brought by generative AI. Data from the LTO Consortium shows that global LTO tape shipment capacity reached 176.5 EB in 2024, a year-on-year increase of 15.4%, and in the first quarter of 2026 it grew by 57% year-on-year. Fujifilm estimates that storing 100 PB of data for ten years with an automated tape system reduces total cost of ownership by about 86% and carbon emissions by about 95% compared with an all-hard-disk-drive solution.

Body · AI translation

According to Fujifilm, this is the largest investment in LTO tape-dedicated facilities in nearly 20 years, and the main driver behind the expansion is precisely the massive data archiving demand brought about by generative AI.

There is a classic joke circulating in the computer industry.

Computer scientist Andrew Tannenbaum once said that you should never underestimate the bandwidth of a car speeding down the highway loaded with tapes. It sounds a bit absurd, but when the data that needs to be transferred reaches the PB or even EB level, loading storage devices onto trucks may indeed be more efficient than transmitting over a network.

In 2016, Amazon AWS launched the Snowmobile service, which directly used heavy trucks to transport large storage devices, moving up to 100PB of data at a time.

Although Snowmobile did not transport tapes, and the service was discontinued in 2024, it illustrates a long-standing problem. Once the scale of data reaches a certain point, how to preserve and move data at a reasonable cost is itself an engineering challenge.

Training large models requires GPUs, HBM, and high-speed SSDs, but after training is complete, enterprises are left with raw corpora, cleaned datasets, training logs, model checkpoints, and historical versions generated through continuous iteration.

Among these, data currently involved in computation requires high-speed storage devices, frequently accessed large-scale datasets can be placed on mechanical hard drives, and materials that are not used for months or even years but cannot be casually deleted are suitable for entering long-term archiving systems.

The industry usually refers to data that is accessed very infrequently as cold data.

The problem then arises. As data accumulates, should all hard drives be kept powered on, spending money to maintain files that may not be accessed even once in several years?

Magnetic tape happens to be good at this.

First, it is cheap, and it saves power. Tapes can be stored offline for long periods, require no continuous power supply when idle, and do not need to keep equipment running and dissipating heat like online hard drive arrays.

Fujifilm once calculated that, under specific conditions of preserving 100PB of data over a ten-year period, an automated tape system can reduce total cost of ownership by about 86% and carbon emissions by about 95% compared with an all-mechanical-hard-drive solution.

Second, it is secure. Tapes can be removed from the system and stored offline; even if servers are attacked by ransomware, physically isolated tapes still have a chance to retain an independent backup.

Moreover, modern tapes are not as slow as one might imagine. LTO-10 has a maximum raw continuous transfer rate of up to 1200MB/s, and automated tape libraries can also use robotic arms to complete loading and reading.

However, tapes need to search for data sequentially, and randomly reading a single file may take a relatively long time, so they are not suitable for frequent access and real-time computing. If SSDs in AI data centers are responsible for making data available on demand, then tapes are more like a low-cost, massive-capacity digital warehouse.

In fact, the scientific community has long been using this approach.

In December 2025, CERN announced that the Large Hadron Collider had cumulatively preserved 1EB of experimental data, equivalent to 1 million TB, most of which is stored on about 60,000 tapes.

Experimental data collected decades ago may regain research value due to the emergence of new theories. CERN estimates that the scale of data that will need to be preserved and processed in the next decade will reach about ten times the previous cumulative storage volume.

And the growth of the tape market is already supported by actual data.

According to data published by the LTO Consortium in September 2026, global LTO tape shipment capacity reached a record 176.5EB in 2024, a year-on-year increase of 15.4%. Although it declined somewhat in 2025, it still reached 160.3EB.

Entering the first quarter of 2026, with the launch of LTO-10 and the growth of archiving demand, global tape shipment capacity increased by 57% year on year. What is counted here is the storage capacity provided by tapes, not tape cartridge sales.

A single tape can store 100TB, with capacity soaring 400-fold in 26 years

When ordinary consumers assume that tape technology has long stagnated, engineers have in fact been continuously increasing its storage density.

In June 2025, Fujifilm launched the LTO-10 30TB data tape, and in December of the same year it released a 40TB version, which began shipping in 2026. The latest product reaches a raw capacity of 40TB per cartridge; calculated at an ideal compression ratio of 2.5:1, the theoretical capacity can reach up to 100TB.

Image | LTO-10 30TB

By comparison, the first-generation LTO tape launched in 2000 had a raw capacity of only 100GB. In just 26 years, the capacity per cartridge has increased by a full 400 times. However, 100TB is the theoretical compressed capacity; for video, images, and model weights that have already been compressed, the actual capacity usually cannot reach this figure.

What is even more interesting is that an LTO-10 tape cartridge is only about 10 centimeters square and 2.15 centimeters thick, yet inside it is rolled a tape 1337 meters long, equivalent to more than three laps around a standard athletics track.

To fit a tape 1337 meters long into a cartridge only about 10 centimeters square, Fujifilm introduced aramid film and new thin-layer manufacturing technology, reducing the overall tape thickness to just 4 micrometers, while adopting ultrafine magnetic particles that combine barium ferrite and strontium ferrite technologies to increase the amount of data that can be recorded per unit area.

As for the storage principle, it is in fact in the same lineage as the audio tapes of decades ago.

The tape surface is covered with tiny magnetic particles. When writing data, the head records digital information by changing local magnetization states; when reading, it detects changes in the magnetic field and restores the data through decoding and error correction.

Modern LTO tapes use linear serpentine recording, which you can imagine as a highway with tens of thousands of lanes. The head writes from beginning to end along one set of tracks, then repositions and writes back in the opposite direction.

The latest LTO-10 has 15,104 data tracks, and the drive can read and write 32 tracks simultaneously, continuously increasing capacity through more precise materials and servo control technology.

Moreover, tape capacity growth is far from over.

In 2020, IBM and Fujifilm achieved a magnetic recording density of 317Gb per square inch in the laboratory, demonstrating the technical potential for a single tape cartridge to reach 580TB in the future, equivalent to 14.5 times the current 40TB commercial product.

Of course, 580TB remains at the laboratory demonstration stage, and there is still a long way to go before commercialization.

Tape also has a long-term cost that is often overlooked: compatibility.

Tape media can be preserved for many years, but that does not mean a new drive decades later will still be able to read it. The latest LTO-10 drive can only read and write LTO-10 tapes, so enterprises need to retain old equipment or periodically migrate historical data to new media.

For data centers managing tens of thousands of tape cartridges, data migration, verification, and equipment maintenance are likewise considerable expenses.

Even so, in large-scale cold data archiving scenarios, tape still has strong cost and energy-efficiency competitiveness. Fujifilm's precision coating, thin-film materials, and magnetic particle technologies accumulated over the past decades have therefore regained commercial value in the AI era.

The story of tape is also playing out in the hard drive industry.

In January of this year, Western Digital revealed that its hard drive production capacity for calendar year 2026 was basically sold out, and long-term procurement agreements with some large customers had already been extended to 2028.

Seagate's situation is even more extreme. According to its late-October earnings call, the vast majority of the company's nearline hard drive shipment capacity had been allocated through long-term agreements to 2028, and procurement agreements with some cloud service providers even cover through 2030.

What is more noteworthy is that hard drives are not only selling more, but prices are also rising.

The reason is not complicated. SSD random read/write performance far exceeds that of hard drives, but for cloud service providers that need to keep massive amounts of data online for the long term, hard drives still have a cost-per-TB advantage.

In particular, as the AI industry continues to expand and training data, model files, and various types of generated content keep accumulating, data centers need not only faster SSDs but also hard drives that can provide tens of PB or hundreds of PB of capacity at a reasonable price.

Of course, HDDs also have performance and maintenance costs. As per-drive capacity continues to increase, random access capability does not improve proportionally, and data rebuild time after a failure may also lengthen.

But for cloud service providers, as long as storage cost, capacity density, and access performance can meet business needs, hard drives still have a reason to exist.

From Fujifilm's announcement of a fivefold capacity expansion, to Western Digital's full-year capacity being basically sold out, to Seagate signing long-term agreements with cloud service providers extending to 2030, a group of traditional technologies that have gradually lost relevance in the consumer electronics market are finding new growth space in AI data centers.

However, tape may also face new competitors in the future.

In February 2026, Microsoft Research published the latest results of Project Silica in Nature. Researchers used femtosecond lasers to write data inside glass, achieving a storage capacity of 4.8TB in a piece of glass about 12 centimeters square and 2 millimeters thick.

Based on accelerated aging tests, the researchers estimate that the data could be preserved for more than ten thousand years. Although there is still a distance before large-scale commercial application, it demonstrates another possibility for long-term archiving.

On one side, Fujifilm is still improving the magnetic powder coating process that has continued for decades, enabling a small tape cartridge to hold tens of TB of data; on the other side, Microsoft has begun trying to use lasers to write information into glass, much like ancient humans carving characters into stone.

As the whole world chases more powerful GPUs and smarter large models, those old technologies responsible for preserving data have instead ushered in their own golden age.

The spotlight never belongs only to new things; it also belongs to those old crafts that find their place in new trends.

Source

cnBeta