6 Failed Storage Technologies: What Microdrive, HiFD and Holographic Discs Teach Us

The article's argument is that technical brilliance does not guarantee a market. That holds up. But a few of its claims need tightening, and one needs correcting. Here is the fuller picture.

The big idea: great engineering, wrong moment​

These six were not all duds. Several worked, and some shipped in real numbers. What they lacked was some mix of cost, speed, reliability, compatibility, manufacturing scale or timing. Treat the list as six different failure stories, not one.

IBM Microdrive: a hard drive that arrived as flash overtook it​

EE Times reported in September 1998 that IBM planned a one-inch hard drive in the CompactFlash form factor. It was to be a single-platter 170MB or 340MB drive with a standard ATA interface, shipping in 1999. It measured 42.8 by 36.4 by 5.0mm and spun at 4,500 RPM.

The drive's pitch was cost per megabyte and power use. Analyst Jim Porter said IBM would likely sell at a "small fraction" of flash's cost per megabyte. IBM said the drives would use less energy than flash chips, which must expend power to erase before writing.

Mechanical parts were the catch. IBM planned head parking and other "ruggedizing" measures to stop the head scratching the platter when the device was jostled. Flash has no head to protect. Apple's iPod nano launched on September 7, 2005, in 2GB and 4GB models, and that is a useful marker for the shift to solid-state storage in small players.

How-To Geek says later Microdrives reached 8GB. That figure comes from the original article. I could not independently confirm it, so treat it as plausible but unverified here.

IBM 3850 Mass Storage System: a robot librarian for data​

Columbia University's computing-history archive describes the 3850 as the "monster storage device of its day". Its installation held the 1980 US Census database: 2,000 cartridges of 50MB each, for 102.2GB. The cartridges lived in honeycomb cells. A mechanical "hand" fetched them and took them to a reader/writer. Data was staged through disk drives, more or less transparently to users.

That is the key to why this was not a simple tape-versus-disk story. IBM's own retrospective calls it one of the first hybrid disk-and-tape systems, an automated tape library "pretending to be disk". It was designed to hold infrequently used data and to look like disk to the host.

Wikipedia says IBM announced the 3850 on October 9, 2024... in fact on October 9, 1974, and withdrew it on August 5, 1986. That is secondary sourcing, but it matches the 1974 date in IBM's retrospective. Capacity varied by model. The smallest held 706 cartridges (35.3GB) and the largest 4,720 cartridges (236GB). A later series roughly doubled the maximum. So "hundreds of gigabytes" is fair, but the Columbia figure describes one installation, not every unit.

The same page notes that staging a cartridge to disk typically took about 15 seconds. That delay is the real cost of the honeycomb. Hard disks kept getting faster and cheaper, and the 3850's size, complexity and cost stayed put.

The idea itself did not die. Automated tape libraries, hierarchical storage management and tiered storage are everyday enterprise practice. The 3850 lost as a product, not as a concept.

Sony HiFD: the floppy killer that couldn't stay alive​

Digi.no reported in October 1998 that HiFD used a 3.5-inch disk with 200MB formatted capacity (240MB unformatted). It promised up to 3.6MB/s and backward compatibility with 1.44MB floppies, so users would not need a separate floppy drive. The first external unit was to cost $199, and media about $14 per disk. The launch had already slipped several times.

Wikipedia's account says the drive suffered crashes and very slow read speeds a few months after launch. Sony then issued a full recall. HiFD returned in November 1999 over USB. By then Zip drives offered 250MB and CD-RW was entering the mainstream, and the format's reliability reputation was damaged. Sony discontinued it by early 2001.

The "every single unit" claim in the original feature is not something I could verify. The recall and the reliability problems are well supported.

Bubble memory: promising, then outpaced​

Bubble memory stored bits as small magnetized regions in a thin magnetic film. Texas Instruments' first commercial chip held 92 kilobits, and Intel later pushed capacities higher.

The original feature frames bubble memory as a hard-drive competitor that lost on price, speed and capacity. That is broadly the market outcome. But bubble memory was non-volatile solid-state memory with distinct characteristics, not a drop-in disk. It also competed with other memory technologies. It failed to displace its rivals at scale, which is a safer summary than a single cause.

InPhase Tapestry: a correction​

The original article says Tapestry "never made it to market". The record is messier than that.

• Computerworld reported that InPhase announced bulk shipments of a 300GB drive. The initial Tapestry HDS-300R drives had shipped in December to beta users, in a write-once format aimed strictly at archival markets.
• TechCrunch reported in February 2007 that InPhase had begun bulk shipping 300GB disks at $180 each, on top of an $18,000 drive.
• TechCrunch noted the 20MB/s access speed against 150MB/s for readily available SATA drives.
• SPIE's professional magazine reported in October 2007 that InPhase was taking orders for its holographic drives.
• An earlier SPIE report cited shock and vibration susceptibility and the reliability of the field-replaceable laser as development problems.
• Wikipedia says the company failed several times to release the reader on schedule, after setting dates of late 2006 and then February 2007.

Sources disagree on timing and scale, and some of this is announcement rather than confirmed volume. The defensible reading is that Tapestry was announced, delayed, and at best shipped in small, expensive quantities to archival buyers. It never became a lasting, broadly adopted format.

The 1.6TB figure was a roadmap target. The company said capacity would expand to 1.6TB per disc within three years. It was never an achieved product capacity. InPhase filed for bankruptcy protection on October 17, 2011, with its assets sold in March 2012 and the technology passing to Akonia Holographics. A company failure does not prove the underlying physics was a dead end.

DataPlay: the tiny disc without an ecosystem​

DataPlay showed off postage-stamp-sized, cartridge-protected optical discs at CES in January 2001. The pitch was 500MB per disc, priced at $5 to $10, for music players, cameras and other devices. Twice reported that mass-production delays in the miniature optical engine and in factory setup pushed the launch from 2001 to early 2002. Wired covered the 2002 market arrival, which came with record-label partners.

A 2004 report says the company went bankrupt in late 2002 and its intellectual property was later acquired. That is secondary reporting, not a verified filing.

How-To Geek blames the rise of the iPod, the cost of the players, the DRM and the entrenched CD. Those are reasonable factors, but the evidence I could verify does not show that DRM alone killed it. A new disc format needs players, content and consumer trust all at once, and DataPlay had to build all three.

What this means for today's storage buyers​

The practical lessons transfer to modern PC and IT decisions:

1. Capacity targets are not shipping products. Roadmap numbers, such as InPhase's 1.6TB, are goals.
2. Ecosystems beat specs. Players, drives, drivers and software support decide whether a format lives.
3. Reliability on day one matters. HiFD's recall shows how fast a reputation can sink.
4. Ideas outlive products. The 3850's automated tiering concept survives in modern archive tiers. Flash won the Microdrive's job, but small mechanical drives lost mainly because a solid-state alternative arrived.
5. Beware single-cause stories. "Hard drives were just better" is tidy but rarely complete.

The same filter is worth applying to the next storage technology that promises to change everything.