Will HDD Shortages Slow AI Training And Implementation? – Forbes

Out of Stock store sign

getty

HDDs store the majority of the worlds information, particularly in data centers, but HDDs, like other storage (and memory) technologies have experienced a significant decline in demand during much of 2022 and in 2023 so far. This is largely because large data centers and enterprise customers bought more storage products than they needed at the time during the supply chain crisis in the COVID pandemic. As a consequence, HDD and SSD company stocks have suffered. HDD companies have cut back on production and laid employees off as demand slackened.

Finis Conner was one of the founders of Seagate Technology back in 1979 and he has been a founder of several other digital storage ventures since then. He is one of the pioneers in modern storage technologies. He recently wrote a blog in LinkedIn on a likely disparity between the demand for digital storage and particularly hard disk drives, in order to support data intensive activities such as AI training, and the availability of these storage devices after an extended period of low demand.

AI requires lots of processors to train models but the data used for training will likely live on SSDs as primary storage for data being actively processed but with HDDs as the inexpensive storage that stores the bulk of this data until required for processing. After an extended period of low demand resulting in cutbacks in production and employment, a sudden increase in demand for storage could quickly wipe out any excess in the industry and could lead to shortages that might hobble efforts to ramp up new AI models until storage producers could hire the workers, restart idled production lines and get their factories running at full speed again. This is particularly a problem for hard disk drives.

The article points out that major cloud storage (hyperscale data center) companies such as Google, Microsoft, Meta and AWS are spending billions of dollars building new data centers. Finis points out the likely storage bottleneck in implementing these new AI workflows. On one level, its connected to an HDD supply surplus from the pandemic. On a more concerning level, it brings the potential for an HDD supply crisis that could, temporarily, bring trillion-dollar companies to their knees.

A Fortune Business Insights article projects that the global data storage market is expected to grow from $247.32B in 2023 to $777.98B by 2030. Despite this projected growth HDD company stocks are down considerably from their 2022 highs. Seagate is down by 48%, Western Digital is down by 32% as is Toshiba. These companies have responded by lowing their production capacity and reducing personnel to reduce their costs, but lowering production capacity, immediately before the AI data storage demand tsunami strikes creates a dangerous situation.

Conner says that it will take a lot of time to start the HDD production engine back up again. While its true that HDD orders are retreating, the wave of AIand its demand for data storageis growing by orders of magnitude, and so is its ravenous hunger for data. It reminds me of a tsunami, and right now the disk drive industry is seeing the waters pull back, and wondering why. My serious question is, will the three HDD makers be ready for it when the tide eventually turns, and the big wave hits?

There are signs that these shortages may be starting to impact industry. The article says that last month Toyota unexpectedly stopped production at 28 assembly lines in 14 Japanese auto plants because an error occurred due to insufficient disk space, causing the system to stop.

To show this demand, Finis referenced a chart from an August 2023 blog by this author, shown below. AI and other big data analysis will drive demand for all storage devices, including HDDs. This will be served by shipping higher capacity HDDs, but that alone is not enough to meet capacity demand and the total number of HDDs shipped for enterprise and data center applications will need to increase.

There are currently three HDD companies, but it is possible that this number could decline to two companies in the coming months as Toshiba is acquired in a $14B deal that will take the company private and perhaps get rid of currently low or non-profitable businesses such as its HDD division. This could further restrain the supply of HDDs.

All these factors led Finis to speculate that HDD companies are undervalued. These mergers and acquisitions in the HDD sector are timely. They show change is already in process. But more than that, consider the roughly $40 billion value of the three HDD manufacturers and compare it to the value of the cloud data storage businesses of Apple, Google, and AWS combined. Thats over $6 trillion. These numbers dont add up to the enabling value HDD suppliers bring to the table. It seems like nobody is paying attention to the importance of these companiesand the overall fragility of the HDD supply chain.

Finis points out that this creates a potential opportunity for HDD companies. If they can equip themselves to ramp up production and make supply chains more resilient, they may be able to help meet this imminent and growing demand for storage. He thinks one way to do this is to bring some production back to the US, similar to current semiconductor reshoring efforts.

My own take is that demand for storage driven by AI and other processing trends will likely use up the excess HDD inventory by the end of 2023 and that we will see significant growth in demand for HDDs in 2024 and future years to support the widespread training and utilization of AI models and other big data processing tasks.

HDD pioneer, Finis Conner, believes that HDD companies may be undervalued and that demand for digital storage to support the training and use of AI models which will transform all aspects of business and work could be hampered by shortages in the near future.

I am the President of Coughlin Associates and a storage analyst and consultant. I have over 40 years in the data storage industry with multiple engineering and management positions. I have many publications and six patents to my credit. I am also the author of Digital Storage in Consumer Electronics: The Essential Guide, the second edition was published by Springer. Coughlin Associates provides market and technology analysis (including reports on several digital storage technologies and applications and a newsletter) as well as Data Storage Technical Consulting services. I publish a Digital Storage in Consumer Electronics Report, a Media and Entertainment Storage Report, and an Emerging Memory Report. I am active with SMPTE, SNIA, the IEEE and other professional organizations. I was the founder and organizer of the Annual Storage Visions Conference (www.storagevisions.com), as well as the Creative Storage Conference (www.creativestorage.org). I was general chair of the Flash Memory Summit for 10 years and am currently the program chair. I am also a Fellow of the IEEE, IEEE President Elect in 2023 and a member of the Consultants Network of Silicon Valley (CNSV). For more information about me and my publications, go to http://www.tomcoughlin.com.

View original post here:
Will HDD Shortages Slow AI Training And Implementation? - Forbes

Related Posts

Comments are closed.