In the first of a two part series, I’ll look at the widespread adoption of cloud computing – which uses networks of remotely located equipment to manage, process and store data over the Internet – that is fueling huge investments in data centers. The largest of these facilities, filled with row upon row of tightly packed servers and data-storage arrays, requires more than 1 million square feet of equipment space, and can consume as much electricity as a small town.
More than 8 million data centers are now operating worldwide, according to IDC, as organizations large and small take advantage of the cost savings and service improvements available from outsourcing their technology infrastructure.
And even though the industry is now trending toward fewer, larger data centers managed by third-party cloud service providers such as Amazon, Alphabet, Microsoft and IBM, spending on data center facilities continues to accelerate. CBRE Group estimates that $18.2 billion was invested in new U.S. data centers during the first half of 2017, more than twice the amount spent in all of 2016.
A recent study from IDC shows that revenue generated from the sale of infrastructure products (servers, storage and switches) used for cloud IT reached a stunning $43.4 billion total for 2017, nearly 22% higher than in 2016. Dell Technologies, Hewlett Packard Enterprise, and Cisco are the industry leaders in selling hardware for these cloud installations, however there is rapid growth attributed to “white box” original design manufacturers (ODMs), particularly for custom designs for the major public cloud providers. Regardless of who is building it, analysts at Cisco predict that 95% of data center traffic will be generated by the cloud by 2021 – unbelievable!
AI boosts data center growth
Additional demand for data center facilities and equipment is being driven by the increasing use of artificial intelligence (AI) for speech and facial recognition, self-driving cars, genetic research and a growing list other data-intensive applications. AI “deep learning” techniques, in particular, are allowing users to rapidly analyze large volumes of unstructured data to reveal previously unrecognized patterns. IDC expects worldwide AI systems revenue to grow by 55 percent annually for at least the next few years, reaching $47 billion in 2020.
As AI is deployed more widely in business, government and science, technology suppliers are racing to develop more powerful server and storage systems and components that can help users harness more of this promising technology’s potential.
Processor designers, for instance, are developing new, more powerful and energy-efficient processor chips that are shaking up the status quo in the $50 billion server market. While Intel continues to dominate the server chip market, Nvidia’s sales to data center users grew 133 percent during 2017 to $1.9 billion as its graphics processors became increasingly popular for AI applications. That’s prompted AMD, Google, Intel, Qualcomm and others to rush their own new graphics, server and AI-specific chips to market.
Memory innovators, including Samsung, are also investing heavily in research and design, leading to dramatic increases in the storage capacity and performance of the latest memory components. Yet these added capabilities can’t come soon enough for today’s data center architects, who yearn for technologies that not only can keep up with their users’ increasing need for more data storage – but also provide faster data access and take up less physical space. One could argue that we are moving from a “CPU-centric world” to an AI-driven “data-centric world”, where memory and flash storage are becoming more and more critical to support new domain specific chips designed for emerging data-intensive applications.
As a Flash memory marketing guy, I understand how hungry data centers and other customers are for higher capacity, better performing data storage options. In the second part of this post, I’ll describe several recent advances in Flash memory technology that Samsung has been developing to meet the needs of the fast-growing data center industry.
For more on that topic, read my second part of this series.