It may be difficult to remember, but there was a time when a company’s entire data center might have been run from a lone server within an office cabinet.

Then came hypervisors, whose use as abstraction layers let apps in virtual machines (VMs) be relocated from one physical hardware installation to another—a key moment in the rise of hyperscale data centers.

In most cases, the original, on-premises data centers of old simply cannot handle the volume of data now generated, especially that created by hyperscaled applications.

Hyperscale data centers (also called hyperscalers) occupy considerably larger physical space than traditional on-premises data centers, which have tended to be sized somewhere in the 10,000-square-foot range.

According to the IDC definition of a hyperscaler (link resides outside ibm.com), to be considered a true hyperscaler, a company must use 5,000 servers or more and devote at least 10,000 square feet to the operation.

Hyperscale facilities are often multiples beyond that—with building sizes often realizing dimensions approaching 60,000 square feet—roughly the size of a regulation US football field.

And while that may be a typical hyperscale size, it’s not nearly the largest example of storage options. That distinction belongs to the big China Telecom data center (link resides outside ibm.com) located in Horinger, Hohhot, within China’s Inner Mongolia region. This facility, which cost USD 3 billion to construct, covers a staggering 10.7 million square feet and uses 150 megawatts of power. (To visualize a facility of this enormity, try to imagine the combined space of roughly 165 adjoined football fields.)

Hyperscalers are largely cut from the same cloth as traditional data centers—but just geared to a much larger scale than typical on-premises data centers. They do this by building and running an enormous hardware and software infrastructure in the hyperscaler facilities. Millions of servers distributed across many data centers provide seemingly endless storage and computing resources.

Since data traffic can fluctuate wildly—especially when running huge applications—hyperscalers accommodate that traffic and stabilize hyperscale operations when increased demand is exhibited within a computing environment. Hyperscalers do this by essentially serving as a form of load balancer, juggling tasks and re-directing computing resources as needed.

As of Q1 2023, AWS achieved a 32% market share (link resides outside ibm.com), making it the largest provider of hyperscale cloud services. Its cloud-themed products focus on aspects like storage options, computing power, automation, databases and data analytics.

Many companies are already consumers of one Microsoft product or another, so those businesses may have an established comfort level with the company and its offerings. That includes Microsoft’s enterprise software, which integrates nicely with Microsoft Azure, its hyperscale product. (Market share: 23%)

Google rose to power and prominence through its mastery of data handling and businesses still seek out GCP (about 10% of the market), especially if those companies are eager to engage in advanced analytics and extend their footprint into artificial intelligence (AI).

With long-established technology bona fides, IBM Cloud leverages the company’s expertise in many areas, like AI and dealing with enterprise data centers. IBM Cloud delivers comprehensive services for Infrastructure-as-a-Service (IaaS), Platform-as-a-Service (PaaS) and Software-as-a-Service (SaaS).

OCI’s advantages include ease of migrating critical enterprise workloads and the ability to construct cloud-native apps. Another selling point for OCI has been its aggressively low pricing, with Oracle claiming to offer the same basic services as AWS at a fraction of the cost.

The quick overview summary suggests that AWS excels because of its global access and advanced scalability. GCP is attractive to businesses who need top-level data management and want to make forays into capabilities like machine learning, and Microsoft’s Azure provides smooth product integration and enhanced security.

Beyond the top three competitors, IBM Cloud is sparking considerable interest with its current work with AI, which informs its hyperscale offerings. And Oracle’s been serving OCI up as a platform built to save money, and host cloud-native applications.

As is evident, each of these providers differs significantly in their approach and their particular areas of expertise, but they do share similarities. For example, the top three of these providers now offer cloud-native services supporting zero-trust network access (ZTNA) protocols that are designed to offer an alternative to VPNs, which can experience security vulnerabilities.

Not all businesses can afford the substantial financial investments needed to set up elaborate hyperscalers or wish to make such a financial commitment. Remember, we’re not simply talking about construction costs. There are massive equipment purchases to be considered as well.

It’s also been observed that some hyperscale facilities use more power than small cities. So, electricity costs and environmental concerns usually must also be factored into the company’s strategy about hyperscaler use.

As an alternative to purchasing equipment, some companies still engage in the practice of colocation, in which servers or other computing equipment are instead rented.

Hyperscalers have also made an impact when it comes to Internet of Things (IoT) devices and how they’re managed. The fact that hyperscale clouds are being used in conjunction with pre-existing equipment (in many instances) helps drive down the price tag of infrastructure investments for the IoT ecosystem—making it a better value for companies.