According to Synergy Research Group, the world’s largest hyperscalers collectively own 700 data centers. Amazon, Google, IBM and Microsoft each have at least 60 locations. While it has taken five years for the number of hyperscale data centers to double, it has taken less than four years for capacity to do the same. When measured by critical IT load, the US accounts for nearly half at 49%. Their size, reach and power cannot be overstated; Even the largest mobile operators cannot compete with them.
In the past, these hyperscalers were just providing Infrastructure as a Service (IaaS). Now, with Amazon, Google, Oracle, Microsoft and IBM (among others) beginning to integrate on-premises, edge and public cloud infrastructure, they are moving into the intelligence layer - artificial intelligence (AI), machine learning (ML), analytics, 5G core, Communications Platform as a Service (CPaaS) - and drifting into the telecom space. AWS, for example, launched its AWS Private 5G managed service in December 2021 to help businesses deploy and scale up private 5G mobile networks in their facilities. AWS is working with partners to provide components from the small cell radio units and servers to 5G core, RAN software and SIM cards for connected devices. It’s important to note that AWS 5G Private is still limited to the small-to medium-size enterprises because anything outside of the of the on-premises coverage - think walking between two factories on a campus - will require public spectrum and upgrades.
This creates a kind of love-hate relationship between the two, or a codependent one at the least. Hyperscalers have millions of enterprise customers - which is a good avenue to sell private 5G and other services - but they still need the footprint and the edge data centers of the operators. On the other hand, the operators need the hyperscalers to actually get to the enterprise market.
All of this to say, it’s not necessarily whether they are friend or foe but rather two entities that must coexist. That's why we are seeing a lot more of these hyperscaler/operator partnerships and integrations. IBM Cloud for Telecommunications, for example, launched in November 2020 to provide the unifying architecture and data sovereignty that telecommunications companies need to transform and innovate. While operators are becoming increasingly software-based, agile platforms—they are deploying proprietary or disparate systems, which makes it difficult to cost-effectively scale in order to deliver more data and media services and adhere to industry standards and regulations.
Google is trying to do the same thing in Europe. In September 2021, Google Cloud and T-Systems announced they will build and deliver sovereign cloud services for German enterprises, the public sector and healthcare organizations, and allow them to host their sensitive workloads on a sovereign cloud. Amazon Private 5G is taking a different path that remains in the public cloud regions. But while each have different approaches, they're all playing in the same space - trying to get into the connectivity layer and the whole symbiotic relationship with the operator for edge computing and 5G IoT mitigation services use cases.
Whether running on a private network or in a hyperscale environment, the main concerns are privacy, security, performance and service level agreements (SLA). It’s not that they don’t have problems, rather these are the areas that require attention. For telco operators running on public clouds, they want five-nines availability SLAs from services providers to ensure no more than five minutes and 15 seconds of downtime in a year because typically hyperscalers won’t give performance-based SLAs. Instead, they’ll state that their infrastructure services are below 4-9s (unavailability limited to 52 minutes and 36 seconds per year). The operator, on the other hand, will want to then run an application on top to give a higher level of performance. Those are the kind of things that will be on the operators’ minds as they look to switch from the current on-prem to hyperscale environment. The first thing is use case determination: What things do they launch in public cloud? Then service-wise, they need to be cognizant of the SLA, performance, redundancy methods—as well as meeting security, privacy and regulatory requirements.
For traditional telecom operators, it’s only a matter of time before they must parse out some of the network elements to hyperscalers. The public cloud’s cost sharing infrastructure and services make it cost effective to run the telco workloads in public clouds than private cloud. In June 2021, AT&T announced it was moving its 5G mobile network to Microsoft Azure cloud, starting with AT&T’s 5G core (the software at the heart of the 5G network).
Hyperscalers now have the ability to roll out new hardware so much faster than operators. Consider trying to get the latest instance of a chipset on a server. It will take a telecom operator a year - versus hyperscalers that come up with new versions of the chipset frequently and are often much better optimized.
The trend of moving to public cloud is in rhythm with the overall shift to software as the foundation of the mobile network. For the entire telecom ecosystem, the process of moving telco workloads to public cloud will benefit from the flexibility of companies that specialize and architect solutions for any cloud. That’s the impact and opportunity the hyperscalers bring to the mobile market.
