Industry players argue that any threat might be more theoretical than real – at least for now.
Ulrich Voigt, vice-president of product management at media technology specialist Vizrt, works a lot with AWS, among other cloud providers – and he suggests that the benefits, including for sustainability and efficiency, outweigh any effect of their dominance. This is “anything but trivial”, especially when total costs of datacentre ownership only make sense at scale, he says.
“You actually get unlimited machines and optimisation,” says Voigt. “You don’t have to order and purchase a fully redundant system. And even smaller hosting providers have sweet spots – one is cheaper for bandwidth, another for the CPU.”
Two years ago, the distributed AWS offering for tape libraries passed the point where you could break even by purchasing a large, physical, automated tape library, he says. Data durability was also much higher than you could do on your own – and the hyperscalers often lead best practice on data management.
“In AWS, it was distributed geographically over three datacentres – highly redundant, completely managed and cared for,” says Voigt.
Does diversity need dominant players?
For Vizrt, AWS is always pushing for more encryption of video files, which is still “not easy” to achieve, he says. Dealing with video requires specific hardware approaches and cloud virtualisation, which entails differentiation when addressing the video needs of different verticals – medical versus entertainment, for example.
“And the more you put services into the cloud, the more you need good connections,” says Voigt.
He adds that overdependence on the hyperscalers can be partly avoided by using more than one – although having an architecture incorporating redundancy available at another hyperscaler is not easy either.
Voigt concedes that some do have concerns about the major cloud providers all being US companies – and that is not all about data sovereignty, which has been dealt with in the UK to some extent at least with post-Brexit US agreements.
But competition between the “three big guys” should keep the pressure on cost, with end-user customers that are tech providers then differentiating through their own added-on services and industry expertise, he says.
Gordon McKenna, chief technology officer for public cloud at managed services provider Ensono, describes himself as “a Microsoft guy for 25 years”. He feels the hyperscalers have helped drag people “kicking and screaming, if necessary” towards greater innovation.
“Those doing the disrupting are nearly all in the cloud, or mostly in the cloud,” he says. “Innovation is happening with the hyperscalers because of the scale, the focus, and the money they’ve got. I think businesses that struggle are often not taking advantage of that.”
Gordon McKenna, Ensono
Microsoft, Google and AWS already have highly containerised datacentres, which make it less and less important to use anything else, he says. Meanwhile, those that prefer hybrid, private or on-premise infrastructures can still choose that – even if they are benefiting from hyperscaler edge tech at the same time.
McKenna hastens to add that he is not “completely all on the hyperscaler side”, due to elements of lock-in that still, in his view, need to be addressed. AWS, Microsoft and Google all have proprietary cloud-native services, microservices or platforms where a migration essentially means a rewrite on another platform.
“For a very large e-commerce or ERP [enterprise resource planning] application, it is almost impossible sometimes to get out,” he says. “But if you want to talk about lock-in, the likes of IBM have been doing that for 75 years. Whether it’s on-premise or in the cloud, companies typically take advantage of positions.”
McKenna would like to see pressure from regulatory bodies on further standardisation where needed – such as to drive sustainability.
Chris Royles, field CTO at Cloudera, broadly agrees, noting that each hyperscaler has its own ecosystem of startups, consultancies and other providers that help customers have choice that meets specific requirements.
“Data and their workloads will be born in the cloud and natively integrated with the hyperscaler’s platform from day zero,” says Royles. “But the types of workload and data can end up in different places.”
Lock-in might be a thing – but choice remains
People might favour Google if they’ve got a website with Google Analytics and aim to generate and use data with the customer journey, says Royles. That lands in BigQuery “whether you like it or not”, with Google adding relevant toolsets and analytics around that.
Royles has found that a lot of organisations focusing on SAP-type workloads – ERP, CRM [customer relationship management], SAP Hana offload and the like – can choose AWS. “Amazon’s done really well at making that easy,” he says.
Conversely, retailers sometimes avoid Amazon because of its e-commerce heritage – while organisations with lots of Microsoft in-house already, such as the NHS, may gravitate to Azure. Business alignment, economics, politics and ease of doing business can all affect the choice, he says.
“Some want to utilise some of the hyperscaler resources and services, but won’t move my data there – so that’s a dynamic as well,” says Royles. “It’s about the ecosystem they build and how that then flows down.”
David Friend, Wasabi
David Friend, chief executive of object-storage services provider Wasabi, points out that when IBM had most of the storage market, no one thought others could compete. However, today’s datacentres have “dozens” of companies’ hardware all working together, often based on standards.
“Amazon has over 300 different products, from storage and compute to blockchain, gaming, analytics, facial recognition, AI, and on and on – and there are independent vendors in each category whose products are probably better,” he says.
Friend suspects that in five to 10 years’ time, hyperscalers may become the “Marks and Spencers” of cloud – a sort of non-specialist department store for commonplace requirements. For particular needs, especially where performance is a factor, customers can gravitate to specialists with expertise in a niche, he says.
“Yet there will always be a need for a form of one-stop shop where you can get it all in one place, on one bill, as well,” says Friend.
That said, he concedes that companies typically do attempt to prevent others nibbling away at profitable pieces of their market. The technique that Microsoft uses might be seen as a kind of bundling – with users of Office 365 often incentivised to buy Microsoft’s storage, compute, and other things as well. That, of course, is just “sales stuff”, he says.
“What bothers me is when a when a company blocks competitors from entry. If a vendor sells through all 14,000 channel partners, telling them that if they carry Wasabi, they’ll pull out, that would in my view be a restraint of trade,” says Friend. “A lot of companies try to do that. Then it’s appropriate for regulators to step in.”
Markets can ‘self correct’ at times
However, Friend too does not see any “monopolistic tendency” in public cloud that is not going to solve itself over time.
David Walker, field CTO at distributed SQL database company Yugabyte, agrees that the hyperscalers have been crucial to the development of cloud, pointing to changes in infrastructures and working practices.
Also, the hyperscalers themselves have worked to aid competition in some ways, including encouraging the roll-out of marketplaces for third-party product, he says: “That drives innovation – and also they’ve had to keep the market relatively open.”
In some countries, too, governments restrict hyperscalers to working with local partners. Overall, market diversity includes the kind of complexity and nuance that ensures multiple niches and flavours of cloud to best suit jurisdictions as well as customers, says Walker.
“A degree of regulation is good if you don’t want to be too big to fail – and being ‘too big to fail’ starts to play on terms like concentration risk,” he adds.
If one company does the majority of something and fails, there can be a huge economic impact, says Walker. When payments are involved, there can even be a run on the banks as people rush to withdraw cash to ensure they can still go about their daily business.
David Walker, Yugabyte
Using components that allow customers to move between cloud suppliers helps to reduce lock-in and concentration risk. Financial services, for instance, may have databases, Kubernetes applications, Lambda functions and the like written in such a way that they can lift up and move all those components through to any cloud provider, he says.
“You can do hybrid cloud. You can do multicloud. So we can build across these things,” says Walker.
“Cloud v.3” means that an ability to deploy core functionality can be deployed in any cloud, feeding through from the ongoing evolution of Kubernetes as well as “bags” of other players, not to mention common application programming interfaces (APIs) for deployment, he adds.
That counterbalances the drive for innovation by ensuring there is a core which is shared, which itself means you can take a risk with one vendor-specific offering for a time and experiment, and then move on, says Walker.
“And at some point that standardises and becomes part of the core, other vendors provide that service and then you move it,” he says. “You can keep on being competitive in the market with these things.”
So what is any upcoming Ofcom investigation likely to hinge upon?
Selina Chadha, Ofcom’s director of connectivity, suggests that the idea is to get in early and identify any emerging concerns with the £15bn UK cloud services market. Gartner has said its analysts think 45% of businesses’ IT spend could be on public cloud by 2026, Chadha noted in the related announcement.
“Collectively, these three firms generate around 81% of revenues in the UK public cloud infrastructure services market,” she says. “Our study will formally assess how well this market is working.”
Microsoft Azure declined to comment for this article. Google Cloud and AWS were contacted but did not respond.