Hyperscale cloud solutions have come under increased scrutiny in recent years. Reports of companies considering alternatives to hyperscale cloud and cloud repatriation reveal fundamental issues in hyperscale cloud environments that, for many businesses, have become chronic.
Problems like escalating costs, platform complexity, security and support concerns have led to an ongoing conversation around reverse cloud migrations and whether hyperscale cloud-first strategies should still be the default choice when it comes to a business’ infrastructure.
Despite more talk of cloud exit strategies and companies leaving the cloud, hyperscale cloud providers are nowhere near the brink of collapse. The global hyperscale cloud market was valued at $320.59 billion in 2023 and is expected to reach $1.44 trillion by the end of 2029 – not figures that would suggest the hyperscalers are going away any time soon.
These two trends might seem like a contradiction but, the reality is that they’re playing out simultaneously. Demand for hyperscale cloud services continues to grow whilst, at the same time, the way some businesses view and use hyperscale cloud infrastructure is fundamentally changing.
One of the best examples of the shift in how businesses use hyperscale cloud is the increased adoption of multi-cloud and hybrid cloud strategies. Multi-cloud strategies, involving the deployment of infrastructure across multiple clouds, have now been taken up by 86% of organizations. And by the end of 2024, it’s predicted that 90% of enterprises will have adopted a hybrid infrastructure strategy, combining multiple compute types.
Organizations want the agility and flexibility that comes with hybrid and multi-cloud approaches. Distributing data and workloads across multiple service providers (or in combination with on-premises hardware), not only helps to optimize costs but comes with increased redundancy, data portability and additional control over security and compliance.
“Hybrid cloud is more than a legitimate strategy,” said Joe Baguley, EMEA CTO at VMware, to Dark Matter. “Most organizations you talk to are already in a hybrid state. I think the next step to that is being consciously aware that you have to have a solid, multi-cloud strategy”.
What’s considered the infrastructure ‘ideal’ is shifting. IT leaders are realizing that the best-case-scenario doesn’t lie in a cloud-first strategy. And, in many cases, it doesn’t lie in a full repatriation either. Best practice involves working out the best place for each workload.
“The question should be: Where do I want to put my workload?” said Taj Patel, Vice President for IT and CIO at Stevens Institute of Technology. “The infrastructure exists to meet what the business wants to do. A hybrid environment can help deliver this”.
Hybrid infrastructure environments combining private cloud and public cloud are also being used to maintain control over costs and data integrity as organizations start to embrace artificial intelligence as part of their workflows and product offerings.
“The excitement and related fears surrounding AI only reinforces the need for private clouds,” said Dave McCarthy, Cloud Analyst at IDC. “CIOs are working through how to leverage the most of what LLMs [large language models] can provide in the public cloud while retaining sensitive data in private clouds that they control”.
Software as a service (SaaS) – a software delivery method allowing end users to access applications over the internet via a subscription model – has boomed over the past decade. It’s now estimated that 85% of business apps will be SaaS-based by 2025. And as one of the three key types of cloud technology, alongside infrastructure as a service (IaaS) and platform as a service (PaaS), it naturally follows that SaaS applications have, to date, been hosted predominantly on hyperscale cloud.
With free credits and instant provisioning, the hyperscalers offered a perfect environment for rapid market entry. But as SaaS vendors began to shift from a scalable growth mindset to a sustainable growth mindset (fueled in part by a post-pandemic slump), some found that that the hyperscalers were actually getting in their way.
This is what happened to Dukaan, a SaaS-based ecommerce enablement platform that helps local store owners sell their products online. Hyperscale cloud infrastructure was fundamental in supporting the startup’s initial market entry and helped the team launch their first minimum viable product in just two days. But soon after, mounting cloud bills of close to $90,000 started to consume Dukaan’s budget. To grow sustainably, the team needed to reduce infrastructure spend dramatically.
By adopting a managed bare metal solution, they were able to reduce their monthly infrastructure bill to $1500 and improve the reliability of their infrastructure stack in the process. Dukaan’s gargantuan savings are an extreme example, but nevertheless demonstrates the significant impact of hyperscale cloud on business’ bottom lines. (If you’re looking to understand the ins and outs of Dukaan’s migration, Founder and CTO Subash Choudhary discusses their cloud exit at length in this video with Arpit Bhayani).
Examples like this one reflect a wider infrastructure culture shift amongst SaaS vendors. In the recent past, it was near guaranteed that any given SaaS product was being hosted on hyperscale cloud. But now, as more vendors seek ways to optimize their infrastructure spend and regain control over their technology, this is simply no longer the case.
The shift in how organizations are using public cloud infrastructure also comes down to shifts in technology. The boundaries between traditional bare metal hosting and hyperscale cloud infrastructure are blurring.
“It all started with [hyperscale] cloud (which was very expensive but very flexible) and bare metal (which was very inexpensive but very inflexible). Now these two things are coming closer together,” said Liam Brennan, Director of Games Technology at Tencent. “Cloud vendors are struggling to some degree because they are very expensive…bare metal is becoming more flexible with hourly pricing models and hourly cancellation policies”.
With services like bare metal cloud, offering bare metal hardware as an on-demand service, organizations have more choice when it comes to their infrastructure. Now able to access dedicated hardware with a level of flexibility and scalability akin to public cloud, more organizations are considering bare metal hosting as a legitimate alternative to the big hyperscalers.
Virtualized IT environments have formed the bedrock of IT strategies since the turn of the millennium. But growing concerns around the reliability and security of these environments, along with high virtualization costs, is leading to a reverse push towards ‘devirtualization’, according to Gartner’s 2024 Hype Cycle for Data Center Infrastructure Technologies report.
It’s something that’s come to the fore since Broadcom’s acquisition of VMware. In reaction to the associated price hikes, organizations are starting to challenge the need for virtualized IT architectures, bringing alternative solutions like bare metal cloud into the fold.
Is the hyperscale cloud going away? Well, no. In fact, the numbers show that the opposite is true. It’s been forecasted that that the average capacity of hyperscale data centers will more than double over the next six years. So, whilst talk around the limitations and challenges of hyperscale cloud infrastructure is on the up, it’s safe to say the cloud boom is far from over.
However, hyperscale cloud use is changing. IT leaders are taking a much more conscious approach to their infrastructure choices and hyperscale cloud-only strategies are making way for highly optimized, workload-specific architectures.