60% of AWS customers use up credits within the first 12 months.
I’m always tempted by the promise of free, especially when it comes to hosting. The idea of spinning up your servers instantly or skipping hardware setup. At no cost?!
Sign me up.
But free hyperscale cloud credit is more like a gateway drug. Once hyperscale vendors like AWS and Azure hook you in, the free credits vanish, and suddenly you’re locked into an ecosystem with spiralling costs.
These generous starter packages are not gifts - they’re bait. New businesses burn through their hyperscale cloud credits fast (usually within the first 12 months) but not fast enough to avoid dependency on one cloud provider.
9 months in, and with no credit left, hyperscale customers soon realise they’re locked in.
60% of AWS customers use up credits within the first 12 months.
Cloud lock-in isn’t obvious at first. The hyperscale cloud marketing machine is sophisticated, to the extent that the cloud is positioned as straightforward and infinite in resource. Customers buy into cloud models based on free credit, quick setup, and promises of extensive support.
In the first few months, hyperscale seems great. Cloud infrastructure spins up in seconds, AI models run smoothly, and APIs are easy to integrate. But customer cloud dependency is also building.
From talking with businesses worried about this dependency, I’ve seen how some services end up being tied to just one provider’s ecosystem. Products such as relational database as-a-service or firewall solutions can be especially tricky, since moving away from them usually means a lot of re-engineering and internal re-training before they’re back on vendor-neutral ground.
What this means is that their product isn’t just on the cloud; it’s completely wrapped up in it, which makes migration prohibitively expensive. Rethinking architectures, or even migrating specific parts of your system, can bring enormous cost and risk. For many customers switching is no longer viable, and this is how lock-in begins.
Getting locked into hyperscale cloud poses risks for any business. From proprietary costs to the technical hurdles of migrating, cloud dependencies make it difficult for customers to move elsewhere. But by far the biggest risk – as you might have already guessed – is technical entrenchment.
When products are tightly integrated into hyperscale cloud, migrating them to another infrastructure can feel impossible. Tools may not work outside the original environment, and core components like authentication systems or monitoring are often deeply embedded in a single stack.
And even if you could untangle those dependencies, there’s another major challenge to consider: data transfer and the associated costs. As datasets grow into the petabyte scale, moving them can come with steep egress fees and significant costs. Typical migration costs range from $10,000 to $50,000 for small businesses, to $1,000,000+ for large enterprises.
This brings me onto my last, but by no means the smallest, risk for companies locked into hyperscale cloud: operational costs. A staggering 84% of respondents in Flexera’s 2025 report said managing costs is the biggest challenge they face. The promise of cloud as a cheaper, more efficient alternative often falls apart as workloads scale, and its tightly integrated products encourage businesses to consume unnecessary resources rather than streamline them.
Before they know it, businesses can end up locked into proprietary services, racking up data with costs that aren’t sustainable and with no easy way out. The way I see it, the easiest way to stay out of trouble is by knowing how vendor lock-in happens, and how to avoid it, before signing on the dotted line.
Over the years, I’ve seen how easy it is for businesses to fall into the trap of cloud lock-in. But with the right approach, it’s entirely avoidable. Here are three strategies I share with companies not wanting to make the same mistake:
If you need the instant scaling capabilities of the hyperscalers, consider placing baseline resources with bare metal. Consistent, high compute workloads can run on bare metal, and can scale rapidly, but hyperscale cloud is better suited for workloads that scale unpredictably.
For example, compute-heavy tasks such as real time bidding in adtech are suitable for performance optimized bare metal, whereas more dynamic workloads, such as the launch of a new game, can be moved to the cloud - you can never truly know how many players are going to turn up!
Cloud architecture and DevOps consultancies can help to design hybrid infrastructure environments, using tactics to minimize lock-in such as packaging software into containers. They can guide teams towards using open-source solutions and platform-agnostic tools, reducing the risk of being tied to proprietary services.
You can also avoid cloud lock-in and data ties by using open-source APIs or standardized data formats like Parquet or Avro. Aligning with portability standards not only ensures improved interoperability across systems, but also makes sure a business complies with regulations like GDPR.
Being locked in doesn’t mean you’re stuck forever. From diversifying providers to introducing bare metal, you can gradually regain control over your infrastructure. The key is not to rip everything out overnight, but to take a staged and strategic approach to your migration.
Here’s what I typically advise to businesses:
Learn what cloud services you have that are proprietary and difficult to migrate. Pinpoint the components that would be most painful to move, as this is where lock-in lives. Once dependencies are identified, technical teams need to understand how deeply these services are entwined. Determine which components can be rebuilt in a cloud-agnostic way, and start mapping out the steps to reduce cloud reliance.
Identify workload spikes and why (and when) they occur. Only a few types of businesses require intense scalability (e.g. streaming platforms for major events) that require the huge capabilities of hyperscale cloud. In most cases, hyperscale is overkill, and resources can be optimized on bare metal.
Every business has different requirements and therefore needs an infrastructure perfectly suited to its workloads. A high-profile streaming business might better suit hyperscale cloud, while a SaaS company with steady usage patterns, or a gaming studio that needs consistent GPU, would better suit a bare metal alternative.
Even if you are a streaming business or game studio, going all in on hyperscale is rarely the most efficient route to take. Having a foundation of bare metal will take care of the predicted baseline with more tailored and cost optimized hardware. Keeping hyperscale cloud on top as a flexible burst layer ensures you can still scale quickly to meet spikes in demand without overspending.
Cloud migration doesn’t happen overnight. Deeply embedded systems or legacy code can’t be untangled quickly, so expect to phase work over months or years. Incremental changes and strategic planning is how to make progress that sticks.
Choosing tools and architectures that are portable and standards-based gives you leverage when costs rise, or when your strategy shifts.
Hyperscale cloud vendors design their platforms to make onboarding easy and exiting very difficult. This is why I recommend building IT infrastructure with your exit in mind. Even if you don’t intend to leave your cloud provider, your architecture should give you the option to.
It’s also worth being honest about your scaling reality. Hyperscale infrastructure is built for rapid and intensive scaling for global streaming platforms or behemoth retailers. Most businesses don’t need this kind of elasticity and are better suited to the steady compute power that dedicated servers provides.
I’m not saying the cloud is inherently bad; far from it. Hyperscale cloud is ideal for some organizations, and free credits can be used strategically to experiment, rather than anchor your business in proprietary tools.
Just don’t fall for the idea that the only way to scale is to tie yourself to one provider.
With previous roles at Multiplay and PEER 1 Hosting, Isaac’s career has spanned all aspects of the global hosting industry including the development, procurement and selling of bare metal, colocation and cloud infrastructure.
![Reasons why companies are leaving the cloud [with examples]](/dA/845f390047/image/preview-why-companies-are-leaving-cloud.webp?language_id=1)
