Cloud providers, hosting and data centers are different layers of the infrastructure market, and you should choose between them along two axes: which service model you need and in which jurisdiction the data must reside.

Axis 1. Service model: what exactly you buy

Cloud (IaaS/PaaS). Virtual resources and managed services on a consumption basis. This includes international and Russian cloud platforms, including Yandex Cloud. You don’t manage the hardware. It’s a model for variable load, a fast start and ready-made services.

Hosting / VPS (for example, Timeweb). Essentially the same rental logic, closer to the end user: websites, small applications, bots. A low barrier to entry and minimal infrastructure worries.

Colocation / dedicated in a data center (the Akashi model). Physical placement of your own equipment in a data center: space, power, cooling, connectivity, fault tolerance, physical access and audit. A model for constant load, full control and your own hardware.

The key idea: these layers don’t exclude each other, they complement each other. A cloud provider physically places its equipment somewhere — often precisely in a data center. “The cloud” can perfectly well run “on top of” colocation.

Axis 2. Jurisdiction: where the data physically resides

For Kazakhstani business this is often the decisive factor. Kazakhstani law requires storing the personal data of citizens on servers physically located in Kazakhstan.

  • Russian clouds and hosting (Timeweb, Yandex Cloud and others) operate out of the Russian jurisdiction — this creates limitations and compliance risks when working with the data of Kazakhstani citizens.
  • A local Kazakhstani data center provides guaranteed localization: the data is physically in Kazakhstan, and the jurisdictional barriers fall away.

A detailed breakdown of the requirements and fines is in the guide on data localization in Kazakhstan.

How to choose a layer for your job

It helps to ask yourself three questions:

  1. What is my load? Variable and small → cloud/hosting. Constant 24/7 and/or your own hardware → colocation.
  2. Who am I in the chain? End user of a service → cloud/hosting. Infrastructure owner or service provider → data center.
  3. Where must the data reside? If it’s the personal data of Kazakhstani citizens, you need a facility physically located in Kazakhstan.

If the choice has narrowed to colocation, the next step is the facility’s fault-tolerance tier: see the breakdown of the difference between Tier III and Tier IV. And to decide between renting a cloud and hosting in a local data center, there’s a separate guide: Timeweb, Yandex Cloud or a data center in Kazakhstan.

Where Akashi sits on this map

Akashi occupies a layer that is still poorly covered on the market — a physical data center in Astana with local Kazakhstani jurisdiction and a colocation/dedicated model. Its positioning is Tier IV level, an under-construction facility, the first of its kind in Central Asia.

This is a choice not for those launching a bot in the cloud, but for those who need to physically host and control their own infrastructure locally in Kazakhstan — including the providers of cloud and SaaS services themselves.

Frequently asked questions

How is the infrastructure market in Kazakhstan structured?

It’s several layers: clouds and hosting (renting virtual resources) and data centers (physically placing equipment). They solve different problems; the choice depends on your load and your data-jurisdiction requirements.

Are Yandex Cloud and Timeweb competitors to a data center?

More often they’re adjacent layers. Cloud and hosting rent you resources; a data center physically houses equipment — including the equipment of the cloud providers themselves.

Which criterion should I choose by first?

By the job and the data jurisdiction. If you work with the personal data of Kazakhstani citizens, start with the localization requirement — it narrows the choice to facilities in Kazakhstan.


Contact the Akashi team — we’ll help you identify the right infrastructure layer for your industry and workload.