Upload a photograph to the cloud and it seems to leave the physical world. In reality, it becomes data stored and processed on computing equipment somewhere. The convenience comes from not having to own or manage that equipment yourself, not from the equipment disappearing.
Cloud services depend on data centres: buildings containing servers, storage and networking hardware, supported by power, cooling and security systems. Your screen hides this machinery behind an account and an internet connection.
A service model built on hardware
NIST defines cloud computing through characteristics such as on-demand access, shared resource pools and the ability to scale resources. Those characteristics describe how computing is provided, rather than a special substance from which it is made.
A physical server can support multiple virtual computing environments. Software allocates processor time, memory, storage and network resources to different workloads. This helps providers offer capacity flexibly without assigning every customer a personally owned machine.
The underlying hardware still has limits. Processors consume electricity, storage devices can fail and network links have finite capacity. Cloud software manages these realities; it does not abolish them.
Inside the building
Server racks connect through networking equipment to other systems and the wider internet. Electrical infrastructure supplies power, while backup arrangements help handle interruptions. Cooling removes the heat generated by operating equipment.
Physical security controls access to the premises and sensitive areas. Maintenance teams replace components and manage facilities. Behind a simple file upload is a continuing industrial operation, not just a website running unattended in an empty room.
Data centres differ in design and scale. Some serve one organisation, some host equipment belonging to others and some form part of large public-cloud platforms. The word describes infrastructure that can support several business models.
Your data may have more than one address
Providers organise infrastructure geographically. AWS, for example, describes Regions and Availability Zones, with zones designed to provide separated infrastructure within a region. Other providers use their own arrangements and terminology.
A service may replicate data across multiple devices or locations to improve availability and durability. But replication is a configuration and service-design choice, not a guarantee that every file exists everywhere. Region selection, service terms and application settings affect where data is handled.
A nearby copy of content can reduce delay for users, while broader distribution can support recovery from certain failures. These goals can interact with cost, privacy requirements and rules about data location.
Why the location still matters
Distance affects latency: information takes time to travel. Local power and water conditions can influence facility operations and environmental impacts. Laws and contractual commitments can affect how data is stored and accessed.
A cloud outage is therefore not evidence that something immaterial has malfunctioned. It may involve software, networking, a facility problem or a dependency shared by many services. Resilience requires appropriate design across those layers.
The cloud is best understood as access to computing through a managed service. Physically, it is racks, chips, cables and buildings in particular places. The remarkable achievement is making that complicated infrastructure feel like a folder you can open from almost anywhere.