What Is Cloud Computing? How Modern Internet Infrastructure Works

You probably used cloud computing before you finished breakfast today. Checking your email, scrolling through social media, streaming music on your commute, opening a shared document at work, backing up photos from your phone. All of it ran on somebody else’s computer, in a building you have never seen, in a city you might never visit.

That is cloud computing. And it is the invisible infrastructure underneath nearly everything you do online.

The global cloud computing market crossed the trillion-dollar mark in early 2026. About 94% of enterprises run workloads in the cloud. Public cloud spending alone is on track to exceed $850 billion this year. Three companies, Amazon, Microsoft, and Google, control roughly two-thirds of the entire market. The cloud is no longer an emerging technology. It is the technology that everything else runs on, including AI systems, streaming services, social platforms, and business tools that most people use every day without thinking about where they actually run.

Understanding what cloud computing is, how it works, and why it matters is not a technical speciality. It is basic literacy for anyone who uses the internet.

The Simplest Explanation

Before the cloud, if you wanted to run a website, store data, or run software for your business, you needed your own computer—physical machines, sitting in a room, connected to the internet, maintained by someone you paid. If you needed more power, you bought more machines. If a machine broke, you fixed it or replaced it. If you needed more storage, you bought more hard drives. You owned everything, managed everything, and paid for everything whether you were using it or not.

Cloud computing replaces all of that with a service. Instead of buying and maintaining your own computers, you rent computing power from a company that owns enormous numbers of them. You use what you need, when you need it, and pay based on how much you use. When you need more, you get more instantly. When you need less, you scale back. The computers are housed in massive buildings called data centres, operated by companies such as Amazon Web Services (AWS), Microsoft Azure, and Google Cloud.

The analogy that makes this click for most people is electricity. A hundred years ago, factories generated their own power. Each factory had its own generator, its own fuel supply, its own maintenance team. Then the electrical grid arrived, and factories stopped generating power and started buying it from the grid. They plugged in, used what they needed, and paid the bill. The generator room became unnecessary.

Cloud computing did the same thing to computing infrastructure. Companies stopped building their own server rooms and started buying computing power from the cloud. They plugged in, used what they needed, and paid the bill. The server room became unnecessary for most businesses.

What Actually Happens When You “Use the Cloud”

When you open a Google Doc, here is what physically happens.

Your device sends a request over the internet to one of Google’s data centres. That data centre is a building, sometimes the size of several football fields, filled with thousands of servers (specialised computers designed to handle many requests simultaneously). Your request reaches one of those servers. The server retrieves your document from storage, processes any changes you make, and sends the results back to your screen. All of this happens in milliseconds.

The document does not live on your computer. It lives on Google’s servers. When you share it with a colleague, they access the same document on the same servers. When you open it on your phone, your laptop, or a friend’s computer, you are always connecting to the same file in the same data centre. That is why changes appear instantly everywhere. There is only one copy, and it lives in the cloud.

This same process- your device sending a request to a remote server and getting a response- is what happens when you stream a movie on Netflix, send a message on Slack, use ChatGPT, run a report in Salesforce, or do virtually anything else online that involves an application you did not install from a disc. The application runs on someone else’s servers. Your device is just the screen.

The Three Models of Cloud Computing

Cloud computing comes in three layers, and understanding these layers helps explain why different companies use the cloud in different ways.

Infrastructure as a Service (IaaS) is the most basic layer. The cloud provider gives you raw computing resources: virtual servers, storage, and networking. You decide what operating system to install, what software to run, and how to configure everything. It is the closest thing to having your own server room, except the servers are in someone else’s building and you pay by the hour instead of buying them outright.

IaaS is what startups use when they need servers but cannot afford to build a data centre. It is also what large companies use when they need temporary computing power for a specific project. Need a thousand servers for a week to process a massive dataset? Spin them up, run the job, shut them down, and pay only for the time you used them. That flexibility is impossible with physical hardware you own.

Platform as a Service (PaaS) adds a layer on top. The provider manages the servers, operating system, networking, and runtime environment. You write your application and deploy it. You do not think about the machines underneath. PaaS is what developers use when they want to build and launch an application without managing infrastructure. Google App Engine, Heroku, and AWS Elastic Beanstalk are examples.

Software as a Service (SaaS) is the layer most people interact with daily. The provider runs the entire application. You use it through a web browser or an app. Gmail is SaaS. Google Docs is SaaS. Slack, Salesforce, Dropbox, Zoom, Netflix, Spotify: all SaaS. You do not install anything. You do not manage anything. You open a browser, log in, and the application is there.

Most people use SaaS every day without knowing it. Most developers work at the IaaS or PaaS layer. Most businesses use all three layers simultaneously for different purposes.

Public Cloud, Private Cloud, and Hybrid

There are also different deployment models, which describe who owns the infrastructure and who can access it.

Public cloud is what most people mean when they say “the cloud.” AWS, Azure, and Google Cloud are all public clouds. The infrastructure is owned by the provider and shared across many customers. Your data is kept separate from other customers’ data through software isolation, but the physical hardware is shared. A public cloud is the most common model because it is the cheapest and most flexible. You pay only for what you use, and you can scale instantly.

A private cloud is cloud infrastructure that belongs to a single organisation. It can be physically located in the organisation’s own data centre or hosted by a third party, but the hardware is dedicated to that one customer. Banks, hospitals, and government agencies often use private clouds because regulatory requirements or security policies prevent them from storing certain data on shared infrastructure.

Hybrid cloud combines both. An organisation runs some workloads on its private cloud and others on the public cloud, moving data between them as needed. This is the most common enterprise strategy: 87% of organisations now operate a hybrid cloud setup. Sensitive data stays on private infrastructure. Everything else runs on the public cloud where it is cheaper and more flexible.

The Big Three: AWS, Azure, and Google Cloud

Three companies dominate the cloud computing market so thoroughly that understanding them is essential to understanding the industry.

Amazon Web Services (AWS) holds roughly 30-31% of the global cloud infrastructure market. It was the first major public cloud, launching in 2006, and it has maintained its lead ever since. AWS generated approximately $37.6 billion in revenue in Q1 2026 alone, growing at about 19% year over year. It offers the broadest range of services (over 200) and has the largest global infrastructure footprint. AWS is the default choice for many startups and large enterprises alike, and it runs some of the largest websites and applications in the world, including Netflix and large portions of the US government’s digital infrastructure.

Microsoft Azure holds roughly 23-25% of the market and is the fastest-growing major cloud provider in absolute revenue terms. Azure grew 40% year over year in its most recent quarter, driven significantly by AI workloads. Azure’s strategic advantage is its integration with Microsoft’s enterprise software ecosystem: companies that already use Windows, Office 365, and Active Directory find Azure a natural extension. Its partnership with OpenAI, which runs ChatGPT’s infrastructure on Azure, has been a significant driver of growth. Azure serves 85% of Fortune 500 companies.

Google Cloud holds roughly 11-13% of the market but is growing faster than either competitor in percentage terms: 63% year over year in Q1 2026, its fastest quarter on record. Google Cloud generated $20 billion in revenue in Q1 2026. Its differentiation is AI infrastructure, specifically its Tensor Processing Units (TPUs), which are purpose-built for AI training and inference. Google Cloud also benefits from integration with Google’s consumer services and from Google DeepMind’s model capabilities. It turned its first profit in 2023 and has been expanding margins since.

Together, these three control about 67-68% of all enterprise cloud spending. The remaining third is divided among hundreds of providers, with Alibaba Cloud the largest non-Western player at roughly 4% global share.

Why the Cloud Changed Everything

Before cloud computing, launching a technology company required significant upfront capital. You needed to buy servers, rent data centre space, hire people to manage the hardware, and pay for it all before you had a single customer. If your product took off, you scrambled to buy more equipment. If it failed, you were stuck with expensive hardware you did not need.

The cloud eliminated that barrier. A startup can now launch a product on AWS, pay a few hundred dollars a month, and scale to millions of users without ever buying a single server. If the product fails, the cost stops the moment it shuts down. If it succeeds, they scale up by clicking a button, not by ordering hardware and waiting weeks for it to arrive.

This is why the cloud fundamentally changed the economics of the technology industry. The capital required to start a tech company dropped dramatically. The speed at which companies could scale increased dramatically. The infrastructure that used to take months to provision now takes minutes.

The same dynamic applies to established companies. A bank that needs to process a year’s worth of transaction data for a regulatory report can spin up hundreds of servers for a few hours, run the analysis, and shut everything down. A movie studio can render visual effects using thousands of cloud GPUs for a few weeks and then release the capacity back. A retailer can triple its server capacity for Black Friday and scale it back down the following Monday.

That elasticity, the ability to expand and contract computing resources on demand, is the single most important thing cloud computing provides. It is the reason 94% of enterprises use it, and why the market is approaching a trillion dollars.

The Cloud and AI

The AI boom is now the single largest driver of cloud infrastructure growth. AI-related cloud spending makes up approximately 19% of total cloud spending, up from 8% just three years ago.

Training a large language model requires thousands of GPUs running simultaneously for weeks or months. No company that is not already a hyperscaler can afford to build that infrastructure from scratch. The cloud makes it possible for any organisation to access that computing power on demand. You pay for the GPUs you use, for the time you use them, and then you release them.

This is why every major AI lab runs on cloud infrastructure. OpenAI runs on Azure. Anthropic uses AWS and Google Cloud. Google’s AI runs on Google Cloud. DeepSeek, Qwen, and Mistral all train and serve models through cloud providers.

The relationship between AI and the cloud is symbiotic. AI needs the cloud because training and running large models requires enormous, elastic computing resources. The cloud needs AI because AI workloads are driving revenue growth faster than any other category. Google Cloud’s 63% growth rate is largely attributable to demand for AI. Azure’s 40% growth is significantly driven by its partnership with OpenAI. AWS is investing heavily in custom AI chips (Trainium and Inferentia) to capture more of this demand.

The rise of AI has also created a new category of cloud service: model APIs. Instead of renting raw computing power and training your own model, you can access a pre-trained model through an API and pay per use. OpenAI’s API, Anthropic’s API, and Google’s Vertex AI all work this way. You send a request, the model processes it on the provider’s cloud infrastructure, and you pay per token. This is effectively SaaS for intelligence: the model runs on someone else’s servers, and you use it through an interface without managing anything.

The Costs Nobody Talks About

Cloud computing solved the problem of upfront capital costs. It created a different problem: ongoing operational costs that can grow faster than anyone expected.

Because cloud pricing is based on usage, costs scale with your workload. A company that grows quickly can find its cloud bill growing just as quickly, sometimes faster than its revenue. Cloud waste, paying for resources you provisioned but are not actually using, is a pervasive problem. Estimates suggest that 30-35% of cloud spending is wasted on idle or over-provisioned resources.

The average enterprise now uses nearly five different cloud providers simultaneously. Managing costs across all of them requires specialised tools and practices, a discipline called FinOps (financial operations for cloud). The FinOps market exists entirely because cloud bills became complex enough to require their own category of management software.

There is also the question of vendor lock-in. Moving from one cloud provider to another is theoretically possible but practically difficult. Applications built using a specific provider’s services, databases, and tools do not transfer easily to another provider. The switching cost can be high enough that companies effectively become dependent on a single provider, which gives that provider significant pricing leverage.

The cloud eliminated the capital cost of infrastructure. It replaced it with an operational cost that never stops, that scales with usage, and that requires active management to keep under control. For most organisations, the trade-off is worth it. But it is a trade-off, not a free upgrade.

What the Cloud Actually Looks Like

A modern cloud data centre is one of the most impressive physical structures in the technology industry, and most people have never seen one.

A large data centre can occupy hundreds of thousands of square feet. Inside, rows of server racks stretch the length of the building, connected by miles of fibre-optic cable. The servers generate enormous amounts of heat, and cooling systems, often using water, ambient air, or specialised liquid cooling, run continuously to keep temperatures within the operating range.

Power consumption is immense. A single large data centre can consume as much electricity as a small city. The total power consumption of all data centres worldwide is estimated at over 1% of global electricity use. That figure is rising as AI workloads drive demand for more computing power.

As of mid-2026, more than 2,200 new data centres are in planning or construction stages across the United States alone. Google, Microsoft, Amazon, and Meta have collectively spent more than a trillion dollars on infrastructure expansion since 2023. The physical footprint of the cloud is growing as fast as the digital demand it serves.

These facilities are also targets. Data centres concentrate enormous amounts of computing power, data, and economic value in single physical locations. Securing them against physical threats, cyberattacks, natural disasters, and power failures is a significant engineering and security challenge. Redundancy is built into every layer: multiple power feeds, backup generators, redundant network connections, and data replicated across multiple geographic regions so that the failure of any single facility does not take services offline.

Why This Matters

Cloud computing is not a product you buy. It is the way modern internet infrastructure works. Almost everything you use online, from the simplest web page to the most sophisticated AI system, runs on cloud infrastructure operated by a small number of very large companies.

Understanding that changes how you think about the technology you use every day. When ChatGPT gives you an answer, that answer was generated on Microsoft’s Azure servers. When you watch a movie on Netflix, the video streams from AWS. When you search on Google, the results come from Google’s own cloud infrastructure. When your bank processes a transaction, there is a very high probability it runs through one of these three providers.

The cloud made it possible for a startup with no hardware to serve millions of users. It made it possible for AI labs to train models that would otherwise require them to build their own supercomputers. It made it possible for businesses to operate globally without infrastructure in every country. It made it possible for you to access your files, your email, and your applications from any device, anywhere, at any time.

It also concentrated an extraordinary amount of the world’s computing infrastructure in the hands of three companies. Whether that concentration is a strength (reliability, economies of scale, continuous investment) or a vulnerability (single points of failure, pricing leverage, data governance concerns) is one of the defining questions of modern technology infrastructure.

The cloud is not somewhere you go. It is the place everything already is.