A brief history of cloud computing

Before the word 'cloud' existed

Time-sharing on mainframes in the 1960s and 70s let multiple users access a single, expensive computer remotely - the same core idea as cloud, decades before broadband made it practical for everyday business use. Virtualisation in the 1990s, splitting one physical server into many logical ones, is the technology that made modern cloud economically viable.

The 2000s: cloud becomes a product

• Salesforce (1999) proved software could be sold and used entirely over the internet, no installation required
• Amazon Web Services (2006) rented out spare data centre capacity as on-demand infrastructure, creating IaaS as a category
• Google Docs and early Office 365 predecessors brought everyday productivity software online
• Broadband and, later, fibre rollout made these services usable outside large enterprises

The 2010s: cloud becomes the default

Microsoft's shift from boxed Office and on-premise Exchange to Microsoft 365 changed how most UK businesses buy software. Azure and AWS matured from developer tools into full enterprise platforms. Cloud stopped being the exception justified by a business case and became the default that needed a reason not to be chosen.

What has actually changed for UK businesses since

• Full fibre availability removed the bandwidth excuse for not running everything in the cloud
• Cyber Essentials and insurers built their expectations around cloud-native security controls like MFA and conditional access
• AI features (Copilot, cloud-based EDR, cloud voice AI) only work because the underlying data and compute already sit in the cloud
• The PSTN switch-off is forcing the last on-premise holdouts, business telephony, into the cloud on a fixed timetable

The lesson from the history for today's decisions

Every wave of cloud adoption was driven by economics and connectivity catching up with an idea that already existed. The businesses that adopted early and cautiously, testing one workload before moving everything, generally did better than those who moved everything at once or waited until forced. That pattern holds for 2026: pilot AI and cloud features on a defined use case, then scale what works.