Data Centres, Compute & Cloud

Built to calculate artillery firing tables, it occupied a large room and needed a team to reprogram by rewiring, but it established that electronic computation at scale was possible, the starting point from which later compute infrastructure descended.

Time-sharing replaced the practice of feeding a computer one job at a time, showing that computing power could be divided among many simultaneous users, the conceptual root of remote, on-demand computing services.

Virtualisation let a single mainframe be divided into many isolated environments, a concept that lay dormant for decades before becoming the technical foundation that made multi-tenant cloud computing commercially practical.

Seymour Cray's design, notable for its speed and distinctive circular shape, established a lineage of dedicated high-performance machines built for scientific and, later, weather and weapons simulation that modern AI training clusters descend from.

General Magic used the cloud metaphor for the universe of 'places' that mobile agents in its Telescript environment could go. The usage is credited to communications specialist David Hoffman, drawing on long-standing networking and telecom convention.

Compaq's plan for future computing and the Internet popularised the expression 'cloud computing'. It foresaw commercial success for online consumer file storage, prompting Compaq to sell server hardware to internet service providers.

Its subscription model, run from the provider's own servers rather than installed on a customer's machines, established the software-as-a-service pattern that later became a standard way of delivering cloud-hosted applications and compute.

Built to model climate and seismic activity, it held the top ranking for over two years, the last time a non-American machine led global supercomputing until Chinese systems took the lead more than a decade later.

AWS was set up to allow developers to build applications independently, marking the practical emergence of cloud computing in the 2000s. It became the foundation of the commercial public cloud market.

The papers described how to split enormous computing jobs across thousands of ordinary machines, an architecture that the wider industry, including the open-source Hadoop project, copied and that underpins hyperscale cloud computing to this day.

Renting server capacity and storage by the hour let startups and later large enterprises avoid buying their own data centres, a shift in the economics of computing that Microsoft, Google and, later, Chinese providers all followed.

NASA produced the first open-source stack for deploying private and hybrid cloud environments. This opened cloud deployment beyond proprietary public providers.

Azure gave Microsoft a full public cloud platform, establishing a major competitor to AWS. Its launch was part of a wave of cloud service introductions in the 2010s.

OpenStack provided a shared open-source software base for building private and public clouds. It gave organisations and states an alternative to proprietary hyperscaler platforms.

NIST identified five essential characteristics of cloud systems and recognised IaaS, PaaS and SaaS as the core service models. This became the reference framework for procurement and regulation of cloud services.

IBM entered the cloud market with its SmartCloud framework. It was part of the broad shift of incumbent IT vendors into cloud provision.

Oracle announced its own cloud platform, extending the competitive field of major providers. It continued the migration of established enterprise software vendors into hosted infrastructure.

The agreement, reached during a presidential visit to the Gulf, cleared the way for the Emirati firm G42 to import substantial volumes of advanced Nvidia processors for AI data centres, part of a wider set of Gulf compute agreements including Saudi Arabia.

Microsoft, Google and regional cloud providers expanded data centre and compute capacity in Johor, building on earlier commitments and positioning Malaysia as a lower-cost alternative to Singapore for cloud and AI workloads serving Southeast Asia.

Chinese cloud providers deployed clusters built on thousands of Ascend processors to train and serve large models domestically, narrowing though not closing the gap with clusters built on restricted Nvidia hardware, and reducing reliance on smuggled or grey-market chips.

The project, developed with domestic technology groups, aimed to give Korean researchers and companies access to large-scale GPU capacity at home, echoing sovereign compute programmes announced by Japan, India and Gulf states over roughly the same period.

The trade ministry extended grants to domestic cloud providers including Sakura Internet to expand graphics processor capacity inside Japan, part of an economic security strategy intended to reduce dependence on foreign hyperscalers for AI training and inference.

The pan-African group, working with Nvidia, opened AI compute capacity intended to let African researchers and businesses train and run models locally rather than renting capacity from cloud providers based in the United States, Europe or the Gulf.

Additional Chinese fabrication plants and toolmakers were added to restricted lists, extending earlier controls on advanced processors to the equipment and specialised memory needed to manufacture them, an escalation contested by allied toolmakers in Japan and the Netherlands.

Investigations and export-control enforcement actions traced shipments of restricted Nvidia processors reaching China via Singapore, Malaysia and other transit points, prompting tighter reporting requirements for regional buyers even though most local demand was genuine.

The fund, one of several vehicles channelling state capital into semiconductors, backed equipment makers and chip designers as Beijing sought compute capacity that did not depend on foreign suppliers subject to export licensing.

Firms including Yotta and E2E Networks were selected to supply graphics processors at government-negotiated rates, building a shared national compute pool intended to give startups and researchers access that they individually could not afford.

ISO refined the list of essential cloud characteristics and added service categories including Network, Communications, Compute and Data Storage as a Service. The standard also classified serverless computing alongside IaaS, PaaS and SaaS.

The law committed tens of billions of dollars in subsidies and tax credits to rebuild domestic semiconductor manufacturing and research capacity, reflecting a bipartisan judgement that compute capacity was a strategic asset rather than an ordinary traded good.

Outposts extended AWS infrastructure, services, APIs and tools into customer data centres, co-location spaces and on-premises facilities. It marked the hybrid convergence of public cloud and private infrastructure.

The machine topped the global TOP500 ranking using Chinese-designed chips rather than American processors, demonstrating that indigenous high-performance computing was achievable and foreshadowing the compute self-sufficiency drive that followed later export controls.

Hackers stole over seven million Dropbox user passwords, seeking payment in Bitcoin. Together with the 2014 iCloud leak, it highlighted the security exposure inherent in consumer cloud storage.

A Commission Communication set out several issues holding back development of the European cloud market. It marked early supranational policy attention to cloud capability.