History

Data and the cloud form the invisible infrastructure behind the other courses: AI trains on data and runs in the cloud, blockchain is a distributed database, and cybersecurity defends the whole.

Data came before the computer. In 1890, Herman Hollerith mechanised the US census using punched cards; his company would become IBM. For half a century, processing data meant sorting cards.

In the 1960s, large shared mainframes gave rise to the idea of computing sold as a public utility. As early as 1961, John McCarthy predicted that computing could be sold like water or electricity: the cloud was conceptually born.

In 1970, Edgar Codd published the relational model: organising data in linked tables that can be queried with a declarative language, the future SQL. Oracle sold one of the first relational databases in 1979.

For thirty years, these databases became the silent infrastructure of the world: banks, airline tickets, payroll, inventory. The 1990s brought data warehouses and Business Intelligence: no longer just recording, but analysing to decide.

To index the web, Google published its recipes: GFS in 2003, MapReduce in 2004. The open-source Hadoop project was born in 2006: big data was launched. The same year, Amazon opened S3 then EC2: renting storage and computing on demand, paid for by use.

The phrase "data is the new oil" popularised the idea that data was becoming a strategic resource, even if the image is imperfect: data can be copied, combined and reused, and it draws its value from a context.

Smartphones and social networks unleashed a flood of unstructured data; NoSQL databases complemented the relational model. Docker popularised containers in 2013 and Google open-sourced Kubernetes in 2014. Snowflake, founded by two French engineers, and Databricks took the lead in cloud analytics.

The decade ended with a political turn: Cambridge Analytica and the entry into application of the GDPR made data a legal and democratic issue.

COVID accelerated the move to the cloud. In 2021, France formalised its "Cloud au centre" (cloud-first) doctrine: the state's sensitive data must be hosted on SecNumCloud-certified clouds. Then generative AI reshuffled everything: data became the fuel of models and the cloud reorganised itself around GPUs.

In mid-2026, the public cloud is worth about $850 billion a year and data centres have become a matter of energy geopolitics. A major AWS outage in 2025 was a reminder of the systemic dependence on the cloud.