Skip to main content

The problem the cloud solves

Summary: before 2006, launching an online service meant buying physical servers, guessing the load three years out, tying up hundreds of thousands of euros, and waiting weeks before being able to write the first useful line of code. This lesson details the six structural problems of this model and why they blocked innovation.

1. The world before — setting up an infrastructure in 2004​

The decisive point: four months and half a million euros before serving the first user. For a young company, this barrier was often insurmountable — hence the role of investment funds, indispensable if only to buy hardware.

2. The six structural problems of the owned model​

Problem 4 is the most counter-intuitive and the most expensive. A classic industry study placed the average utilization rate of enterprise servers between 6 and 15%. In other words: out of 100 euros of purchased hardware, 85 to 94 euros of capacity never served anyone.

3. The sizing problem, illustrated​

This is the founding economic argument of the cloud, and it is worth understanding well.

With owned hardware, there is no good answer — only the choice between wasting and going down. It is precisely this dilemma that the cloud's elasticity removes.

4. A realistic scenario — the success that kills​

Here is a classic sequence of events from the pre-cloud era, which explains well why the model changed.

With the cloud, this story does not happen. An auto-scaling rule would have taken the capacity from 2 to 30 instances in a few minutes, absorbed the peak, then scaled back down automatically. Cost of the peak: a few tens of euros.

5. What the cloud concretely changes​

The most underestimated benefit is the sixth one. A team of three people can today use a database replicated across three continents, a worldwide content delivery network and machine learning models — capabilities that twenty years ago required an organization of several hundred engineers.

6. What the cloud has NOT solved​

For intellectual honesty, we must also name the problems the cloud has created or moved.

The cloud is not a universal solution. It is a trade-off: you exchange cost predictability and autonomy for elasticity and speed. For most organizations, the exchange is favourable — but there remain cases where it is not, detailed in lesson 6.

7. When the cloud remains less relevant​

A phenomenon to know: cloud repatriation. Since 2022, several well-known companies have partially left the public cloud after realizing that their load, having become stable and predictable, cost less on owned hardware. It is not a rejection of the cloud — it is a sign of maturity in the trade-off.

Remember in 30 seconds​

  • Before the cloud: four months and hundreds of thousands of euros before the first user was served.
  • Six structural problems: guessing the future, tied-up capital, slow experimentation, wasted capacity, the geographical wall, the plumbing on your shoulders.
  • The average utilization rate of enterprise servers was between 6 and 15%.
  • Without elasticity, the choice is wasting or going down — no good option.
  • The cloud brings: elasticity, expense instead of investment, speed, worldwide reach, delegation, advanced services.
  • But it also creates: an unpredictable bill, moved complexity, provider dependence, the sovereignty question.
  • Repatriation is a real phenomenon since 2022, a sign of maturity in the trade-off.

Next: What is the cloud really? NIST definition and history →