---
title: "AI has killed the \"cloud-first\" strategy: why hybrid IT is the only way forward today — Marsanne Studio AI watch"
description: "Given the costs and constraints of AI, Deloitte and a former FedEx architect recommend hybrid infrastructure rather than all-cloud."
url: "https://marsanne.studio/en/ai-watch/ai-has-killed-the-cloud-first-strategy-why-hybrid-it-is-the-only-way-68c512/"
lang: "en"
---

[Marsanne Studio](https://marsanne.studio/en/)

[← Back to home](https://marsanne.studio/en/ai-watch)

[← The AI watch](https://marsanne.studio/en/ai-watch)

Analysis · 6 January 2026

# AI has killed the "cloud-first" strategy: why hybrid IT is the only way forward today

Original title, in French: “L'IA a tué la stratégie "cloud-first" : pourquoi l'informatique hybride est la seule voie à suivre aujourd'hui”

**Publication**

ZDNet France (France)

**Author**

Joe McKendrick

**Published on**

6 January 2026 · read in the page structured data · updated on 2 September 2026

## Summary

Drawing on a Deloitte analysis, the article explains that the rise of AI is challenging the "cloud-first" strategy adopted by many companies over the past decade. Deloitte points to four problems: unpredictable escalation of cloud costs for AI, latency constraints incompatible with certain uses, resilience needs in the event of connection outages, and data sovereignty requirements. The firm recommends a three-tier hybrid approach: cloud for training elasticity, on-premises for production inference with predictable costs, and edge computing for decisions requiring minimal latency. Milankumar Rana, a former software architect at FedEx Services, confirms this trend: he recommends that his clients keep latency-sensitive or regulation-sensitive applications on-premises, while using the cloud for flexibility.

Summary based on the full text of the article.

[Read the article on ZDNet France →](https://www.zdnet.fr/actualites/lia-a-tue-la-strategie-cloud-first-pourquoi-linformatique-hybride-est-la-seule-voie-a-suivre-aujourdhui-487638.htm)

## Figures cited

-   280 reduction factor
    
    fall in the cost of AI tokens over two years
    
    global, two-year period, according to Deloitte
    
    > « Les coûts des jetons d'IA ont été divisés par 280 en deux ans, observent-ils, et pourtant "certaines entreprises voient leurs factures mensuelles s'élever à des dizaines de millions". »
    
    Source: Deloitte
    
-   70 %
    
    cloud cost threshold beyond which on-premises investment becomes more cost-effective
    
    predictable AI workloads, according to Deloitte's analysis
    
    > « Cela peut se produire lorsque les coûts du cloud commencent à dépasser 60 à 70 % du coût total d'acquisition de systèmes équivalents sur site, ce qui rend l'investissement en capital plus intéressant que les dépenses opérationnelles pour les charges de travail prévisibles en matière d'IA. »
    
    Source: Deloitte
    

## Classification

**Country**

Worldwide scope

**Company size**

All sizes

**Industries**

Software vendors, IT services companies, telecoms · Transport, logistics, supply chain · All industries, or industry not specified

**Companies**

Azure (supplier), FedEx Services (deploys AI), Deloitte (study author), AWS (supplier), GCP (supplier)

**Keywords**

cloud infrastructure, hybrid IT, AI costs, on-premises, edge computing, data sovereignty

Entry written by Marsanne Studio from the cited article, with the help of an AI model (claude-sonnet-5). The rights to the article and its illustrations belong to ZDNet France and its author. Entry translated from French by Claude Sonnet 5.5.