Analysis ·
Twelve months of AI budget used up in four: the hidden cost of scaling up
Original title, in French: “Douze mois de budget IA engloutis en quatre, le coût caché du passage à l’échelle”
- Publication
- Alliancy (France)
- Author
- Fiona Slous
- Published on
- 5 October 2026 · read in the page structured data
Summary
In a talk at the Salon de la Data et de l'IA in Nantes, Arnaud De Temmerman, CTO and co-founder of Treez, presented AI FinOps as a new maturity test for generative AI deployments. He cites the case of Uber, which used up in four months the AI budget planned for twelve after rolling out Claude Code to around 5,000 engineers, as agentic AI multiplied calls to the models. He also describes a project to migrate 13,000 Google Data Studio reports to Google Looker for a major retail player, where a model generated a deterministic program rather than processing each report individually. He puts the monthly cost of one million conversations at between 15,000 and 75,000 dollars with frontier models, against 150 to 800 dollars with small specialised models, and recommends governing spending in relation to the value produced.
Summary based on the full text of the article.
Read the article on Alliancy →Figures cited
-
5,000 engineers
Number of Uber engineers using Claude Code at the time of the rollout cited
Uber, spring 2026
« la diffusion de Claude Code auprès de quelque 5 000 ingénieurs a pris de vitesse les prévisions financières du groupe »
-
4 months
Time taken to use up Uber's planned annual AI budget
Uber, spring 2026, budget originally planned for 12 months
« Uber avait prévu douze mois de budget IA. Quatre ont suffi pour le consumer. »
-
13,000 reports
Number of Google Data Studio reports to be converted to Google Looker in the migration project for a major retail player
Major retail player, three-month project
« Le chantier portait sur 13 000 rapports Google Data Studio à convertir vers Google Looker en trois mois. »
-
75,000 USD
Estimated monthly cost of one million conversations with frontier models
For one million conversations per month, estimate by Treez's CTO
« Pour un million de conversations mensuelles, le cofondateur de Treez a situé la facture entre 15 000 et 75 000 dollars avec des modèles frontières, contre 150 à 800 dollars avec de petits modèles spécialisés. »
-
800 USD
Estimated monthly cost of one million conversations with small specialised models
For one million conversations per month, estimate by Treez's CTO
« Pour un million de conversations mensuelles, le cofondateur de Treez a situé la facture entre 15 000 et 75 000 dollars avec des modèles frontières, contre 150 à 800 dollars avec de petits modèles spécialisés. »
Classification
- Country
- France
- Company size
- Large companies, All sizes
- Industries
- Software vendors, IT services companies, telecoms · Retail and distribution, e-commerce · All industries, or industry not specified
- Companies
- Treez (cited), Uber (deploys AI)
- Keywords
- FinOps, generative AI, AI costs, agentic AI, scaling up, open-weight models
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 Alliancy and its author. Entry translated from French by Claude Sonnet 5.5.