Data study ·
Tokenomics: the figures behind an emerging FinOps discipline
Original title, in French: “Tokenomics : les chiffres d'une discipline FinOps en construction”
- Publication
- Silicon.fr (France)
- Author
- Clément Bohic
- Published on
- 25 September 2026 · read in the page structured data · updated on 25 September 2026
Summary
The Tokenomics Foundation, a sister organisation of the FinOps Foundation, conducted its first community survey, with nearly 500 respondents representing around ten industries, with a median revenue close to $2 billion. The study shows that 96% of organisations use an AI model provider directly, 87% use token-based inference services, and 64% use embedded AI tools. Closed models still dominate, but 70% of respondents plan to use more open models within 12 months. The study also reveals limited confidence in spend visibility and ROI measurement, as well as governance that is mostly entrusted to IT.
Summary based on the full text of the article.
Read the article on Silicon.fr →Figures cited
-
500 companies
number of survey responses
first community survey by the Tokenomics Foundation
« Elle a enregistré un peu moins de 500 réponses. »
Source: Tokenomics Foundation
-
2,000,000,000 USD
median revenue of the companies represented
Tokenomics Foundation survey sample
« Le revenu médian des entreprises qu’ils représentent est proche des 2 milliards $. »
Source: Tokenomics Foundation
-
51 %
share of respondents favouring closed models
survey sample
« Plus de la moitié (51 %) ont largement penché en faveur des modèles fermés (8, 9 ou 10). »
Source: Tokenomics Foundation
-
70 %
share planning to use at least as many open models in 12 months
survey sample
« ils sont environ 70 % à envisager d’utiliser au moins autant les modèles ouverts. »
Source: Tokenomics Foundation
-
96 %
share of organisations dealing directly with a model provider
survey sample
« 96 % en l’occurrence. »
Source: Tokenomics Foundation
-
87 %
share using token-based inference services
survey sample
« Elles sont également nombreuses (87 %) à utiliser les services d’inférence au token »
Source: Tokenomics Foundation
-
31 %
share using local and/or edge AI
survey sample
« On tombe à 31 % pour l'IA locale et/ou en périphérie. »
Source: Tokenomics Foundation
-
31 %
share renting GPUs from cloud providers
survey sample
« Même pourcentage pour la location de GPU chez des clouders. »
Source: Tokenomics Foundation
-
29 %
share owning their own hardware
survey sample
« 29 % ont leur propre matériel, dans leur datacenter ou en colocation. »
Source: Tokenomics Foundation
-
14 %
share very confident in spend visibility
survey sample
« Très confiants 14 % 9 % »
Source: Tokenomics Foundation
-
9 %
share very confident in ROI measurement
survey sample
« Très confiants 14 % 9 % »
Source: Tokenomics Foundation
-
35 %
governance entrusted to IT
survey sample
« La gouvernance, justement, échoit le plus souvent à l'IT (35 % des répondants). »
Source: Tokenomics Foundation
-
38 %
share taking energy considerations into account
survey sample
« Le sondage révèle une certaine prise en compte de l'enjeu énergétique (38 % des répondants). »
Source: Tokenomics Foundation
-
7 %
share spontaneously citing the FOCUS specification
survey sample
« 7 % des répondants ont spontanément mentionné FOCUS. »
Source: Tokenomics Foundation
Classification
- Country
- Worldwide scope
- Company size
- Large companies, All sizes
- Industries
- Software vendors, IT services companies, telecoms · All industries, or industry not specified
- Companies
- DeepSeek (supplier), Amazon Bedrock (supplier), Windsurf (supplier), Cursor (supplier), Vertex AI (supplier), Databricks Genie (supplier), Vercel (cited), OpenAI (supplier), Anthropic (supplier), OpenRouter (supplier), Azure Foundry (supplier), Snowflake CoCo (supplier), Google (supplier), Tokenomics Foundation (study author), LiteLLM (supplier)
- Keywords
- finops, ai cost management, ai governance, open-weight models, inference, tokenomics
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 Silicon.fr and its author. Entry translated from French by Claude Sonnet 5.5.