Most large firms plan higher open-weight AI use to cap bills, study shows


FILE PHOTO: The words "AI Artificial Intelligence", a keyboard, and a robotics hand in this illustration taken October 19, 2026. REUTERS/Dado Ruvic/Illustration/File Photo

Nov 12 (Reuters) - ⁠Nearly 6 in 10 senior executives at large companies ⁠plan to increase their use of cheaper open-weight AI ‌models over the next year as AI bills overshoot budgets, a Capgemini study showed on Thursday.

This trend echoes SAP, whose finance chieflast month said it was ​replacing some of its costliest AI models ⁠with cheaper ones where performance ⁠held up. Anthropic, which launched itsIPO roadshow this week, also warned ⁠in ‌its prospectus that customers may shift to cheaper options, including open-weight models.

Open-weight models, like those from DeepSeek or ⁠Mistral, can be downloaded and runanywhere, while closed ​options, like Claude, ‌ChatGPT or Gemini models, are only accessible from the cloud.

Some ⁠59% of ​the surveyed executives said AI spending over the past yearhad been significantly above theirexpectations or budgets, by 24% on average, while 53% said ⁠they were worried about AI costs — last ​year, 45% of executives hadexpressed such concerns.

Open-weight models handled 72% of traffic onOpenRouter, a popular multi-AI provider, in September, though that represented ⁠just 27% of AI spending, Capgemini said.

The IT consulting firm said open-weight tokens were on average 87% cheaper than closed ones. Closed models broadly remain in the mix, with half therespondents saying ​they would retain closed models where the ⁠complexity of the task warranted the additional cost.

Capgemini, which is helping ​companies embed AI into their existing ‌platforms, surveyed 500 executives at enterprises ​with annual revenue of more than $1 billion in 10 countries.

(Reporting by Leo Marchandon in Gdansk, editing by Milla Nissi-Prussak)

Follow us on our official WhatsApp channel for breaking news alerts and key updates!

Air Pollutant Index

Note: API is calculated based on 24hr running average of the most dominant parameter.

Highest API Readings

    Select State and Location to view the latest API reading

    Others Also Read