Artificial intelligence is costing Wikipedia dearly. Literally. Hundreds of thousands of dollars a month. The AI takes your content for free by consuming your bandwidth, that is, electricity and server resources, which cost money. So AI companies are going to start paying.
The Wikipedia Foundation has announced an agreement with Amazon, Meta, Microsoft, Perplexity, and the French Mistral AI. These companies will pay an undisclosed amount to use Wikipedia data to train their language models. In exchange, the popular digital encyclopedia will offer them already ordered so that AI can take advantage of them.
This agreement breaks with the funding philosophy that Wikipedia has used for 25 yearsbased on individual donations. But he has no choice but to bring order to the chaos, because the AI was bringing his servers to their knees.
AI is going to pay Wikipedia to use its bandwidth
Created in 2001, Wikipedia is celebrating its 25th anniversary these days. It is one of the largest and most important projects on the Internet: an encyclopedia created by users, completely free. It already contains 65 million articles in 200 languages, maintained by more than 250,000 volunteers.
The project is non-profit, it is maintained by the Wikipedia Foundation, which lives off donations, the vast majority from individual users who contribute what they can. But AI is putting this financing model at risk.
Wikipedia is one of the most visited websites in the world, and to keep so many millions of articles accessible at a reasonable speed, it has servers all over the world. These servers cost money, both to purchase and to purchase, electricity and maintenance, and artificial intelligence is adding a lot of pressure.
Every day hundreds of robots extract content from Wikipediato train different AIs. They literally copy the entire Wikipedia, congesting your network bandwidth. In 2025, non-human traffic on Wikipedia increased by up to 50%, and the proportion of human visitors fell by 8%. As it is continually updated, data mining to train the AI is continuous.
Wikipedia is a free service created to spread verified knowledge (all its content is based on verifiable sources, otherwise moderators delete it), so they don’t want to prohibit AI from using it either. That is why they have made the decision to make you pay, but not for copying your data, but to cover the extra expense on electricity and resources.
”I am glad that AI models are trained with Wikipedia, since it is a source curated by humans,” acknowledges its founder, Jimmy Wales, in an interview in AP. “AI companies should contribute and pay the part of the cost they are imposing on us.”
This is a double-edged sword, because many people might think that, if AI pays… Why donate? But surely, it’s the only thing they can do.
It is almost impossible to prevent robots from continuing to extract content, as it is a project open to everyone. AND If AI uses Wikipedia to train it will benefit everyonebecause it is data curated by humans where errors have been eliminated, and it is verifiable information with documented sources.
After 25 years, Wikipedia changes its funding model. Nothing is immutable with AI.