Shares of Israeli gaming company Buff Technologies jumped about 40% on the Tel Aviv Stock Exchange on Monday after the company said it had signed a pilot agreement with one of the world's five largest artificial intelligence companies to provide data on gamer behavior for AI model training.
Buff did not identify the AI company and declined to reveal its name after the announcement.
Buff Technologies founder and CEO Elay De Beer
(Photo: Buff Technologies)
The pilot agreement is valued at $300,000. Buff said that if the pilot proves successful, the AI company is expected to purchase significantly larger volumes of data. The company also indicated it does not expect the customer to be the only AI firm interested in such information.
Buff, the first gaming company to list on the Tel Aviv Stock Exchange, operates a gaming platform built around a loyalty program with about 20 million registered users. Players earn points for in-game activities that can be redeemed for products in the company's online store. Until now, Buff's revenue model has primarily relied on advertising shown to gamers.
The agreement reflects growing demand among AI companies for large datasets on human behavior to train machine-learning models. Alongside investments in semiconductors and data centers, AI developers are seeking increasingly diverse datasets that capture how people behave in different scenarios.
During its six years of operation, Buff has developed infrastructure to collect, manage and process large-scale data on its users, positioning itself as a potential supplier of training data for AI companies.
Noam Lanir, an Israeli technology entrepreneur and investor who acquired a 7% stake in Buff when it went public, said AI companies are seeking what he described as "distilled wisdom of the crowd" that is less susceptible to the biases and misinformation found on the internet.
"Just as there is a race for chips and a race for data centers, there is also an enormous race for clean data that can train these models," Lanir told ynet. "It's not only about learning physical activity, but also understanding how users act in different situations."
He compared the process to the development of generative AI systems such as ChatGPT, which are trained on vast amounts of written material to generate relevant responses. Buff's gaming data, he said, could similarly help AI companies better understand human behavior in specific environments.
Founded in 2018, Buff remains unprofitable. The company reported revenue of $8.5 million for 2025 and a net loss of $2 million.
Buff went public at a valuation of 90 million shekels (about $27 million at the time). Its market capitalization later fell to about 20 million shekels before Monday's rally lifted it to roughly 33 million shekels.
Founder and CEO Elay De Beer said the company recently realized it had "enormous potential" to sell data to AI companies developing next-generation models.
"AI companies come to us and tell us they need to teach their models specific actions," De Beer said. "Through our platform, they can ask gamers to perform specific tasks in specific games, while we reward the players for completing them."
He said the approach allows AI developers to obtain targeted behavioral data. For example, companies developing automotive-related AI models could analyze data generated by players of racing games, while defense companies could use data from military-themed games to help models better understand patterns of human decision-making in those environments.