Thompson Reuters sees profit boost from AI activity

Thompson Reuters, the information company behind the Reuters newswire service, beat analysts’ profit expectations with its Q4 results as AI services and licensing boost the business. 

Why an information company’s AI operations matter

Large language models need data and lots of it. The problem is that if you build your LLM around the open internet, you end up with a model built on information that is at best sketchy and at worst just uselessly wrong. 

Reuters’ extensive archive of news, as well as its less well-known Westlaw legal database and Checkpoint accounting service is now being licensed by parent company Thompson Reuters. A training data set is being built on rigorously collected, high-quality news for AI firms to use – but at a cost. 

The news offers the perspective of the other side of AI: one in which publishers can extract value from the technology. The big question is whether this is a short-term bump of cash in return for much longer-term threats. Either way, it appears to be the direction of travel. 

There is another angle too: LLMs are developing quickly, with the technology moving less in the direction of one LLM to do it all, but with specific tasks or working areas for developing specialised models.

What’s going on

While the company grew revenues 3% – in line with expectations – its underlying profitability (EBITDA) grew 56%, growth that the company ascribes to licensing its content library to train LLMs.

“We're in growth and investment mode: 2024 is an investment year for us," Steve Hasker, CEO of Thomson Reuters, told the newswire in an interview. 

"We see growth opportunities in 2025, ‘26 and beyond around generative AI, but not exclusively generative AI," he added. 

It confirms reported talks with LLM providers in January. The company already offers up training data as a service on its website. 

In context

• Earlier this week, Microsoft announced extensive partnerships with news organisations and journalism schools, but these have remained relatively vague about the actual work they are doing. 
• Axel Springer and Bloomberg, meanwhile, have moved fast to offer a licensing capability. 
• There are also accusations of subterfuge in the news/AI nexus. The New York Times recently sued OpenAI for copyright infringement, one of a number of lawsuits currently circling around the technology. 

Sourced from Reuters, Investopedia, Bloomberg, Microsoft, New York Times