AI is a huge hype currently. They stole everything they could online and use it to provide answers to users in well constructed language.
However, training these models take a while. While it will improve in the future the models memory is always outdated (just ask your favorite model for investing advice, often their price reference is completely off). That is what they use initially to filter information. Missing the relevant current information. That is only augmented with a second step; web search.
With AI providers having stolen all public information already data is shielded better in the future. Also, more specific models will be used in the future (rather than a model that is supposed to know everything). These models (and the all knowing ones) need information. Relevant information and preferably curated information.
They will need to pay for it. The power will be with the companies that have proprietary and relevant information. The LLM will be a regular commodity, data will be the gold.
I believe in following themes. This is like an ugly step child of the AI theme but I think this ugly duckling will grow into a swan. As such, I sold exposure to Marvell for instance (and other picks and shovels for AI) and put money in RELX, WKL (this one has especially been punished by AI on the market), Fugro and Eurofins amongst others.
In probably early but data is key. Anyone working with AI knows how much it depends on it, currently it's doing a poor job at it. That's why mostly coding assistants do well, as it's useful for the boring stuff but brings relatively few new things to the table.
Random rumblings. TLDR: Proprietary highly curated and up to date data will be the new gold. Now it's seen as lead, fortunes will turn.