I think the value it brings to Google is several-fold. I'm probably missing a few, but here's what comes to mind.
1- Brand: Google seems very data scientist friendly. Data scientists will want to use Google technologies because Google is the 'data science' company.
2- Information on trends: Are data scientists suddenly spending a lot of time working in some language, or abandoning another? Are the Kaggle forums overflowing with a lot of Tensorflow related ML problems? Google could use this information to know what decisions to make (bonus: the people who'd curate and analyze the information are probably data scientists).
3- Promote/amplify: Google can hawk the crap out of Google related data science products to that market. And Google can probably provide GCP credits to registered Kaggle users, so that they can spin up their own little server, run their models, and get accustomed to using GCP in the process.
Those are just some thoughts. They likely have even better plans!