In May 2025, the Wikidata graph was split into two.
Scientific articles indexed in the database have been isolated into their own dedicated graph, which can be accessed through a specific SPARQL endpoint.
Interestingly, this operation resulted in a nearly symmetrical split, creating two distinct graphs, each with about 8 billion triples. This highlights the significant role that scientific articles play within the Wikidata knowledge base. Their indexing is also supported by initiatives like WikiCite, a Wikimedia sub-project dedicated to tracking sources for information on the world's largest encyclopedia.
While this was the first internal split of the massive Wikidata graph, it may not be the last. The amount of data is growing at a rate of 1 billion triples every year, and although there's intense work being done to optimize and stabilize the infrastructure, further segmentation might be necessary in the future.
