The W3C’s ESW wiki has a page (Large RDF triple stores) with

“references to quotes of deployment of large triples stores rather than predictions of what some software might scale to.”

Mentioned are reports of systems using MySQL, Kowari, Postgress and other systems including some applications with over 200M triples. (Spotted on the SWIG scratchpad)

Related posts:

1. Stress test your RDF triple store
2. How YouTube scales MySQL for its large databases
3. An intelligent geospatial reasoning system
4. The nine billion names of Google
5. NOSQL: distributed key-value data stores