Cyc and DBpedia from what I understand are in the 10-100 million+ triples scale and are the largest curated ontologies I've seen. The Billion Triples Challenge this year (2010) has 3.2 billion triples, but a billion triples does not an ontology make. What is the current largest curated ontology?
Depends on your definition of an ontology.
I think that the ncbi/uniprot taxonomy as published on www.uniprot.org/taxonomy/is quite large. It is a curated hierarchical ontology and has about half a million statements.
If you consider, the curated and reviewed section of UniProt, Swiss-Prot as an ontology of protein classes then we are talking about some 180 million statements.
If you take UniParc as an ontology of observed and predicted protein sequences then you go into the billions of triples range.
It's gotta be: Linked Open Numbers (for A-Box at least... ~14bn triples).
Note: on that, see Figure 1 in their paper.
In theory, the chess dataset is bigger, but you could not store all the triples it would theoretically produce if you resolved all the possible URIs!
http://data.totl.net/chess/state/rnbqkbnr_pppppppp_8_8_8_8_PPPPPPPP_RNBQKBNR_w_KQkq_-_0_1