QGIS is really coming on and has almost reached a point where I think I could live completely without ArcGIS (if I add a bit of scripting in GDAL/OGR plus PostGIS etc). However, ArcGIS cannot be considered to be a single package. It's more of a sliding scale as you move up through the ranks of Arcview/ArcEditor/ArcInfo. Then there are the additional extension modules. It all depends on what you can afford or what your organisation has decided to buy in comparison to what you need. So, there cannot possibly be a single answer to your question. In VERY CRUDE terms (and ignoring specialist combinations of ArcGIS extensions like Spatial Analyst, 3D Analysts, Military Analyst etc.) I'd suggest a scale something like this:
ArcView -> ArcEditor -> QGIS -> ArcInfo
But even that doesn't really give you a true comparison because, for instance, if your ArcGIS installation does not have Spatial Analyst Extension, and you want to do raster analysis, then obviously QGIS has to be the preference.
As soon as you go beyond staightforward GIS, if you throw some money at ESRI you will probably be able to make most versions of ArcGIS rival or surpass QGIS... but if you are comfortable with the FOSS world you can probably extend QGIS with a plugin, use some other FOSS solution or roll your own in C#, Python, or whatever. At this point ArcGIS wins on convenience even if not on functionality.
As for speed, ESRI claim great speed increases in v10. I've not noticed anything that's got me excited so far and I've been using v10 since it came out and have upgraded to 10.1. Empirically, I find GDAL/OGR functions written in Python faster and since QGIS is built partially on this functionality you get convenience and speed.
Lastly, if you are hiring and want "plug-and-play" GIS professionals, you'll get a better response to your job advert if your organisation uses ArcGIS than if you use QGIS.
"You pays yer money and takes yer choice" :)