You build an ad network that 'performs arbitrage on competing ad networks'. Opera makes money on the margin that potential advertisers save. This works because Opera controls the pipeline from web to phone.

I might be particularly dense today, but I don't get the link between controlling the pipeline and building an arbitrage ad network. Why can't you do one without the other?

Edit: to spell out my confusion. If you want to arbitrage ads, content providers would (presumably) need to include your ads in their content one way or another, so I don't see what advantage controlling the pipeline gives you. If you just want to inject ads into the content you're serving simply because you control the pipeline, then you can just as well inject regular ads, I don't see where the arbitrage part comes in.

I think a key point of arbitrage is that you do both of the transactions (the buying and the selling) at the same time.

This insulates the person doing the arbitrage from losing money because one or both of the markets changed (it is 'theoretically risk free').

So they would be theoretically be trading page impressions as they occur, to the highest bidder. Compare this to AdWords, which offers 'some amount of Cents' per impression.

You sell your adspace to opera, opera sells that AdSpace to someone else, dynamically based on market prices. (They could even sell it Google or whoever). Its like an ad stock market.

At least, that is my (very) speculative understanding.

edit: The key point is that they control the cache, so if you sell them ads and your page goes through the cache, they can insert it (whoever's ad it is) quickly. Otherwise it would be an extra level of indirection loading the ad.

I haven't thought about this extensively, but quickly considering this, I'd imagine one would be able to target users with ad networks much as you can do that with ads today. Given a number of networks, display ads from the network most likely to get a click-through from the user under consideration.