Good post at GigaOM about the current state-of-play in broadband. 281m global subs - so I'd guess probably around 500-600m broadband users if it averages around 2 per house. Add in some more who use broadband at work, school or in public Internet locations and I guess we must get to around 800m regular high-speed Internet users. That also seems to fit pretty well with other numbers like 150m-odd Skype users.
Now, obviously that's still much less than the 2.5bn-odd mobile users (ie equating to around 3bn subs) - around 30% of the figure. But I'd guess in terms of many metrics like purchasing power, propensity to "generate content", sophistication of expertise and so on, that the numbers are much closer to parity. 800m isn't just "early adopters", either - it's early majority. And I strongly suspect it's also higher than the total sum of people who use anything more than voice & SMS on their mobiles.
In other words, pretty much everyone on the planet who's worth advertising at heavily, or who has sufficient disposable income for mobile "content", or is open-minded enough to experiment with new Web 2.0 community-type stuff, is already using PC-based broadband. Which sets expectations in terms of speed, latency and of course, price. As well as disaggregation of access and application. It also means that 800m people aren't that bothered about QoS for many activities.
One last thing - I guess we must be getting very close to 100m WiFi households, then. In other words, the same philosophy will also apply to future femto/in-home cellular deployments - almost every home installing a femto will already have WiFi, and business models & device/service strategies that ignore this will have problems.
[Side note: something I absolutely hate is "spurious accuracy". Given that most operators report figures rounded up/down to the nearest 10k/100k, citing 281,592,583 subs is mathematically unjustifiable. Even 3-sig-fig accuracy of 281m is questionable, given the definitional inconsistency of the data sources]