I was once asked: "What do you think the personal computer of 2020 will be like?" I was standing in front of an audience of experts, who had paid to listen to me, and who seriously expected (it seemed) that I would have something to say that would never have occurred to them.

To summarise my response, I said that something that is described as five years in the future means that two or three people are trying to get a prototype built; 10 years' time means that it's a theoretical possibility, and 15 years from now is basically saying: 'I've had an idea, and I know it can't be done, yet, but wouldn't it be nice/awful if ...?'

So when I talk about what will happen in 2020, the one thing I can be sure of is that I'll be wrong. If it's feasible, I won't know about it; if I know about it, it will turn out to be unfeasible.

But what I do suspect is that the battle will be about where the data is. Most of the data in the world is not on my personal computer. If I want some of it, I ask Lord Google for permission to access it.

Down comes a choice, on a platter: I help myself to what I need, knowing that I can have more if I want it.

Let's assume that, in a decade, anybody with a network connection to the internet that runs slower than 10Mbps will be regarded as a lightweight, or a pauper. Most homes will have 100Mbps (two high definition TVs and some extra connectivity) and the leading edge will have Gigabit links to the world.

So, how on earth do you distinguish between the data on your personal computer, and the data Google holds in trust for you?

My own data will be instantly accessible, whether I've left it on my hard disk at home, put it on my portable machine, or just retrieved it from Lesotho. The actual data doesn't mean a thing: the trick is to integrate it.

For example, suppose I have data that describes the streets near where I am, data relating to my own tastes and data relating to local businesses and bus timetables. Is that data? No.

The data is that I like Indian food, and that there's a restaurant just around the corner that meets my known preferences and is reachable in five minutes. The data that is my own says: 'If you get on this bus, you can be eating an affordable but excellent curry in five minutes.'

All the rest is 'out there' somewhere. Exactly where it is stored is academic. Or is it? Suppose government and commercial snooping has reached the level where it's worth spending serious effort keeping it out of the hands of higher powers? Suppose I want to access this data on my own storage device, where nobody else can see it?

In that case, the question of where that storage device is becomes vital. I might want instant access to it, which is OK, because my own storage device can be reached from anywhere, but I don't want this integrated, analysed data to be available to anybody else.

It arrives on my personal communicator in highly encrypted form, direct from a machine in my home safe/vault.

Either scenario is possible, even likely. And the difference is this: in the first scenario, I'm a consumer of data, and the amount I send back to the internet is small.

But in the second, my home vault is a massive provider of information, and needs substantial upload capacity to the internet cloud, from where it reaches me. See what I mean by 'impossible to predict'?