Here's a response to Robin's comment -- from Scott Golder of the HP Labs Cloudprint team:
A common concern expressed about the rise of cloud computing is that, when a third party is hosting or transmitting your data, by definition they have access to that data. Email is the biggest way this problem is manifested now; the major free, commercial webmail services routinely scan and index email, whether to do spam filtering or to serve ads. Interestingly, from a computer science perspective, these activities are similarly invasive, but spam filtering is seen by users as innocuous or even desirable, whereas serving ads is somewhat suspect. Similarly, activity history -- e.g. search queries and clicking patterns -- as well as user profile and other uploaded data, can be examined, in order to provide suggestions (of movies, books, new contacts, and so on), but in some cases, such "suggestions" are framed as "targeted advertising." So it's clear that users see some uses of personal data as desirable, and others as less so. Logging data has become the path of least resistance, however, and the databases and server logs and IT infrastructure that make desirable services (from email to social networking to cloud computing) possible also create opportunities for misuse/abuse (e.g. advertising and identity theft). A commenter on this blog (http://www.communities.hp.com/online/blogs/scaglia/archive/2008/06/23/opportunity-knocks.aspx#83627) recently expressed concern about Google Desktop. I think the commenter is expressing a feeling that is shared by most people, that the data on their desktop is a safe place, and the data is as private as can be. But services are migrating to the cloud and the barrier between the desktop and the internet is getting fuzzier. There is a little bit of discomfort generally, since users generally know (even if it's slightly exaggerated) that once things go onto the net, they're there forever. Users, then, are rightly cautious.
I'm not sure what the answer is, but I can think of two possible resolutions, cryptography and trust. I know of many people who encrypt their email using public key encryption, but cryptography is generally hard to understand and, in practice, very hard to use if one isn't a computer scientist or other kind of expert. If cryptography tools were easy to use and perhaps invisible to the user (maybe "encrypt everything I send with this key") then maybe people would use them. If every file -- or print job or email or what have you -- were encrypted when I posted it to the cloud and decrypted when I retrieved it, without ever having to enter a key or think about it, that could provide some sense of privacy.
The second, and maybe more mundane, answer is trust in organizations and credentials. There's some interesting social theory that suggests that when societies grow too large, there are too many individuals to have good information about the reputation of each person, so instead you trust institutions. For example, you don't have to have detailed reputation information about any particular doctor, but rather by trusting the AMA as an institution (and perhaps the fact that the doctor has a degree from a recognized school "brand") you have pretty good assurance about a baseline level of competence. The same kind of thing is true for IT companies. Undoubtedly one of the reasons people/enterprises buy HP products is trust. Perhaps they know others who had good experiences with our products in the past and trust our track record. They also know that support will be available if necessary, because HP's size means that it won't disappear tomorrow. These things are true for more established companies like HP, and are less so for smaller companies or startups; While this kind of institutional credibility is always important in business, it will likely be a significant asset for HP specifically in the cloud computing space.
That doesn't mean people won't continue to be healthily skeptical of cloud services. However, the combination of encryption and brand trust will hopefully allow consumers to be reasonably assured that their data is only accessible to entities who are most likely trustworthy.
I also think there are going to be tremendous lifestyle benefits from cloud computing, and users will be willing to take managed risks in order to share in those benefits. The simple existence of the mass internet as a source of information and means of communication really changed things in the past 10-15 years, but it has only been in the past year or two that really good, nearly-ubiquitous mobile net access in the form of the smartphone has been around, and I think that has changed things as much or even more. Instant mobile access to the net and to my own data has really changed by life. Even though I'm clearly in the early-adopter category, I think this will soon be commonplace for everyone who has a mobile phone. The instrumental value and pure enjoyment of this kind of access will render moot the question of whether to have cloud data; it'll be a given. The question is not just who will build the infrastructure first; it's also who will users trust enough.