After a comment, I realised that I hadn't explained my position well enough, but only argued for a single facet of my opinion. Here is a fuller version.
There are cases where making a wide design can be beneficial, and a good designer will use these opportunities. But such conditions occur rarely. If they are not present, both a narrow fixed-width design and a combination of a space filling design for wide screens and narrow design for narrow screen are good, legitimate choices. Creating a wide version only and delivering it to everybody is never good.
A designer has to design for some width, which I will call the primary design width. The designer will aim to maximize the usability at the width chosen; I will call the achieved amount of usability the full usability of the design.
It is frequently a requirement that the design keeps as much of its baseline usability as possible when the width is reduced (responsive design). This works by making "stretchy" designs and sometimes changing the design elements (e.g. reducing the number of columns when a very narrow width is recognized). But it is recognized that at some point, each design will degrade (have reduced usability as compared to the baseline) when the width gets too small. Even a simple text column will be less readable at, say, 1 cm width.
It is humanly impossible to design and test at all widths at once, but a good designer will usually work on the narrower and wider widths in parallel to the primary one, or at least keep them in mind and make design choices at the primary width which allow for easy transition to wider and narrower.
Reasons for not filling up the space when there is more width available than the primary design width
Because more space efficiency does not automatically lead to higher usability. "Wasted space" is not a negative factor in usability. Even the "above the fold" rule is losing credibility, and when vertical scrolling is needed anyway, a few extra turns of the wheel are typically not a problem.
The designer has already achieved full usability at the picked primary design width. She can try to increase it within the increased freedom provided by the larger space, but unless the space was already a severely limiting bottleneck at the chosen primary design width, it is unlikely that the usability will be increased above full usability to some state of extra usability. Of course, if the designer has found a solution which increases the usability when compared to the narrower width, she should use this solution.
An example of designs which do benefit from the added width would be the Remember the milk web application and the Mendeley desktop application, which don't use the classic master-detail pattern, but place the detail editing form in an extra column. This allows the user to select multiple masters and set one or more detail fields for them simultaneously to the same value.
But such examples are rare. In most cases, width is not a limiting constraint, and filling up a wider space does not increase usability. The designer can, of course, create a more space efficient version which gets delivered to users with wide screens. But if the usability remains the same between the wide and narrow version, there is no need to use the wide version. It is the designer's (or her customer's) choice whether to make a wide-screen-optimized version, to let the primary design width version stretch dynamically, or to leave the design at a fixed width so that a wider window is filled with the empty background pattern. When there is no difference in usability, the only criterion is aesthetics, and this will differ between people, there is no way to pick one which pleases everybody.
In the worst case, making a more space efficient version will even reduce usability. A user performing a specific tasks has specific information needs. Any information and functionality not needed for the current task is actually distracting. If you have ever complained of "bloated" software, you know what I mean by saying that packing more information or buttons on the screen can be detrimental. And even if there is no distraction, clutter and lack of white space will typically reduce the usability of the design.
But why not pick a primary design width larger than the classic ~1000 px @ 100 dpi?
(I am personally a proponent of 960, but have seen many people who use 1024 too)
Simply: Because it punishes users who don't conform to your expectations of a desktop computer with a modern wide screen. Basically, the more assumptions you make about minimal screen width, the closer you get to earning the dreaded "Works on my machine" distinction.
One major advantage of a wide screen is that it is possible to have multiple windows, tiled. If the (web?) designer makes the assumption that the user will always maximize the window, he/she stands in the way of some very efficient workflows. Basically, just because the user bought a large screen, they didn't give you the permission to occupy all of it with your application.
There is a multitude of devices and platforms (and has always been). Just because desktop computers tend to get larger screens lately, this doesn't mean that all devices which run your application have them. This is especially true of web applications, which have to run on all kinds of platforms, but it can happen that users want to run your Windows desktop application on a netbook sized computer, a small screen embedded somewhere in an industrial setting, and so on.
Users are not obliged to buy the latest and greatest hardware. Many non-IT-affine people out there still use ten year old monitors in a 3:4 form factor and low resolution. And they have an expectation that your application runs at their computer too.
With the advent of higher dpi screens, more applications start allowing decent scaling. The more applications allow pixel-independent scaling, the more it gets likely that users with worse eyesight, or users who have placed the monitor farther from the eyes due to space restrictions on their desk, will want to zoom in your application. If it is too wide, it will have no space to zoom into. (Admitted, this is something of a future worry, because most applications are very far from good pixel-independent zoom yet, and operating systems don't have a good native support for it - the option exists on Macs, but I think it needs third party software).
And don't forget that most websites out there are still about text. Longer lines are very detrimental to readability. The optimal reading configuration contains 60-70 character wide lines which follow vertically in a linear manner. And while the user is engulfed in reading, he doesn't want to have anything beside the text. But if you put something else there, he will ignore it anyway. This is a special case (applications can sometimes make a good use of a right column, see Remember the Milk and Mendeley for decent examples), but it is a very common case.
And why start with a narrow primary design width at all?
Keeping the full usability at reduced width gets trickier when the difference between the primary design width and the reduced width grows. It is likely that the designer will have to introduce more element changes as opposed to simple shrinking/reflowing. This is more effort for the designer and ends up in a less consistent design (and consistency itself is a factor of usability, through increased learnability). For this reason, a designer will prefer to choose a primary design width which is as small as possible (so that reduced widths can't get much smaller) while still allowing enough freedom to establish a really good baseline usability.
The distribution of widths available to the user is far from smooth; there are large jumps at standard display sizes. So a designer will frequently pick a standard size for the primary design width, as opposed to just a few pixels less (this will restrict the designers freedom, without reaching more users).
Of course, the primary design width shouldn't be chosen too narrow either, or it won't be sufficient to achieve full usability. There are people who design mobile-first (picking a primary design width corresponding to a smartphone screen size), but they will typically also make a desktop version with substantial changes, which will also have increased usability.
Why users' calls for a wider designs are not always a good reason to deliver them
The first reason can be applied to practically everything in design and requirements engineering. It stems from the fact that evaluating the usability of a finished product is easy, but evaluating the usability of a proposed design is a difficult learned skill. It is a situation of "I can't tell what [a usable design] is, but I will know it when I see it". It can be that a user pays money for a new, wider screen, then sees lots of white space on it and feels that the space is somehow wasted. He comes up with the solution "just put some design in there", evaluates it for downsides, and sees none. In many cases, the user will also have previously experienced occasions where going from a low width to a large width increased usability (e.g. when he upgraded a 13 inch monitor to a 17 inch one) and will not notice that this effect does not hold for all width increases. Then he can start demanding a wider design. We designers cannot expect our users to know what is good for them, but we are also not required to fulfill their wishes when they are not leading to a situation which is good for them. (Unless there is a contract which says that we have to do just that).
The second reason is specific to screen width. The majority of people start learning computers, achieve a state of intermediate knowledge, and are not willing to put more effort for achieving more advanced skill levels. This is a very efficient learning strategy for a non-changing or slow changing technology, as learning the skill to an advanced level is subject to diminishing returns. In computers, we are faced with a situation where tiling windows would be a great improvement over using each window maximized, but this was not so 10 to 15 years ago due to the small screens and low resolutions prevalent then. So people who have long ago achieved a comfortable level of computer literacy don't ever think of tiling and never try it. OS creators don't have much incentive to offer proper tiling, as no users demand it. As a result, while using two or more windows tiled on the current large monitors would frequently lead to much higher usability than having one application using up all the space, users just don't think of that, and demand the suboptimal "fill my expensive space with your application/website" solution instead.