(The other answer currently offered does well at explaining the advantaged of camera based stabilization. This answer will look at the disadvantages of camera based stabilization.)
Putting stabilization in the camera doesn't work as well in the focal lengths where it is needed the most. That's pretty much it in a nutshell. The best lens based stabilization systems provide up to four stops of room to slow down the shutter speed when hand holding a camera. This is true regardless of the focal length of the lens in question. No existing camera based system can provide one half of that at any focal length beyond the medium or normal focal length range defined as the length of the sensor's diagonal.
You don't need to wait until you are in ultra expensive Super Telephoto focal length ranges to demonstrate this either. At 80mm the same amount of stabilization provided by a camera based system is one half as effective as it would be at 40mm. By 160mm it is one-quarter as effective, and by 320mm (not much past the fairly cheap zooms that top out at 300mm) it is only one-eighth as effective. That four stop benefit from IS at 40mm is now a half stop benefit at 320mm. And in a marketing environment where smaller and lighter is becoming more and more popular, the amount of space to fit the room for movement (assuming the size of the image circle cast by the lens is large enough - it often would not be. Making lenses that have larger image circles with higher image quality further from the center is expensive...) and the servos to do this just isn't there, even if capable servos could be produced with current technology. They can't. And if/when they could they would likely be prohibitively expensive, kind of like those Super Telephoto lenses. If your camera has a much smaller sensor it doesn't matter, either. Because instead of 40mm now 20mm is your normal focal length, and the 160mm lens needed to get the same field of view as the 320mm lens attached to a camera with a sensor twice as large is still 8x the focal length of your normal starting point.
To put it another way, to get the same amount of stabilization in camera for 320mm that you get for 40mm, the sensor would need to move eight times as far in the same amount of time at 320mm as it moves when correcting for camera motion at 40mm to get the same amount of stabilization in terms of how many pixel widths of blur can be accommodated before the blur begins to show up in the resulting photograph.
There are other considerations as well. If in camera stabilization is accomplished by moving the sensor this can be problematic for parking the sensor exactly where it needs to be to when stabilization is turned off and absolute image quality obtained by insuring the sensor is motionless and perfectly perpendicular to the optical axis and at the exact registration distance from the lens mounting flange is desired. Even if it comes out of the factory perfectly aligned, repeated usage and the resulting wear and tear will require periodic calibrations to insure that such alignment is maintained. And measurements and adjustments of this type are very labor intensive and require ultra expensive test equipment.
The same about wear and tear can be said of lens based systems, but it is generally easier to swap lenses on an interchangeable lens camera than it is to swap cameras on an interchangeable lens! And in the case of higher end camera bodies where the users in this category (wanting absolute image quality with no regard for stabilization since the camera is already locked down on a tripod) are probably concentrated, many lenses cost less than the camera body. And it is generally cheaper and less labor intensive to replace a defective IS unit in a lens than it would to replace a sensor assembly (including integrated sensor) in a camera!
Brands that offer no equivalent high end cameras and lenses such as those offered by Nikon and Canon have less to lose by placing stabilization in the camera. They can leverage promoting the advantages (buy once, works with every lens, even older ones made before stabilization existed, etc.) in their marketing while ignoring the disadvantages. After all, that's what marketing is! Brands with many high end users would need to create virtually two separate systems for users that desire stabilization and users that desire absolute image quality. Then, in order to market the non-stabilized bodies, they would need to explain why their stabilized bodies are not ideal for high end work!