As someone who works in the industry developing software, it's because for the high-end VFX studios, software like Blender and Natron just don't cut it in terms of efficiency, power and flexibility.
Artist time costs more money than software licenses do, and as the open source software isn't generally good enough for their needs, using better off-the-shelf software that have fairly decent and flexible plugin APIs is good enough for them.
A lot of the code VFX studios have is proprietary (as in unique to the individual studio who wrote it) plugins for host apps like Maya, Katana and Nuke, and shaders/integrators for renderers like PRMan and Arnold.
There are open source projects that are used for glue in the VFX industry:
But these have almost exclusively been internal proprietary projects started within studios that have been open-sourced, to help transfer assets between studios, and are mostly open file formats for images, volumes and geometry, as well as other intermediate tooling.
I guess I've just always been a little surprised that studios didn't open source more of their internal software. For example, Nuke was originally internal software at Digital Domain before they sold it to The Foundry, I believe. Rhythm & Hues also has a lot of its own internal compositing and 3D applications. And Pixar's Renderman software is also not open source (I know some aspects of Renderman are open source, but the actual Pixar Renderman software is not open source). It just surprises me that more of these internal applications didn't get open sourced. Blender is a good example of one that did. While I'm sure these studios were getting an edge by having their own software in the early days of cg, now I would think the studios would actually benefit more by open sourcing their apps so that other developers can contribute to the project. And the studios that make this software might be able to cut back on their dev teams. Just something I've always wondered about.
But having their own software still gives them an edge - that's why some companies are actually going the other way entirely - like Weta writing Manuka renderer and Animal Logic writing their Glimpse renderer. It allows them to iterate extremely quickly on-depand for shows that have issues and design things properly, instead of shoe-horning existing (and often badly designed and implemented) open source software that can do simple stuff, but wasn't designed with full flexibility in mind.
Different pipelines work in completely different ways, and what works for one studio doesn't necessarily work well for another.
However, I do agree in principle that given a clean start (Which is never going to happen given the hundreds of thousands of lines of code VFX studios have in plugins for proprietary systems) it would make sense for all VFX studios to share a loose framework application that very flexible and advanced plugins can be written for. Image Engine's Gaffer is probably the best fit currently.
In some respects it looks like Fabric Engine might at some point fit this bill, however it's not open.
Artist time costs more money than software licenses do, and as the open source software isn't generally good enough for their needs, using better off-the-shelf software that have fairly decent and flexible plugin APIs is good enough for them.
A lot of the code VFX studios have is proprietary (as in unique to the individual studio who wrote it) plugins for host apps like Maya, Katana and Nuke, and shaders/integrators for renderers like PRMan and Arnold.
There are open source projects that are used for glue in the VFX industry:
https://github.com/alembic/alembic https://github.com/PixarAnimationStudios/OpenSubdiv https://github.com/imageworks/OpenShadingLanguage https://github.com/ImageEngine/gaffer https://github.com/dreamworksanimation/openvdb https://github.com/openexr/openexr
But these have almost exclusively been internal proprietary projects started within studios that have been open-sourced, to help transfer assets between studios, and are mostly open file formats for images, volumes and geometry, as well as other intermediate tooling.