Why would a private company take all the hassle to implement a BOINC project in the first place?

It is actually incredibly easy to implement a project on a full BOINC server. You can get a BOINC server up and running, generating work units, sending them out, receiving the results back and validating them all in the space of a single afternoon. The only real work at this stage is populating your work units, but most of the heavy lifting for that would need to be done regardless of what compute source you are using to solve your problem. If we are talking about hardware optimisation, then yes you are correct and we run into a whole lot more work.

The reason that a company or research institute would want to use BOINC is twofold:

• First and foremost, Gridcoin already rewards BOINC crunchers for projects on its whitelist, for free. There is no cost to the project at all. Getting a project whitelisted is trivial if you are a little active in the community and have a solid research proposal with a steady supply of WUs.

• Using BOINC requires practically zero hardware investment, and would be far cheaper than renting hardware for your compute cycles. Not only is the upfront cost far higher to invest in hardware, but ongoing costs too. If I get a research project onto the GRC whitelist and double its GRC mint with rain out of my own pocket, you can safely bet cruncher will flock to the work. It would still only set me back some USD$30/day, which is incredibly cheap.

I'm not just making all this up, for what it's worth. I'm a PhD student working on HPC models of the brain, and spend a lot of time tackling endless compute problems. I have also set up my own BOINC server. BOINC is unparalleled in potential for most research applications that are not timestepping simulations (but that is another discussion).