There is an organizational problem with doing long-term research internally in a high tech company: it is almost always more cost-effective to divert the researchers into solving some short-term problem related to a current project.
Outsourcing the research via a long-term commitment to funding an independent organization that has its own priorities (e.g. a university department awarding PhDs to its student "employees") is a good way to resist that short term pressure.
My employer (a multinational engineering company) doesn't have any delusions that every university research project will produced something "useful." It's more like investing in Broadway shows were one big hit pays for all the flops, but you can't predict in advance which show will be the hit.
The conflict between open publication of results in academic papers and PhD theses, and the commercially sensitive application of those results to benefit the sponsoring company, needs to be managed, but that isn't an insuperable problem. For example new analysis techniques or computer algorithms can usually be demonstrated using well-known problems addressed in earlier academic papers, or using sanitized data, while the "real" application remains confidential to the sponsoring company. If a competitor decides to learn how to use the research starting from what is openly published, it's their choice to fund the costs of that task, both in money and elapsed time.