(first my disclaimers...) I'm a research scientist. I've worked in academia, for a government lab managing grants, and in private industry.
There are many good reasons to change the way science funding is done in the USA.
First, we all know here that there is a surplus of certain STEM labor, including a large number of the researchers (postdocs, grad students, etc.) funded by the government.
Second, there is a serious and long running lack of practical progress being made in science. By some metrics (# of degrees, # of papers), we are doing great, but by others (# of companies founded, return on investment, research efficiency) we are at a generational low-point.
Third, some practical STEM fields (i.e. medicine, manufacturing engineering) DO exhibit a labor shortage, and also rely on training programs largely outside the research grant driven model.
The budgets we're looking at in the government grant space are enormous. It doesn't seem that way to many researchers, but the annual NIH budget alone is about equal to all of the funding provided to all startup companies annually. There's a lot we can do with that, provided the right direction. NIH, for example, could be re-focused on grants for training medical doctors, PAs, nurses, etc., instead of researchers. Yes, that would slow research down, but it would also contribute significantly to lowering the cost of medical care, and it would be appropriate for the mission and people at the NIH. A mature approach to climate change might cut some climate research funding, but increase funding for faster roll-out of a power and transportation infrastructure free of fossil fuels. Surely such an infrastructure could be an obvious point of agreement between the right and the left; start the construction in coal country.
A thoughtful approach to science funding would encourage researchers to look beyond their next federal grant to other (private) funding sources, and would encourage (force) private funding sources to invest in transitional research. The UC pension system has been instrumental in fueling the startup economy for a long time by devoting 1% of it's money to funds investing in startup companies. If other groups did the same (... were forced to do the same...), we would increase the total amount of science funding by several orders of magnitude more than the total federal R&D budget. Prior to the 1990s, all large DoD contractors were required to spend 15% of their budget on R&D projects that were reviewed by government scientists to ensure they were actual R&D projects. Removing that requirement shut down a lot of very good industrial research programs. We learned then that most companies performing internal R&D can't compete with companies using subsidized academic R&D. That's an important lesson that the pharmaceutical industry is just now discovering, and it's an economic fact we need to fight. Reinstating requirements like minimum and audited internal R&D budgets for government contractors would also increase private spending on real research.
Not all research can use a "transition to private funding" model, so there is a need for continued blue sky research funding from the government. However, right now, we are saturated with the results of blue sky research and in serious need of support for transitional and applied research. As a nation, we are paying for this basic research, but we are not seeing the benefits of it. Some small amount is commercialized here, some is commercialized elsewhere, but a whole lot just gets forgotten. That's a waste, and it's stupid.
So basic research could be de-emphasized for a while, and non-government resources could be directed to lead to an overall increase in work and funding for researchers (while also delivering a profit... usually). That's another way of saying that a decrease in federal research funding could be done in a constructive way. We could even look at the labor market for cues as to whose graduate education we should be subsidizing. However, this is not what Trump is suggesting here... but it's nice to daydream about what an intelligent jobs-and-commerce science budget would actually look like.