He's absolutely right. Long-shot research with little to no immediate applicability has been the basis for innumerable breakthroughs over the years. If DOGE existed back in the 70s, 80s, or 90s, we wouldn't have mRNA vaccines[0], Google[1] (or the modern internet for that matter[2]), and many modern cancer treatments[3], to name but a few examples of research that would have been easy to brand as wasteful and not "important science".
[0]: https://pmc.ncbi.nlm.nih.gov/articles/PMC9975718/
[1]: https://www.sciencedirect.com/science/article/pii/S016975529...
[2]: https://www.nsf.gov/impacts/internet
[3]: https://www.nih.gov/about-nih/impact-nih-research/improving-...