Is 64GB of DDR4 a better long-term choice than 32GB of DDR5?

For your use case, the bigger issue is the platform rather than DDR4 versus DDR5 in isolation. DDR4 won’t suddenly stop running applications, and capacity is often more important than memory bandwidth when working with virtual machines, development tools, or local ML workloads. However, AM4 is essentially at the end of its CPU upgrade path, while a DDR5 platform gives you more room to upgrade the processor later. If you can manage with 32GB initially, I’d lean toward the newer DDR5 platform. If your workloads genuinely need 64GB now, the DDR4 system is still a reasonable value.

If you are buying a completely new computer and truly want to keep the core platform for five to seven years, I would favor DDR5 if the budget can stretch to it. You get a newer CPU generation, current motherboard features, and a clearer upgrade path. DDR4 is not useless or obsolete for everyday work, but choosing it means accepting that future CPU upgrades will probably require replacing the motherboard and memory as well.

Capacity should be your deciding factor. If your VMs, datasets, containers, or development environments can exceed 32GB, slower memory with 64GB will be much better than a faster 32GB system that starts paging to storage. For ML specifically, system RAM is useful, but GPU memory and the software’s CPU requirements may matter even more. Check the requirements of the models and tools you actually plan to run rather than trying to future-proof against unknown workloads.

For gaming and normal desktop use, DDR4 is still perfectly capable. The practical difference between comparable DDR4 and DDR5 kits is usually much smaller than the headline bandwidth numbers suggest, because DDR5 also tends to have looser timings. A good 32GB DDR4 kit can outperform a poorly configured or undersized DDR5 system in real workloads. Developers also have to support the large installed base of older systems, so applications are unlikely to be designed around DDR5 bandwidth alone.