It's helpful to plan for scale out if it will be needed, though. It can be difficult to build that later if it wasn't planned for.
On the other hand, you do need to consider scaling up. You can get a HPE ProLiant DL385 Gen10 Plus with dual 64-core Epycs and 8 TB of ram and almost half a petabyte of flash storage. If you're starting from zero, it's likely a long way until that's not enough, by then bigger servers might be easily obtainable, IBM has a power server that goes to 64 TB, but then you're dealing with IBM and power.
If you do need a distributed system, if it's possible, you want to design things as you said, with as little coordination as possible during processing. Thinking about coordination in the system design early can help you make it easier to separate later. Coordination on a single host isn't as expensive as across hosts, but it's not free either, so it's not wasted work to consider early.
On the other hand, you do need to consider scaling up. You can get a HPE ProLiant DL385 Gen10 Plus with dual 64-core Epycs and 8 TB of ram and almost half a petabyte of flash storage. If you're starting from zero, it's likely a long way until that's not enough, by then bigger servers might be easily obtainable, IBM has a power server that goes to 64 TB, but then you're dealing with IBM and power.
If you do need a distributed system, if it's possible, you want to design things as you said, with as little coordination as possible during processing. Thinking about coordination in the system design early can help you make it easier to separate later. Coordination on a single host isn't as expensive as across hosts, but it's not free either, so it's not wasted work to consider early.