We talk a lot about software bloat. How good/fast do chips really need to be for defense purposes? Chips are so good now it seems you could be 2-3 generations back and still get the job done.
For Japan it's a crucial bet, a lot more than other countries in America or the EU as it's the crux of the economy.
I doubt rice cookers and toasters are using the latest chips for TSMC. E.g. the ESP32 microcontrollers are manufactured by TSMC on 40nm, and there are a bunch of different companies that have similarly sized/powerful microcontrollers being manufactured (e.g. STMicro).
The whole effort of bringing TSMC factories to Japan was triggered by the shortage a few years ago, when I think a Toyota factories had to slow down as they couldn't get enough chips to meet the production goals.
They'll probably rely on TSMC to make the deal work even if external competitors offer better prices or availability in the short term.
Using Russia's approach you can easily stay many semiconductor manufacturing generations back. Using the Western approach you will prefer staying up to date so you can continue using the latest the latest proprietary manufacturer-supported FPGA tooling.
The US, NATO in general, and especially Israel, are the other end of the spectrum: they prefer expensive, but powerful and very accurate weapons (and in the case of vehicles, prioritize survivability). For Western countries, wars are unpopular so they want to keep the fatality rate low, and Israel doesn't have that many people in the first place so their Merkava tank designs focus survivability even more.
When RADAR is the eyes and ears on the modern battlefield, you want as many computers on it as possible to find as many targets as possible and to differentiate them.
Just because most equipment must be trusted, it doesn't mean you can't use lots and lots of untrustable hardware.
Human soldiers can be killed. If e.g. Russia bombed a US military base in <country>, then the US has an instant cassus belli (and is politically forced to respond drastically and be drawn into the war) and <country> knows this, which makes the US military base a far more effective red line against Russia as a result. In other words, human solders act as a political tripwire that more effectively binds allies together.
Drones are not there to take over for humans, drone-based weapons are ridiculously effective compared with anything before them and there's plenty of space for more computing power to make them even more effective.
Actually, this is the main reason why drones will be needing more and more compute; to ensure that they can be as autonomous as possible.
Whether they will be able to reach or surpass humans in this regard is still disputed by some, especially those who think that the human brain is powered by some kind of non-material soul.
> human solders act as a political tripwire
This really only applies before a war starts. Once it has started, you want to have the best equipment you can within your budget.
Now there are other political reasons why autonomous drones may be resisted, primarily that they can be seen as very scary or even evil by onlookers. But the same could be seen by current remotely piloted drones, but that hasn't stopped the US from frequently using them for assassinations at least since Obama's period.
Once drones become able to operate without a human to control them remotely, they will. Especially if jamming limits their ability to communicate.