For one, WSDL/SOAP web services back then weren't used the same way. They were generally used to split apps in "tiers" (eg: business logic, frontend, data layer). Sometimes someone would extract a few more for scalability or whatever.
That's why "the cloud" (things like AWS) is an enabler here. You didn't have that back then.
We're not talking splitting the app in 3-4 pieces. I'm talking splitting an app maintained by 100 devs in 2500 services. Effortlessly (thats the key and the only reason its viable).
Tools like Mesos/kubernetes and continual deployment setups make this viable.
At work, if I want to make a new service, it's not much more difficult than adding a function: Run the generator, open the service, add some code, push, hit a button or type something in slack. Done.
We're talking 2 minutes (and on that, 1 minute is the time it takes for IntelliJ to open. You can cut that down using vim, lol). We deploy services and apps to production an average of 12 times per day per developer. Because it's "free" to do.
Microservices are completely non-viable if you can't do that. If you don't have the infrastructure to do it. You need to get the benefits without paying a significant cost.
Doing it like we did it 20 years ago (yes, I was a out of college then too) would have been batshit insane. And splitting stuff in just a few pieces isn't the same.
To make an analogy, it's like comparing svn branches to git branches. I wouldn't make 15 branches a day in svn.