A long post, but having worked both in Massa land, Desh , short stints in Panda land and nearby areas, I believe the actual truth is somewhere in between.
One thing which cannot be disputed is that India is cheap for IT like how panda land is close to unbeatable for manufacturing. Doing all the software in Massa is similar to trying to do all the manufacturing there. Sure it would work for high end jobs , similar to high end manufacturing but from my personal experience 60-%-75% of software is generic in nature and it just not be cost effective to do everything in America. Hence irrespective of tariffs and barriers , what could be outsourced will be outsourced. Unfortunate truth , but one that cannot be avoided.
Also Panda land (or its brother Taiwan) has a huge benefit in developing software in Embedded space due to primarily the time it takes to ship something to India, time lost in customs, in-ability to make minor/major changes quickly and retest ..etc. Hence while it may be 2X expensive, it is still better in a time to market perspective to do the embedded development in Panda land or its neighbors (from personal experience).
In traditional software development or the emerging software (Cloud, Machine learning, Deep learning, SOA, Microservices..etc) they lag much behind than us primarily due to the language barrier and lack of ability to quickly read hundreds of articles, google, ieee/acm papers and merge them together into mix. We are still strong and cheap in these areas and from my travels I don't see anyone else in Asia challenging our dominance anytime soon.
Saying that things are not rosy for traditional commandos and generals in our software services field. I was extremely lucky in my current organization to work in all the emerging technologies and quoting from my personal experience,
a. I assumed that machine learning was a fad till i learned the whole thing properly and it was able to solve some extremely complex logic problems for which my puny switch cases and if-else logic never worked.
b. Then I assumed that Deep learning was a bigger fad, till it worked and suddenly reduced 90% of my time in not doing complex feature engineering. It still cannot do everything but when it works, it works really well (or) at-least better than manual.
c. Finally I never liked the SOA/Micro-services/Docker architecture till my backed become more and more complex that I was afraid to even touch the same. Now 95% of my deployments is in Docker/Micro-services.
d. I always liked the Cloud and NoSql databases (primarily Cassandra) so was not wrong there
e. Nowadays, frankly, I don't see the requirement or appeal of leviathan back-ends in J2EE, expensive sharding, semi-manual geolocations replication in SQL databases, heavy manual testing, Systems admins who can only configure some firewalls and routers...etc anymore. Not to mention the kind of automation which can be performed.
My point is that 90% of people in Indian services company would never have had the exposure or opportunity to have worked on these changes and one fine day they would be laid off and it would be too late to orient to any of these.
Hence while I don't see our dominance ending anytime soon it would still be rough ride for generals and commandos (~> 10 years experience) in Indian IT who can do only people management or have only legacy skill sets.