From the first statement, the "price" refers to performance implications. So the second refers to that too and the article mostly talks about performance. Pure

From the first statement, the "price" refers to performance implications. So the second refers to that too and the article mostly talks about performance. Pure FP generally scales well with caching/memoization, because functions are pure, and with concurrency, because data is immutable. However the paradigm typically allocates more garbage and often creates more indirection due to its declarative nature (think function composition/pipelines/laziness and so on). So it is not as suitable for fine-grained/low level optimizations as imperative programming.