transistors - Why are sum-of-products implementations more popular than product-of-sums implementations?

In my book about circuit design (Fundamentals of Digital Logic with VHDL from Stephen Brown and Zvonko Vranesic), the writers always prefer a sum-of-product for representing and implementing simple circuits.

In Boolean Algebra, this preference is used as well, but I think mostly because writing sum-of-products is just easier and shorter. And maybe easier to understand for readers.

But when implementing using logic gates I would suppose other considerations than these are made as well. Like costs and delays of the gates.

So, is there a specific reason why preferably sum-of-products implementations are made? F.e. are AND-gates cheaper than OR-gates? I read about the transistor realisation of these gates, but I can't recall such a statement.