There is a lot of 'probably' going on here. He structures his entire P=NP argument on the basis that most computer scientists believe it to be false. That's fairly weak.
He has shown that markets are only efficient if P=NP, but that doesn't say much about whether they are therefore inefficient.
"In short, if P≠NP, then the market cannot be efficient for very long because the ability of investorsto check all strategies will be quickly overwhelmed by the exponential number of possible strategies there are to check."
He also states this, but even an overwhelming amount of strategies can be condensed into algorithms that makes 99% of them useless. See the game of chess for similar ideas. 99% of the possible moves are useless, so it seems as if there are an exponential amount of good moves when only a fraction are viable. Chess is NP, but still partially solvable under P conditions. I believe the market is like this too.
He has shown that markets are only efficient if P=NP, but that doesn't say much about whether they are therefore inefficient.