He says that if 10% of men and 8% of women are suitable for tech, and google hires unbiased from suitable people, then google should expect a 55:45 gender gap.

That doesn't mean that the female colleagues are unsuitable for tech, since they were exactly hired among people who were suitable.

He is also not saying that this explains the complete gender gap, just that there is no point in blindly aiming at 50:50 without considering research on the underlying distribution.

He may have the research wrong, but otherwise the point seems to stand?

EDIT: Of course, this argument only works for a binary 'suitable/not suitable' distinction. If 'tech talent' was said to be normal distributed, with men having a slightly higher expectation, it would follow that the average men above some cut was also better than the average woman above the same cut. So in that way he does attack his colleagues.