OK, so back to some comments from earlier again:
Aside from the cost to buy a chip+bridge chip versus just a chip, what about the impact on the expense of manufacturing and testing an actual card that uses one chip versus two?
The cost difference for designing a distinct version of the chip is an up-front, and one time unique cost per chip (if done without rushing, presumably), but the savings for simpler board design and validation should be on each and every board made and chip sold of a revision. Correct?
With an OEM computer seller push for "PCI Express" as a "checkbox feature", it seems to me like there will be plenty of opportunity to offset that initial cost with numerous cards in OEM "PCI Express" checkbox machines as part of the PCI-Express launch effort, as it leaves more room for maintaining a higher margin, room for competitive pricing to increase marketshare while still making money, and an attractiveness to board makers for less costly board manufacturer.
In terms of chip sales profit, however, having the native PCI Express chip only needs to happen in time to balance clearing out stock and having PCI Express cards available and priced competitively and maintaining profitability, with validated board designs of lesser complexity to go with them, and having the designs tested soon enough not to have to "rush" and spend extra money to do so. nVidia seems like they can still have native PCI Express chips in time for this, except that they haven't provided widespread confirmation of this effort being non-problematic yet while ATI seems to have done so. However, if they don't spend extra money on rushing things, and have things ready when necessary, there should be no real impact on this aspect of a native PCI-Express chip design.
The disadvantage for nVidia here seems to be in [relatively minor] the perception of "PCI Express leadership" amongst the "tech savvy" (like us, and the filtration of the impression that makes it into article commentaries and forum discussions across the net), [major, IMO] the confidence and desirability to OEM board manufacturers of their schedule, and how that trickles down to units moved to OEM PC builders on the "PCI Express" bandwagon.
I think this could easily relate to such things as Abit manufacturing ATI graphics cards as one example, and, with a grain of salt, ATI talking with confidence about OEM perception of their product line-up (don't have the quote handy).
Other factors might be involved, such as performance expectations outlined by each IHV (for chips and PCI Express), or conclusions OEMs are making about IHV marketability, or aggressive pricing steps taken by ATI that don't have a cost offset, but information about that isn't readily available at the moment...