They would go along with the PCI-SIG suggestion of what they've ratified in conjunction with what the ATX x.x standard put forward as a method of power supply. i.e. "we've made an 8pin pcie connector" - and PCI-SIG adjust their documents accordingly once it's passed through the ecn.
375w wouldn't be put forward as a limit due to heat disappation within a pc case as they know full well you can put 4x card in xfire/sli within a said case. 8x if you could the productivity servers you can install parallel gpu setups in.
To insist 375w was max and to allow the further addition of pcie x16 electical slots would be silly, don't ya think
EDIT: I have a feeling that maybe you're not seeing my view of how the system works, so I've put together a "brief" explanation in as best laymen's terms as I can:
In the model we're looking at, there are 3 prominent bodies:
ATX
PCI
Graphic card manufacturers
ATX come out with the standards of which power supply manufacturers are suggested to abide by when producing psus for the industry.
PCI have standards that apply to the use of busses, in the main, that are most commonly looked at as slots (even though they can be integrated also).
Graphic card manufacturers obviously produce the cards in our little model that display stuff - basically. AMD, Intel, nVidia, Silicone, Matrox and a few more.
Aside from these 3 there are many groups, with their own standards to which the above three bear in mind when putting forward their own studies/papers/standards, which range from safety people, environmental people, electrical, motherboard and other component people, the list does go on quite a bit.
Between all these groups there is a whole load of interaction, co-operation and studies. As an example, a graphic card manufacturer will come forward wanting to make an oem card that oems can use in their mass produced pcs aimed at business and the public. The oem has told them, as they usually do -very strongly- that they don't want ANY external power connections to this card, but it has to be more powerful than the present integrated/embedded selection. In this case, the gfx people can think of power and look directly at what PCI have put forward. They'll comply with their papers on what they've had motherboard manufacturers in turn comply with.
As time goes by, the consumer market gets more demanding. When PCI came out with their new standard for their gfx slot, they gave everyone the boundries at which the power would/could be used. As the gfx cards advance, their manufacturer saw the easy option of adding an additional molex cable to the side of their cards pcb to go beyond what PCI had stated would be available. Everyone spoke, and PCI added their errata/addendum to their previous paper on power use. It now includes "to achieve the power required for blah blah, a single molex is used" and so on.
Now the PCI's paper will include this addition. The bar has been raised as far as the gfxcard people are concerned, and time continues to move on, advances are made on this new power level.
Oems (HP, Dell, Acer, etc) btw are still insisting on NO extra pcb plugs.
The gfx people have reached a new era in their rnd, they need to surpass this poxy molex supply. They talk with the ATX people, who come up with a new type of connector (the 6 pin pcie for example). They produce their new paper ATX x.x and in turn PCI will catch wind of this - run a bunch of tests, do some studies, tell everyone they're happy, and bring out their new errata/addendum to the existing PCI standard.
Now the PCI's paper will include yet another raising of the power level supporting the *new* 6 pin pcie connector.
Repeat for the use of 2x 6pin, 8pin and 6+2pin, 2x 8pin and so forth.
Theoretically, the gfxcard people and the ATX people could be in discussion about a new 10 pin or 6+2+2 pcie connector. Each of the people will talk to each other, tests will be done, as per usual, regulations and studies will be re-issued with a new proposed power level. The ATX people say they'll bring out psus with 1x 8+2+2 connectors for the lower end of the market and 2x for ... possible dualing of these newer cards. Bringing a possible new power threshold that 2x 10pin connectors on a single gfxcard can handle. (purely theory, I can't see this happening with the proposed new die shrinks also - but who knows - it is possible)
As each of these groups talk with each other, conduct their own internal test, bars are continuously raised. A quoted paper stretching back to 2007 regarding the power levels for pcie use can only suggest about what is currently available. It, at that time, has little idea that 2x 8pin may become that popular.
One thing is for sure, stress to pcbs due to the plugging and unplugging, pull and such like of additional power sources is not favored. Which is why alot of oems dislike them. One of the defenses against a 2x 10pin power arrangement is that it emulates the mobo power connectors which would come too close to stressing the rear or top of the card. But ingenious inventions could work around it somehow. 3x 8pin is obviously suggesting a similar cable to what mobos have now. Sticking those at the back of an 11inch pcb is not wanted I don't think.
