True, but if AMD is struggling to fit 4.3B of TSMC's 40nm process tech transistors into a power-footprint of 300W for the ATX spec I can't imagine what kind of clockspeed reductions Fermi will have to take on in order for Nvidia to shoehorn 6B xtors from the same process tech into the same 300W footprint...
We don't even know the clockspeeds yet. Maybe Fermi's clocks don't have to be so high? Also, I can't remember, but does Fermi also have something designed to make it consume less power, idle and loaded, like Cypress/Juniper?
Whenever Fermi is out, I hope TSMC has 40nm yields up to snuff already. If they have ~50% right now, I hope they up that to ~70-80% by march (is 10% improvement per month too much?), so that Fermi doesn't suffer too much production problems, and the rest can be harvested. Perhaps that's one strategy nVidia will use to be able to roll out a complete Fermi line of graphics cards in spite of TSMC's 40nm problems (if they do still exist at that point in slightly less measure).