So what's driving the projected margin growth in the third quarter? According to you guys it's certainly not GF104 (which should be destroying margins according to Charlie).
As you said, the margin growth is merely a projection at this point, but let's say it's a correct one.
I don't know about GF104 "destroying margins" but there are a few simple facts that are hard to dispute:
- GF104 is bigger than Cypress, therefore costlier to make (and Charlie isn't the only one saying so);
- GF104 is slower than Cypress, therefore sells as $199 and $229 cards;
- Cypress sells as $199 to about $399 cards.
Those are facts. And I would add that GF104 is only available in a crippled version, and even the 256-bit version (still crippled) is apparently a bit harder to get than the 192-bit one. Plus, it is a derivative of GF100, whose yields were rather poor at best. So I'm going to add that due to this plus the size factor, GF104's yields are significantly lower than Cypress's. That's not a hard fact, but I think we can put it down as "very likely".
From there, the conclusion is simple: either AMD enjoys really stellar margins on Cypress and NVIDIA makes decent money on GF104, or AMD enjoys good margins on GF104 and NVIDIA doesn't make much on GF104.
As to the source of improving margins, it could be a more favorable product mix, a result of lower 40nm wafer prices, a by-product of the chipset market dwindling (it's a low-margin market, so if it became less significant, average margins could go up), Quadro and Tesla picking up… or all of the above.
By the way, I couldn't find any Fermi-based Quadro, has NVIDIA released them yet? Do they plan to do it soon?
