I don't buy that, over its lifetime, the chipset has been only marginally profitable.
I just rechecked, and my claims were only partially correct regarding chipsets.
The reason is that the last numbers I checked were for a seasonally slow quarters, and that the chipset division has respectable gross margins but fairly thin operating margins, due to high R&D expenses. So the time they really shine is during seasonally strong quarters, and the annual operating profit was respectable in 2006, at $77M (vs $584M for GPUs, $85M for CE/PS3 and a *loss* of $41M for handhelds).
Historically, I'm not convinced MCPs made any money at all, but it's hard to say because operating income is only known up to 2004 and revenue up to 2003. Before that, they merged MCP revenue figures with that of the XBox deal, which were a significant majority of that 'division'. Ugh!
Anyway, here are the official figures:
2003: $162M revenue, unknown operating income.
2004: $176M revenue, -$40M operating income.
2005: $352M revenue, +$33M operating income.
2006: $661M revenue, +$78M operating income.
As can clearly be seen, MCPs were actually dragging income down and not up in 2003, as I remember hearing back then because MCP figures without XBox in them weren't available until much later. Of course, the original nForce R&D allowed them to get the XBox deal, but that didn't force them to stay in the market...
I think if you estimate what nForce's operational profitability in 2003/2002/2001 must have been (as in, PC-only R&D, excluding anything that could be sponsored by the XBox), you'll likely conclude that the losses should very roughly compensate 2006's gains IMO.
But even if that were true, it's not very relevant (investors couldn't care less about what happened 6 years ago.)
Correct, except from the POV of looking at management's track record and the success of their past diversification attempts. Considering NVIDIA is now trying to diversify further quite aggressively, I could see some people being interested in this kind of data if they knew of its existence.
What matters is the current profitability of the chipset business. As far as I remember, this business unit has been growing quarter after quarter for the last 2 or 3 years. It must be a sizable part of the overall company now.
See the numbers above. It's certainly a big part of the business revenue-wise, but much smaller in terms of operating profits. This is because the R&D expenses are high relative to the amount of revenue, and gross margins are still lower than for GPUs.
You can't have overall margins of 45% if a major and growing part of your products are underperforming to the extent that they barely outperforms treasuries.
NVIDIA's operating margins are not anywhere near 45%; you're thinking of gross margins there. As for NVIDIA's MCPs current profitability, they made a meager $20M of operating profit on $309M of revenue in the first six months of 2007.
Q1 and Q2 are seasonally down quarters though, so it'll be interesting to look at the 2H07 numbers once they're out. One important factor is that their market share growth in the AMD market has halted because of the ATI acquisition. Their diversification into the Intel IGP market will be key to improve MCP profitability.
To summarize, just as Fox5 said, NVIDIA's gross margins for their MCPs are AFAIK the best in the business (potential exception: *maybe* Broadcom, since they only sell server chipsets, but they don't really count). But their R&D expenses are also very high, which is why in terms of net/operating profit, the return on investment is nothing to brag about. Going forward, however, things are improving nicely but it's hard to say without the 2H07 data.
If you're interested in a quick read/look at NVIDIA's business model and gross margins evolution:
http://media.corporate-ir.net/media_files/irol/11/116466/Analyst07/MarvBurkett.pdf
(Otherwise an interesting post. I like the profitless prosperity quote!)
So do I. One thing I tend to be concerned about with NVIDIA is they are likely trying to maximize gross margins too much, sometimes at the expense of overall gross profits.
However, it's a complex trade-off, and doing it that way probably helps their brand image. Similarly, while MCPs weren't very profitable on their own in late 2004/early 2005, the nForce 4 made SLI possible and gave NVIDIA an edge in platformization and brand power. Those things are significant, but very hard to account for financially.
Anyway, that was a big OT - but since the overall message was related to NVIDIA's track record also being far from perfect, hopefully it helped inject a small dose of relativity into this thread regarding AMD's own track record, and ATI's current losses both in GPUs and handhelds.
Anyway, more seriously...