NVIDIA shows signs ... [2008 - 2017]

[Jawed]

They can still sell part or all of the inventory that was written off (even after the write off they are pointing out they still have high inventory levels) which will then be, effectively, very high margin sales in at least Q3. It is difficult to conclude that GF104, as it stands right now, is good margin based on future predictions of margin - it is also likely that as GTX 465 and GTX 470 inventory is depleted they will move GF104 into a higher segment, improving the overall margin mix for GF104.

I think the key point is that GF104 was itemised as achieving very high margin. I doubt NVidia has an inventory of GF104 which has incurred a write-off, on the basis "that we don't expect to count that as revenue".

So I think GF104 is in the clear. It's simply too new.

Separately, I don't think GF104 can retail for much higher prices. There are already substantially overclocked cards out there which are retailing lower than HD5850 - they've shot their load. Only demand outstripping supply can take it higher, I reckon.

These results are a classic "throw the baby out with the bath water" reporting in order to look better for subsequent quarters.

I agree. I also agree with the general point that "inventory tactics" are part of the game.

The interesting thing this autumn is going to be whether AMD replaces Cypress first or not. I have a feeling it won't. Coupled with an October launch I think GF104 has got a very solid period of plain sailing ahead of it. The only risk being if Juniper's replacement is 30%+ faster at $150 (20% would seem likely to be the limit), which would cut off GF104's Christmas sales.

G92's clocks/specification rose substantially over its lifetime, which is part of the reason it was able to hold its pricing against RV770 and run for so long. GF104 does look like that kind of chip - the 28nm b version could be loving us long time.

 

[Kaotik]

How much has the whole "bumpgate" cost nV so far? IIRC this wasn't the first time they paid / marked losses for it?

 

[neliz]

Aug 12 - 2008 was their first "one time recharge" for "weak die/packaging material" in "selected products.
$196M - 6.7M reimbursed.

"We're not expecting more write-downs in the future," Huang said during a conference call with financial analysts. "We think we have a pretty good handle on the situation. We thought we were relatively conservative, but we'll see how it goes."

Replacing the chips "isn't something that we absolutely need to do," he added. "But we stepped up to do it because we think it's the right thing to do."

Oct. 25 - 2009

which excludes a $93.9 million net charge related to the weak die/packaging material set that was used in certain versions of our previous generation chips

Mar. 18 - 2010

During fiscal year 2010, we recorded an additional net warranty charge of $95.8 million against cost of revenue to cover anticipated customer warranty, repair, return, replacement and other costs arising from a weak die/packaging material set in certain versions of our previous generation MCP and GPU products used in notebook systems. This charge included an additional accrual of $164.4 million for related estimated costs, offset by reimbursements from insurance carriers of $68.6 million that we recorded against cost of revenue during fiscal year 2010. During fiscal year 2009, we recorded a net warranty charge of $189.3 million charge against cost of revenue for the purpose of supporting the product repair costs of our affected customers around the world. This charge included an accrual of $196.0 million for related estimated costs, offset by reimbursements from insurance carriers of $6.7 million that we recorded against cost of revenue during fiscal year 2009. Although the number of units that we estimate will be impacted by this issue remains consistent with our initial estimates in July 2008, the overall cost of remediation and repair of impacted systems has been higher than originally anticipated. The weak die/packaging material combination is not used in any of our products that are currently in production.

Aug. 12 - 2010

The charge, of $193.9 million, includes additional remediation costs, as well as the estimated costs of a pending settlement of a class action lawsuit consolidated in the District Court for the Northern District of California in April 2009 related to this same matter.

The problem is you can only find the reimbursements in the annuals.

Over half a Billion $ blown on something that Charlie dreamed up and has no proof for.

 

[Alexko]

He must have a pretty vivid imagination…

 

[neliz]

He must have a pretty vivid imagination…

Mind over matter, every night he worships his Shrine of Riva and uses telekinesis to fornicate the beejebus out of your NV40/50/60 bumps

 

[BRiT]

Charlie's certainly a Sage on a Rampage.

 

[ShaidarHaran]

Over half a Billion $ blown on something that Charlie dreamed up and has no proof for.

I'm assuming this is a sarcastic remark. I can't count the number of HP laptops with bad NV chips I've seen over the years.

 

[Jaaanosik]

I'm assuming this is a sarcastic remark. I can't count the number of HP laptops with bad NV chips I've seen over the years.

Of course it was sarcasm. I am one of the victims as well. My HP laptop died the last year. It was not even 2 years old.

 

[thatdude90210]

So far, that's half a billion in "remediation costs." But what about the cost of lost business opportunities? I know if I was in charge of a business and one of my suppliers screws me over that badly, it'll be a long time before they get any business from me. Because when the product goes bad, the customers will blame me, not my supplier.

In this case HP, Dell, and others' reputations suffer because of Nvidia's mistakes. I'm surprised Nvidia do as well as they do now.

 

[Silent_Buddha]

In this case HP, Dell, and others' reputations suffer because of Nvidia's mistakes. I'm surprised Nvidia do as well as they do now.

Not too surprising as their high end chips were relatively unaffected. As well, in the notebook space there was the release of Optimus tech which is quite good.

And none of the OEMs are stupid enough to ignore Optimus tech in notebooks no matter how much they may have disliked the situation with bumpgate. Doing so would have given a rather large advantage to their competitors.

Regards,
SB

 

[3dcgi]

I don't buy it and agree with Speccy, this is just an attempt to get all the bad news out at once and benefit from the depressed COGS next quarter. Also, I doubt any of these companies ever dump chips. Anything can be sold given the right price, although not necessarily for a profit.

They probably got blindsided by the pace with which AMD produced down-market Cypress variants. That completely wiped out their competitiveness in that segment. And then they had serious troubles in the high-end segment with GF100 too. Recipe for what we're seeing now.

I don't know the truth, but a flat out lie on a conference call would seem to be risky business to me. Plus, I can foresee a situation where it doesn't make sense to sell inventory. I haven't done the math, but there must be some point where it makes sense to landfill your products because newer products will make more money. Of course if it's the competitor's new products then dumping inventory might be a good idea.

 

[Psycho]

Not too surprising as their high end chips were relatively unaffected.

I've seen loads of ppl with 8800s (both g80 and 92) needing the "the oven trick" (which usually only works for a few months btw). And while I thought that was an issue with the card (which is still hurting reputation) - not the chip, I'm not so sure anymore.. Especially as the baking also works on some 8600M cards..But ofcourse most of these 8800 problems occur too late for the warranty and are not as expensive as a new laptop motherboard.
As Charlie predicted, the thermal cycling is worse for the laptops, so the problems just occur faster..

 

[Silent_Buddha]

I don't know the truth, but a flat out lie on a conference call would seem to be risky business to me. Plus, I can foresee a situation where it doesn't make sense to sell inventory. I haven't done the math, but there must be some point where it makes sense to landfill your products because newer products will make more money. Of course if it's the competitor's new products then dumping inventory might be a good idea.

Even if a flat out lie it would be hard to prove. You can always make projections that show that you don't anticipate being able to sell X amount of inventory. The keyword he used in his statment was believe. So basically they don't believe they can sell those items.

He also didn't state that they would not sell those items. Only that they wrote off what they believe was excess inventory that they didn't think they would be able to sell. And then further covered their arses by saying they didn't use that inventory in the guidance for the next quarter.

It's classic CYA and PR speak. If they happen to sell some of that inventory (as they have in the past) that will just be a happy coincidence that just happens to benefit the company and its shareholders. And if they don't manage to sell much or any of it, then they can claim to have correctly predicated changing market conditions.

Regards,
SB

 

[3dcgi]

And then further covered their arses by saying they didn't use that inventory in the guidance for the next quarter.

I disagree. This comment is the reason they cannot be counting on selling this inventory in order to reach the target margins. Selling this inventory would put them above the ~45%, but they're saying it's not necessary to reach it. Obviously these predictions come with the usual caveats, but it would be very disingenuous of them and IMO unethical to be telling Wall Street they didn't consider selling this inventory if they actually are counting on it.

I agree it would be hard to prove such a lie. Nvidia was quite cagey in discussing the written off inventory though so they've cast doubt on their motives and left this open to speculation. IMO reputation is still worth something in this world so it would be unwise for Nvidia to risk tarnishing their good reputation (as I perceive it) with analysts.

 

[Kaotik]

http://www.reuters.com/article/idUSTRE67C3QP20100813
http://www.businessweek.com/news/2010-08-13/rambus-nvidia-agree-on-memory-controller-license.html

They got to an agreement.

But now the interesting part for all of us.
Exactly how many SDR using products nVidia has sold from their birth to today?
And same for DDRx using products.

So we can add up how much they have to pay retroactively

edit: nvm, rambus is only seeking money from the past 8 years, and still needs to win that case in civil court

 

[Jaaanosik]

Three Months Ended May 2, 2010:
................................................ GPU ..............PSB..............CPB................All Other..........Consolidated
Revenue ..........................$ 780,853......$ 189,730.......$ 31,230 ..........$.−.................. $ 1,001,813
Deprec and amort exps.... $ 34,506........$ 5,395......$ 6,893..........$ − ....................$ 46,794
Operating income (loss) ..$ 115,344 ....$ 73,865 ...$ (41,816) .......$ − .................$ 147,393

This is from the NV financial statement. GPU - GeForce, ION; PSB - Tesla, Quadro; CPB - Tegra

PSB is 1/5 in revenue and a little bit more than 1/3 in income.

It's far from 3/4.

From here (page 24).

Revenue line Q2'11 update:

................................................ GPU ..............PSB..............CPB................All Other..........Consolidated
Revenue ..........................$ 550,400......$ 215,100.......$ 45,700 ..........$.−.................. $ 811,200

 

[Silent_Buddha]

Some nice growth in the Professional category (PSB), but that's to be expected with Fermi slotting in quite well there.

That GPU hit is just huge though (~1/3). Really puts into perspective how much the problems related to Fermi have impacted their core business. The delays, non-optimal size and operating conditions for a consumer space that has a performant competitor, etc. As well, that it quite likely delayed introduction of the rest of the GF1xx line, including GF104.

Q4 will definitely be better if for no other reason than GF104 is going to do well and continue to do well. Hopefully the rest of the GF1xx chips will also be competitive.

Regards,
SB

 

[Alexko]

Q4 will definitely be better if for no other reason than GF104 is going to do well and continue to do well.

I'm not so sure about that. It's doing well now, but the relationship between GF104 and Cypress is quite similar to that between G92 (non b) and RV770: GF104 is bigger and slower.

Right now the HD 5800s are expensive, mostly because AMD is still supply-constrained, but if the GTX 460 starts to really eat into AMD's sales, they could easily drop the HD 5850 to $229 or so, which would make the GTX 460 1GB irrelevant, and the 768MB variant less than attractive.

Plus, Q4 should see Southern Islands launched.

 

[Silent_Buddha]

I'm not so sure about that. It's doing well now, but the relationship between GF104 and Cypress is quite similar to that between G92 (non b) and RV770: GF104 is bigger and slower.

Right now the HD 5800s are expensive, mostly because AMD is still supply-constrained, but if the GTX 460 starts to really eat into AMD's sales, they could easily drop the HD 5850 to $229 or so, which would make the GTX 460 1GB irrelevant, and the 768MB variant less than attractive.

Plus, Q4 should see Southern Islands launched.

As long as AMD doesn't suddenly develope an aversion to high margins and Nvidia doesn't release a product that poses a greater threat to 5850/5870, GF104 should continue to do well at its price point.

Right now the biggest potential threat that I can see for GF104 is whatever replacement AMD might have for the 57xx cards this fall. Is it going to approach GF104 perf? 58xx cards going down in price if a 68xx line is released isn't going to be much of a factor. Demand should quickly soak up any excess inventory of 58xx with a price reduction in that hypothetical case. And there'd be no reason to continue manufacturing 58xx if there's a 68xx replacement.

I think GF104 is going to be safe for a while. Hell, if you think about it. We all know GF104 can be clocked quite high. Enough that it can approach 5850 territory. However, if they were to have released at GF104 at those clocks it would have forced AMD to adjust prices.

Once they do that, there's a chance it might trigger yet another price war, one which would have the potential to drive down prices of both GF104 and 5850 below what GF104 currently sells at.

Now, look at Nvidia's Q3 filings. Not a particularly attractive situation to trigger a price war that could in the end result in selling GF104 at a lower price than it is currently selling at.

Right now it's a bit of a delicate balancing act at Nvidia to maximize ROI without triggering a price war which could make their situation far worse.

Regards,
SB

 

[Alexko]

Interesting. The idea that NVIDIA deliberately clocked GF104 pretty low to avoid triggering a price war hadn't occurred to me, but I admit it makes sense. On the other hand, there are quite a few overclocked GTX 460s offered by NVIDIA's partners, and they often offer much better value than both the standard GTX 460 and the HD 5850.

I wonder whether that might be enough to prompt AMD to lower their prices. Plus, if the GTX 460 manages to get enough market share for supply to surpass demand on AMD's side, I think a price drop really would happen. Perhaps not a big one, maybe just $20~30, but I don't see AMD just standing idly by while they lose market share in spite of their substantial performance/cost advantage. Besides, since Southern Islands is coming, they'll want to make sure that they don't have too much Evergreen inventory when it launches.