Nvidia Q4313 results

[mrmt]

Nvidia published their last results here:

http://phx.corporate-ir.net/phoenix.zhtml?c=116466&p=irol-IRHome

If you don't want to go through the numbers, they made record revenues, and are predicting a slump in Q114. PSB business went up again but Tesla is still a blip on their revenues. GPU finally is starting to slump and because of that they are expecting a an over 15% drop in revenues to Q114.

They refused to give yearly guidance and I doubt the reasons is because they can't really forecast the PC market, but because they cannot gauge (or don't want to disclose) the amount of damage Haswell will do to them. The fact they didn't provide guidance as they did in 2012 with IVB is an indicative that Haswell will have a sizable impact on dGPU sales, mainly on mobile. They are also expecting to gain even more market share from AMD, and looking at their margins, they have the strength for that.

One thing that caught my attention is that their MPU strategy with Tegra, focusing on the very high end only, is prone to boom and busts. 20-30% QoQ swings aren't out of the ordinary. They must somehow address that because if they miss a cycle their MPU business will be essentially zippo. As of now, they are expecting a decline in Tegra in Q114, the bust part, and are entirely dependent on tegra 4 to perform.

They are also with high hopes for their LTE modem, which if successful will give a steady income stream, offsetting the problem I pointed out in the last paragraph.

I couldn't fail to see a flair of rush in their call. While their current position is strong, they seem to be walking in very thin ice there. There are too many projects and too few results to speak off of those new products. I wonder if they would be better taking less fights but carrying a bigger stick to each of them.

 

[ShintaiDK]

Thats quite an impressive result.

Quarter Q4 FY'12 Q3 FY'13 Q4 FY'13
Revenue $953.2 million $1,204.1 million $1,106.9 million
Net income $116.0 million $209.1 million $174.0 million
Gross margin 51.4% 52.9% 52.9%

Year FY'12 FY'13
Revenue $3,997.9 million $4,280.2 million
Net income $581.1 million $562.5 million
Gross margin 51.4% 52.0%

 

[krumme]

Very nice but predictive results.

NV have made many good strategic decisions over the years, that last ones fx.:

- Going away from big die GPGPU intensive fermi, to lean Kepler. - while AMD going the opposite direction because of typical stupid technical engineering thinking.

- Tegra quad core on 40nm. A powerfull weapon on the marketing side because of quad core, and nice move because of the delayed 28nm. A brilliant decision.

- Optimus is what saved the low and mid end GPU market for notebooks.

I agree their current position is very strong, but that they are walking on thin Ice.

- 28nm is very cheap now, perhaps only 1/3 of one year ago. I think that pretty much outweight the Haswell GPU advantage. I think the last years have proven there is still a live for dedicated in notebooks. But ofcourse its a matter of time, but perhaps it will take longer than we expected? - optimus is a proven successfull technology, 28nm is cheap now, and i have problems seeing how Intel and AMD can fight those low cost.

- Its very hard to see high profit in the future, but NV have proven historically to be very innovative and extremely good on the market side; the LTE is a very strong addition, but they are in a very compettitive market. Probably they will invent somethink not on their current map. I think NV is good at walking on thin ice, and far better than Intel and AMD.

- I think its also important to stress the effect of AMD winning all the consoles. It could have very big effect on the long term. Perhaps someone could outline them?

 

[Acanthus]

Hard to increase dGPU revenue without releasing any new products.

I still think it's a bad decision to slow the product cycles.

 

[sontin]

They refused to give yearly guidance and I doubt the reasons is because they can't really forecast the PC market, but because they cannot gauge (or don't want to disclose) the amount of damage Haswell will do to them. The fact they didn't provide guidance as they did in 2012 with IVB is an indicative that Haswell will have a sizable impact on dGPU sales, mainly on mobile. They are also expecting to gain even more market share from AMD, and looking at their margins, they have the strength for that.

Makes no sense. nVidia is not expecting that the attach rate is shrinking.
What they can't predict is the overall market. Last year they gave a 2013 guidance which they needed to adjust because of the hard disk fallout and the 28nm supply.

One thing that caught my attention is that their MPU strategy with Tegra, focusing on the very high end only, is prone to boom and busts. 20-30% QoQ swings aren't out of the ordinary. They must somehow address that because if they miss a cycle their MPU business will be essentially zippo. As of now, they are expecting a decline in Tegra in Q114, the bust part, and are entirely dependent on tegra 4 to perform.

The whole segment is "entirely dependent on tegra 4 to perform" because it's the only product which matters.

I couldn't fail to see a flair of rush in their call. While their current position is strong, they seem to be walking in very thin ice there. There are too many projects and too few results to speak off of those new products. I wonder if they would be better taking less fights but carrying a bigger stick to each of them.

Less fights? Right now their Geforce business is at their highest point. Grow is only possible by taking market share. They have nearly 90% of the professional business with Quadro. Grow is only possible with Tesla and new software products for Quadro.

The only business which have a natural grow potential is Tegra. And in this business they increase their investment with Tegra 4, Grey, i500 modem and Project Shield.

There is not really anything new they can talk about: 2013 is a transistion year for them. 2014 is much more interesting with Project Denver, Grey, 20nm GPUs...

 

[krumme]

I still think it's a bad decision to slow the product cycles.

Its the only way, and it should have happened - at least for AMD - years back. What alternative is there?

The benefits - meaning technical benefit that translates into profit from the market - is shrinking. Doing more of the same, doest work. They need something new.

I think they have executed very well on the Tegra side, LTE is comming, but the basic problem is, there are other players, and they invest too. And it have proven difficult to leverage on the Gforce brand.

I think generally NV is doing extremely well, from the strategic very, very difficult position they have. Its easy to be a armchair CEO, but even then i dont have a f.....g clue what they should do

 

[mrmt]

- 28nm is very cheap now, perhaps only 1/3 of one year ago. I think that pretty much outweight the Haswell GPU advantage.

iGPU comes essentially for free as there is no mobile SKU without iGPU today, so no matter how cheap 28nm is there is no chance for them to beat iGPU on price.

What will dictate whether people will go dGPU or not is the performance levels of the iGPU. Why would you pay for something you can have for free? And this "free" threeshold is what dictates a dGPU purchase. With compelling enough performance consumers will give up dGPU for most of their purchases, and this is where I think we'll definitely go with Haswell. Nvidia might get a market niche or two when attaching dGPU with cheaper, fused off parts but the margins will be very low, as they will have the price difference between full and fused off SKUs to earn their COGS, OPEX and profits.

What we are starting to see here is tick-tock biting Nvidia in the same way it bite AMD. Kepler was released with IVB, now with HSW Nvidia will still have Kepler, and when Kepler successor arrives BDW will be around the corner, and you better pray for Kepler successor to be on 20nm because if it is 28nm things will get ugly fast, very fast.

- Its very hard to see high profit in the future, but NV have proven historically to be very innovative and extremely good on the market side; the LTE is a very strong addition, but they are in a very compettitive market. Probably they will invent somethink not on their current map. I think NV is good at walking on thin ice, and far better than Intel and AMD.

They are trying to diversify from the GPU market since the first Fermi and it was not for lack of trying that they didn't succeed. so far LTE, Tegra were nice additions but nothing of the killer market they need.

- I think its also important to stress the effect of AMD winning all the consoles. It could have very big effect on the long term. Perhaps someone could outline them?

I may be missing something but I don't see those winnings as a game changer, but as a steady low-margin revenue stream and nothing more. AMD had the Wii and Xbox 360 and it didn't bring anything out of the ordinary for them. Now with the consoles, why would it?

What codes run much better on Jaguar than on Core, or on GNC better than on Kepler? I doubt you won't find too many examples, and even when heavily optimized would those codes turn the balance in AMD's favor? But even if those codes could shift the balance, there is another possibility. Intel and Nvidia have both the resources to design better and more powerful chips than AMD, chips that could run AMD-optimized code better than AMD's own chips, and given that the consoles chips will be with us for a while you can expect those companies to follow this route if needed.

 

[mrmt]

Makes no sense. nVidia is not expecting that the attach rate is shrinking.
What they can't predict is the overall market. Last year they gave a 2013 guidance which they needed to adjust because of the hard disk fallout and the 28nm supply.

Where did you find that? They didn't give guidance on attach rates this call.

 

[sontin]

Where did you find that? They didn't give guidance on attach rates this call.

And they said not the opposite.
Look at their Opex. nVidia is investing in the future. So they expect that their revenue stream and margin will be stable. If they would fear that Haswell will cut into their Geforce business they would decrease their Opex.

 

[mrmt]

And they said not the opposite.

Okay, constant attach rates are your assumption, not based on something they said at the conference or anywhere else. Thank you for the clarification.

 

[Raghu]

Project Grey taped out in the 5th week of 2013 (or earlier?). Open "Fiscal 2013 Annual and Fourth-Quarter Results Presentation" document from this link

http://phx.corporate-ir.net/phoenix.zhtml?c=116466&p=irol-IRHome

Direct link to PDF

http://phx.corporate-ir.net/External.File?item=UGFyZW50SUQ9MTcxNjM4fENoaWxkSUQ9LTF8VHlwZT0z&t=1

Page 12 - Silicon picture against "Project Grey". Zoom in to see characters "ZS TAIWAN 1305A1". Thats 5th week of 2013 for A1 silicon.

 

[sontin]

It's not the tape-out date.
But i don't know if that is a real picture of Grey. nVidia is not really correct with such things.

 

[Piroko]

NV have made many good strategic decisions over the years, that last ones fx.:

- Going away from big die GPGPU intensive fermi, to lean Kepler. - while AMD going the opposite direction because of typical stupid technical engineering thinking.

Umm... They do have a 500mm² die of Kepler, but this time around they kept it to Tesla and Quadro lines early on (also releasing it last instead of first) instead of struggling with Desktop demand, yields and supply.
Also, how is 294mm² lean and 365mm² 'stupid technical engineering thinking'? That's not logical.

 

[pablo87]

Great results by NV. Now trading at a cash flow multiple of 5 net of cash. And seemingly continuing to invest in r&d.

 

[Genx87]

I was impressed in how they have grown their Tegra business. Up 50% YoY.

 

[krumme]

iGPU comes essentially for free as there is no mobile SKU without iGPU today, so no matter how cheap 28nm is there is no chance for them to beat iGPU on price.

What will dictate whether people will go dGPU or not is the performance levels of the iGPU. Why would you pay for something you can have for free? And this "free" threeshold is what dictates a dGPU purchase. With compelling enough performance consumers will give up dGPU for most of their purchases, and this is where I think we'll definitely go with Haswell. Nvidia might get a market niche or two when attaching dGPU with cheaper, fused off parts but the margins will be very low, as they will have the price difference between full and fused off SKUs to earn their COGS, OPEX and profits.

What we are starting to see here is tick-tock biting Nvidia in the same way it bite AMD. Kepler was released with IVB, now with HSW Nvidia will still have Kepler, and when Kepler successor arrives BDW will be around the corner, and you better pray for Kepler successor to be on 20nm because if it is 28nm things will get ugly fast, very fast.


.

How long time did you have Intel GFX for free for the notebooks. A decade now? and still people buy seperate gfx, even though its just twice as fast. It beats me, but they do. I dont see how Haswell should be a gamechanger in that matter, excepts its more expensive to manufacture if its going to be faster than IB. Same.

If everything was going to be so ugly for Intels competitor i am sure i would be reflected in the share price. They still have a market cap like qualcomm. Intel might potential be more worth, but sure its very risky to invest in that tractor.

 

[ShintaiDK]

If everything was going to be so ugly for Intels competitor i am sure i would be reflected in the share price. They still have a market cap like qualcomm. Intel might potential be more worth, but sure its very risky to invest in that tractor.

Its actually Qualcomm thats the risk company. Since its currently massively overvalued.

 

[krumme]

Umm... They do have a 500mm² die of Kepler, but this time around they kept it to Tesla and Quadro lines early on (also releasing it last instead of first) instead of struggling with Desktop demand, yields and supply.
Also, how is 294mm² lean and 365mm² 'stupid technical engineering thinking'? That's not logical.

Why deviate from the 4x-6x strategy that gave them high marketshare?

Why on earth put loads of DP computing power into the cards, when practically noboby buys them for that reason?

What do a buyer of a gfx card care about future fusion strategies?

 

[krumme]

Its actually Qualcomm thats the risk company. Since its currently massively overvalued.

Please share some of your wisdom, the rest of the stupid market does not understand.

 

[ShintaiDK]

Please share some of your wisdom, the rest of the stupid market does not understand.

It doesnt seem like you wish to understand in the first place, with a comment like that. So why waste one anothers time.

But I assume you know that stocks reflect future expectations.

 

[Idontcare]

Why deviate from the 4x-6x strategy that gave them high marketshare?

Why on earth put loads of DP computing power into the cards, when practically noboby buys them for that reason?

What do a buyer of a gfx card care about future fusion strategies?

Is that really any different than the justification for adding ISA extensions in CPUs. Who cares about SSE2 the day the first processor with SSE2 hits the market? No one really except a handful of programmers. But a couple years later, as the market is seeded with the hardware, people do care.

Please share some of your wisdom, the rest of the stupid market does not understand.

"Market Correction" do happen, and they happen for reasons the rest of the "stupid market" failed to foresee. You can't hold up "the market" as some infallible benchmark.

QCOM is over-priced, waiting to crash, right?

APPL was P/E = 16 in Sept, then "crashed" to P/E = 11. More like a bubble that shouldn't have existed in the first place has been corrected.

QCOM P/E = 17 right now. At some point that P/E is going to become ~11.

 

[mrmt]

How long time did you have Intel GFX for free for the notebooks. A decade now? and still people buy seperate gfx, even though its just twice as fast. It beats me, but they do. I dont see how Haswell should be a gamechanger in that matter, excepts its more expensive to manufacture if its going to be faster than IB. Same.

If you look at the TAM for AMD and Nvidia in 2008 and look for the TAM now, the consumer GPU market is far smaller today than it was four years. Add AMD and Nvidia consumer sales today and you'll have a number smaller than Nvidia consumer sales on that time. So in a sense, the game is changing. Slowly, but definitely changing.

Make no mistake, if Nvidia was in front of a booming GPU market, why would they spend millions to build a GPGPU market or throw themselves in the middle of a bunch of ARM manufacturers? Nvidia burned their margins to make GPGPU happen with Fermi and right now they are bleeding money and cash to make Tegra happen. The only plausible reason for them to do that is that the returns on those markets, even when weighted against the high risks are bigger than whatever returns they can have in GPUs.

I don't expect Haswell to wipe Nvidia out, but I do expect Haswell to make a sizable dent on Nvidia's revenues, a dent big enough to make them to not disclose their financial guidance for the year.

If everything was going to be so ugly for Intels competitor i am sure i would be reflected in the share price. They still have a market cap like qualcomm. Intel might potential be more worth, but sure its very risky to invest in that tractor.

The next two years won't be too happy for Intel shareholders. They have a lot of CAPEX commitments and they need to carve market share in mobile. But I wouldn't put my money on Qualcomm either. Qualcomm doesn't have the fundamentals for a 100 billion CAPEX.

 

[cytg111]

How long time did you have Intel GFX for free for the notebooks. A decade now? and still people buy seperate gfx, even though its just twice as fast. It beats me, but they do. I dont see how Haswell should be a gamechanger in that matter, excepts its more expensive to manufacture if its going to be faster than IB. .

Haswell wont, as current hd3000/hd4000 didnt. Consoles did. Specifically console ports, as they set the standard. Ie. my daughter is running LoL damn decent on an i3, no descrete. Not possible pre SB.

 

[krumme]

Is that really any different than the justification for adding ISA extensions in CPUs. Who cares about SSE2 the day the first processor with SSE2 hits the market? No one really except a handful of programmers. But a couple years later, as the market is seeded with the hardware, people do care.

I think the major difference is ISA extensions in the CPU doesnt take nearly the space adding fx. DP to a GFX does. The die cost far outweight the benefits, the cards will be changed before people care. The added cost to the GPU exactly takes the profit out of the products. Control of marginal cost is what drives your business.

"Market Correction" do happen, and they happen for reasons the rest of the "stupid m

Completely agree. But comments like "Its actually Qualcomm thats the risk company. Since its currently massively overvalued" adds absolutely no argument to the table, and is just part of this childish NV, AMD, Intel nonsense fight.

 

[krumme]

Haswell wont, as current hd3000/hd4000 didnt. Consoles did. Specifically console ports, as they set the standard. Ie. my daughter is running LoL damn decent on an i3, no descrete. Not possible pre SB.

I agree, if there was a significant change is was the jump to SB. But the point was, the transistion to no midrange GPU in notebooks took longer than - at least i - expected. We even see AMD put separate GPU in GPU strong APU notebooks. It means there is still a market 2013 for NV. I think its also about habit. People still think they need a separate GFX because of the early Intel solutions, and that habit takes years to change. But ofcource its a matter of time before it ends.

But adding the fact that people still buys GFX for the notebooks (take nv 620m / now 630m as the most incredible examples - why peoply buy them is beyond me), and that habit takes years to change, i still think there is a market for NV here. But mrmt noticing NV dont have financial guidance for it, shows its a matter of time. I just think we have at least 2 years left for the midrange.