NVIDIA: Record Q3 + $1.85B in cash

Arun

Unknown.
Moderator
Legend
Non-GAAP EPS: $0.44 (vs $0.37 estimate)
Revenue: $1.12B (vs $1.01B estimate)
Non-GAAP Net Income: $264M
Cash flow: $278M
Cash: $1.85B

CC Transcript said:
[Jen-Hsun]
- Record revenue, record net income and record gross margins.
- First billion dollar quarter; grew 19% sequentially and 36% from last year.
- 46.2% record gross margins.

- Remains one of the fastest growing semiconductor company in the world.
- Record desktop and notebook revenue. Standalone notebook GPU share grew to 72%.
- Shipped first Intel IGP, approximately 40M addressable units next year.
- Launched 8800 GT, best performance and best HD video. Comparable to Ti4200.
- Workstation had record revenue, several new product launches.
- Believe the CAD market will continue growing etc.
- Began shipments of the Tesla GPUs. [insert CUDA marketing]

[Marvin]
- Only difference between GAAP and non-GAAP is stock compensation and its tax effect.
- Concerns never materialized, very strong quarter.
- Desktop GPUs grew 14% sequentially.
- MCP business grew 26% sequentially, beginning to see Intel IGP traction.
- Delivered higher margins mainly thanks to GF8, PS3 royalties and workstation.
- Had to accrue a one-time catch-up because of higher full-year estimates.
- 195(?) out of 2xx (?) of the new employees are in engineering (couldn't hear).
- New testers due to higher production and bought land for $27M.
- Cash ended at $1.85B despite buybacks of $185M (not sure?) during the quarter.

- On the outlook, we currently see strong demand for our products.
- Revenue growth within the 5-7% range is to be expected.

- For gross margins, we made excellent progress and will continue to work on it.
- In summary, we expect an excellent Q4.

[Jen-Hsun]
- [marketing about GPUs, video game market, Vista and Google Earth, etc.]
- [marketing about the importance of the GPU driving growth]
- We believe the PC game market is continuing to grow.
- [huge growth in the 'enthusiast' market, but undefined so hard to judge]
- [marketing about Intel IGPs and Hybrid SLI market opportunity]
- [marketing about ESA and how it advances the capabilities of the PC]
- Growth of workstations is driving the growth of Quadro. China and India opportunity.
- [marketing about Tesla]
- CUDA will enhance video, image and physics processing for consumers.
- We expect our next-generation Application Processor to sample at the beginning of next year.
- [marketing about PC-like handhelds]
- AP15 (codename of their next-gen) is the the combination of several hundred manyears of R&D.

[Q&A]
Q: Congrats. GeForce 7000 motherboard GPU; what kind of ASP are we seeing there on the Intel side?
A: MCP73 ASPs are higher than MCP68. Targeted at the mainstream price segment.
ASPs are intended to be in the teens to maximize the addressable market.

Q: What do you expect for notebooks? vs IGPs? Market share in general?
A: Hard to say what our market share will be in the future.
We have been competing with IGPs for quite a long time, and our ability to succeed will depend on coming up with great ideas like SLI and Hybrid SLI. If we do our job well, we will continue to grow, otherwise our competition will grow.

Q: Power consumption?
A: How is their integrated graphics lower power than ours? [marketing about G8x power efficiency]

Q: Update on double-ordering, too good for too long, backlog, etc.?
A: The strength of the orders in Q2 continued straight through Q3. There was no change in orders or shipments.
You're always concerned about double-ordering, but we don't see the evidence of that.
[Jen-Hsun]: We don't know what we don't know, but the marketplace is rather robust. For us, new products always help. 8800 GT is arguably our best product in 5 years, very very big hopes for it. MCP73 is our first product of that category, wonderful adoption.
[Marketing about China, workstations, importance of the GPU, Tesla, etc.]

Q: What areas are capacity-limited?
A: MCP73 is tight, some GF8s are still capacity-limited, or rather production-limited.


Q: Q4 guidance relating to share gain etc.?
A: We don't count on share growth, that happens after you're done counting. We don't typically count that. We look at our strategic position and the new products we're introducing, etc. - but we try not to be too bullish on share gains.
[Marvin] Also not expecting ASP growth.
[Jen-Hsun] So expecting robust demand and entering new markets, such as Tesla.
8800 GT is an existing market that we serve but it's very disruptive, so it promotes very very quick adoption.

Q: Motherboard GPU replacing discrete?
A: GF7 IGP isn't fast enough to replace even our entry-level discrete GPUs. So that's really targeting the segment underneath that we aren't serving today.

Q: 8800GT - great product, but do you think it might have an effect on other products? Gross margin effects?
A: We always try to be thoughtful about cannabilizing our own products. But if we can do that, we'd rather do it ourselves than somebody else. We want to remain aggressive. I am not inclined really to slow down; my own business will have to figure out how to deal with my own business.
[Marvin] Margins on the 8800GT are excellent, well above corporate margins.

Q: Tesla applications for computational finance etc.
A: Very big opportunity for finance. Very very large step-up in performance, quite a bit of development in that area. Development is happening in London, in New York, etc. - very interesting for GPU Computing.

Q: G92 cross-over relative to G80?
A: Uhm. Cross-over. Let me explain it that way. We will ship more 8800 GTs this quarter than we shipped 8800 GTSs last quarter.

Q: Most aggressive ramp in history?
A: This is one of our most aggressive ramps, since it's such a great product. Really rich backlog of orders for it. And although the market now knows what the competition has, they're voting with their orders. We're going to be very very aggressive with our products. [insert marketing about 8800GT vs Ti4200]

Q: Think you can take more market share?
A: There's always competition in our space. It's one of the most vibrant industries. The reason for that is Moore's Law is still GPU's friend and vice-versa. One of the few technology segments where more transistors can still deliver better experiences. Other markets are seeing smaller die sizes and ASPs, but not the case for GPUs.

Q: OEM vs Channel? (for something specific? if so couldn't hear and Jen-Hsun didn't repeat)
A: We are expecting to be very successful with both OEMs and the channel. You'll see it soon.

Q: Notebook market has been doing very well, expectations regarding replacing desktops etc.?
A: The notebook market is certainly doing very well. I expect a lot more growth in the mobile gaming and mobile workstation segments, and we expect to be able to add a lot of value to these two platforms in the future. While people tend to have 1 or 2 PCs at home, they can have maybe 4 laptops. I have 8. So I think it's not completely a replacement market, it's additive.

Q: Notebook market is more competitive because of Intel bundling?
A: Intel is definitely a very competitive company, and we've competing against their bundling for quite some time. The bottom line is we've just got to add value. NVIDIA is not a commodity players. We serve the marketplace where visual computing really matters, if 3D matters to you... then we can add a lot of value to you.

Q: ASPs, on a mix-basis were ASPs able to improve from Q2 to Q3?
A: No, they were relatively flat.

Q: Margins were up, you talked about growth from Sony being a driver? Could you help us understand the impact of that?
A: Any increase in royalties is 100% gross margins. I would say PS3 royalties more than doubled from Q2 to Q3. Was it the driving factor? Probably not, but it was one of the factors that improved gross margins. It is probably true that it accounts for most of the increase in the consumer business.

Q: Professional business? Handset strategy, impact of a lack of baseband solution?
A: On the first one, we really haven't and we don't give too much guidance on that level.
All I can tell is it's been doing very well and we add a lot of value to that business.

With respect to our AP business, we are going to focus on the segments of the market place where computing is important. If you look at all the handhelds in the world, everyone wants to be a computer but besides the iPhone, everyone has a very long way to go. We can add a lot of value, and we are seeing a lot of interest in the work we are doing.

Q: Mainstream growth/low-end, growth domestic vs overseas? China and India? Secondly, current mobile phone business vs expectations?
A: Developing markets are being driven by online games, really a craze in China and India. Games are free and you pay for the service and as you go. Real growth driver there. Their purchasing power is improving too, and GPU is the best way to improve your experience for a small premium, irrespective of Vista. Vista is doing very well everywhere, not just domestically.

With respect to the handheld business, we haven't shipped many application processors, and haven't shipped any AP15s yet. Revenue for non-APs is likely to decline, we aren't really focusing on getting more design wins there.

Q: Gross margins in the next 12 months?
A: We have not raised our margins by raising prices, and the reason for that is it's hard to grow while raising your prices. Operational performance is where gross margin improvements can come from, and I still have a lot of work to do there. Execution etc.
[Marvin] Mix can improve margins, but we want to improve gross margins in all segments too.

Q: Professional services, ASPs for the quarter? Tesla contribution?
A: Tesla was just beginning to ship, expecting it to be small for the first year relative to the size of the company. With CUDA, we hope to bring GPU Computing to the masses though.
[Marvin]ASPs are in the hundreds of dollars. Complex mix I won't go into the details. Relatively flat over Q2.

Q: Expectation for ASPs over time? Tesla?
A: The more programmable we make the GPU, the more valuable we make it, the more people will be willing to pay for it. This has been the case in the last few quarters. I expect further value increases thus

Q: Other enthusiast refreshes?
A: We are *at* the end of the year, and we decided that the refresh we'd do at this time is 8800 GT. From 8800 GT, you could obviously do SLI. And soon you'll be able to do tri-way SLI. You're going to be able to put a lot of GPU horsepower in your system.


Q: Havok bought by Intel, are you in the market for that kind of IP?
A: Uhh, uh, uh no current plans.

Q: Repeat.
A: [...] Still a lot of innovation to happen in that market, possible negative synergies [...] I don't know why they did it.

[Jen-Hsun ending comments]
 
From CNNMoney

Third Quarter Fiscal 2008 and Recent Highlights:

-- As the first GPU company to reach the $1 billion quarter level,
NVIDIA continues to be one of the fastest growing semiconductor
technology companies in the world.

-- The GeForce(R) desktop and notebook GPU product lines each achieved
record revenue. The desktop GPU product line grew 33 percent
year-over-year and the notebook GPU product line grew 120 percent
year-over-year.

-- NVIDIA shipped its first-ever single-chip motherboard GPUs (mGPUs)
for Intel-processor-based desktop PCs - the NVIDIA GeForce 7000 mGPU
family. The GeForce 7000 mGPU family delivers the performance of an
entry-level discrete GPU when compared against traditional
integrated graphics solutions.

-- NVIDIA launched the new GeForce 8800 GT GPU, which combines
best-of-class gaming performance and HD playback at the $199 price
point.

-- The Quadro(R) Professional Solutions Business achieved record
revenue, growing 37 percent year-over-year. NVIDIA launched seven
new Quadro solutions during the third quarter of fiscal 2008, most
notably the Quadro FX 370 and 570, which deliver ISV-certified
hardware at the sub-$200 price point.

-- NVIDIA began shipments of the Tesla(TM) C870 GPU computing processor
and D870 desk-side supercomputer products during the third quarter
of fiscal 2008. The adoption of Tesla and the CUDA C-language
programming environment and tool suite continues to accelerate.
 
But Doug Freedman said that their revenue peaked last quarter. An he is a professional analyst! Of course, he also downgraded the stock back in January when it was at ~$22! And he is the professional analyst!!!! And he used such great arguments to justify his positions as "sources inside AMD" and "RV670 is clocked at 825MHz!!!!! PWMN. 8800GT is 600MHz WTF lame!" Did I mention he is a professional analyst?
 
Just posted the transcript, enjoy. Adding bolding now...
EDIT: Done. I didn't bold much, since I'd rather emphasize what really does matter.
 
A couple of quick points:
1) Jen-Hsun doesn't seem to fully understand why Intel bought Havok. In case anyone from NV with Jen-Hsun's phone number is reading this, the answer is risk management for Larrabee. And for the same reason, Ageia would be a pretty good buy for less than $50M.
2) 8800 GT gross margins are 'well above the corporate average'. What can I say but wow? Given the die size, NV's yields must be amazing. I wonder how they're counting things though, since there will be other G92 SKUs coming from the same wafers eventually.
4) MCP73 outlook looks very good, seems to hint at good Holiday cycle OEM deals IMO.
3) NVIDIA's AP strategy beyond their initial product remains vague.
 
Arun, you should bold the outlook for next quarter, because that's what's going to drive the stock back to 37+ tomorrow.
 
Last edited by a moderator:
Heh, I thought 'excellent quarter' was enough, but fair enough! ;)

Interestingly, that's revenue guidance, so profit growth will be even better. As for margins, MCP73 shipments will go up, but professional looks to be growing in Q4 too IMO. So with that change in mix, it'd have been handy to know the gross margins on MCP73 - ah well! Probably above 35% anyway, so not going to be a significant drag. 8800 GT margins are also out of this world, sooo!

I cannot help but feel NVIDIA is now in an excellent position to go on an acquisition frenzy. Their growth in their core business in 2008 will likely be lower, so it's the ideal time to spend that extra cash and diversify further through strategic acquisitions, such as:
- VIA (CPU/PC Analogue/etc.)
- Clearspeed (HPC Solutions)
- Ageia (Physics/Consumer GPGPU)
- Icera (Programmable Low-Power 3G+ Baseband)
- Anything else to get Bluetooth/WiFi/FM IP and engineers.
- Wireless USB IP/engineers if they don't already have those.

My estimations indicate that it should be possible to get *all* of the above for only slightly more than $1B in cash ($600M for VIA, $50M for Ageia, $150M for Clearspeed, $250M for Icera, less than $50M for other wireless). The most dubious of those is Icera, which they might in fact have to pay even more for. 3Plus1 is a decent (but more exotic) alternative, I guess.

Of course, that doesn't mean that is NVIDIA's current strategy. It is certainly the strategy I would deploy, but hey, I don't have a strategic analysis job at NVIDIA (although I wouldn't deny I'd quite enjoy that) so that doesn't really matter.

I stick to my prediction of NVIDIA acquiring VIA within the next 6 months (my cost estimate substracts VIA's current cash assets), but everything else is just wild speculation at best, obviously.
 
I think G92 VS RV670 margins will be interesting to watch. On one had, AMD is going to collect ~40% more chips per wafer. On the other hand, I wonder what a 55nm wafer costs compared to what by now is tried-and-true 65nm and what kind of yeild they will be seeing.
 
BTW, am I the only one who is reading (too much) into the overboard suck up campaign re. Apple? Reminds me of them inexplicable starting to lavishly praise a different company few years back, followed a few months later by a big announcement.
 
A couple of quick points:
1) Jen-Hsun doesn't seem to fully understand why Intel bought Havok. In case anyone from NV with Jen-Hsun's phone number is reading this, the answer is risk management for Larrabee. And for the same reason, Ageia would be a pretty good buy for less than $50M.
2) 8800 GT gross margins are 'well above the corporate average'. What can I say but wow? Given the die size, NV's yields must be amazing. I wonder how they're counting things though, since there will be other G92 SKUs coming from the same wafers eventually.

1. I am not sure JHH was being entirely literal when he said he didn't know. I think he was diplomatically trying to say he wasn't sure it was a great move by Intel. Probably for the much the same reasons that the Renderware purchase by EA did not really pan out so great.

2. Memory prices are probably low right now, so that helps with board costs. So they can probably still charge a pretty decent price on the 8800 GT. And anytime they can charge a pretty decent price i think they are going to have higher than average corporate margins.
 
rofl...
Analyst Doug Freedman of American Technology Research said Nvidia's numbers are "good," but he said he is "still not sure what share shifts will happen next year."

Not sure, are you Doug? Strange, you were pretty damn sure last week when you told your clients to sell the stock since the revenue "peaked" the previous quarter.
 
1. I am not sure JHH was being entirely literal when he said he didn't know. I think he was diplomatically trying to say he wasn't sure it was a great move by Intel. Probably for the much the same reasons that the Renderware purchase by EA did not really pan out so great.
Fair enough, but then I must respectfully disagree with Jen-Hsun: Intel's acquisition of Havok was an absolutely splendid idea IMO, and a brilliant strategic decision given their position.

It gives them the capability to justify Larrabee and raytracing even before anyone uses it for rendering, if ever. Fundamentally, it gives them the opportunity to position Larrabee as a 'game accelerator' rather than a 'graphics accelerator', and leverage non-graphics areas as a trojan horse to interest developers.

That doesn't mean it will be enough to make Larrabee a success, and I still don't think it will be outside of HPC and a small part of the workstation market (even for the 'second gen'). But this certainly improves their chances to be successful overall, and even if they aren't it's likely to minimize how bad the failure is.

So from a risk management perspective (given Larrabee is likely a much larger investment than that acquisition!) it looks really good to me. Larrabee's architecture is also uniquely suited for physics, given their hybrid MIMD-SIMD scheme. This is something other GPUs won't have for quite some time.

2. Memory prices are probably low right now, so that helps with board costs. So they can probably still charge a pretty decent price on the 8800 GT. And anytime they can charge a pretty decent price i think they are going to have higher than average corporate margins.
Yes, it looks like GDDR3 yields and clocks are now very good, and DDR2 pricing must be having consequences in GDDR3 pricing too. I wouldn't be surprised if the premium for 900MHz GDDR3 vs 700MHz GDDR3 wasn't very big anymore.

However, there is more to it than that. AMD likely isn't getting higher margins on RV670, and yet their die size is much much smaller. I'm starting to think their yields must be downright amazing, which makes me suspect they might have further improved their redundancy mechanisms for G9x, compared to G8x. We'll never know for sure though, I suppose. Also, it's worth considering their margin calculations likely *indirectly* consider future SKUs.
 
Just noticed something funny.
In recent years (from what i recall, anyway), there was never any GPU from Nvidia that didn't started its commercial life with the fully working part being launched either first or simultaneously with the cutdown part (6800 non-GT/GT/Ultra, 7800 GTX/GT, 7900 GT/GTX/GS, 8800 GTS/GTX/Ultra, etc).

Yet the G92 core had its debut precisely with a -supposedly- lesser variant this time.
Lessons from Cell ?

Did they "plan" to achieve this kind of yield in detriment of the top-end right from the start, or was it just pure luck with TSMC's process this time ?


Oh, and does the continuous insistence on Nvidia's CC, comparing the 8800 GT with the Ti 4200 (besides the obvious similar price/performance level), mean we get to see two more high performance parts out of G92 until it's retired or what ?
You know, there were the Ti 4400 and the Ti 4600 above it... ;)
 
Oh, and does the continuous insistence on Nvidia's CC, comparing the 8800 GT with the Ti 4200 (besides the obvious similar price/performance level), mean we get to see two more high performance parts out of G92 until it's retired or what ?
Heh, funny way to look at things! ;) My most recent analysis based on clocks and SKU differentiation requirements gives me a GTS at ~280 and a GTX at ~330 (~390 for 1GiB). I would also expect at least another 'niche' high-end SKU, an Ultra and/or a GX2.

There's definitely room to get the single-chip G92 at least up to $379, though. You'd be naive to think otherwise, IMO. As for going down, it's not hard to imagine a ~$150 192-bit SKU. Anyway, let's not go OT, there are other threads for that...
 
I am telling you guys, something is up with Apple. If I am wrong, I will eat one*.





Granny Smith or Red Delicious, your choice.
 
A couple of quick points:
1) Jen-Hsun doesn't seem to fully understand why Intel bought Havok. In case anyone from NV with Jen-Hsun's phone number is reading this, the answer is risk management for Larrabee.

I always thought it was to establish a already proven devloper relations program to go head to head with the likes of TWIMTBP.
 
I always thought it was to establish a already proven devloper relations program to go head to head with the likes of TWIMTBP.

Even in the context of current CPU development, those extra cores need to go to use. And "Good physics, only on Core 3!" makes for a pretty nice tag line.
 
But Doug Freedman said that their revenue peaked last quarter. An he is a professional analyst! Of course, he also downgraded the stock back in January when it was at ~$22! And he is the professional analyst!!!! And he used such great arguments to justify his positions as "sources inside AMD" and "RV670 is clocked at 825MHz!!!!! PWMN. 8800GT is 600MHz WTF lame!" Did I mention he is a professional analyst?

LMAO, remember this?

http://forum.beyond3d.com/showpost.php?p=1083833&postcount=665

Freedman knows shit about gfx chips.


You hear me Sound_card? ;)
 
And for the same reason, Ageia would be a pretty good buy for less than $50M.

I think you're very alone with that opinion. Ageia is a turd.

EDIT: Geeforcer, what about Apple?
 
Last edited by a moderator:
I think you're very alone with that opinion. Ageia is a turd.

EDIT: Geeforcer, what about Apple?

Re. Ageia: I think the advent of pervasive multi-core CPUs have renders them largely irrelevant as far as their hardware is concerned.

Re. Apple: Jen and the rest of the Nvidia people just could not shut up about how great iPhone is and how nice Leopard is, etc, which in my off-hand estimate represents an infinity percent increase in the number of times Apple has been mentioned during any of the previous CC in recent memory. Remember how Jen all of the sudden started gushing over Sony and PS3 and then a few months later RSX contract was announced? If this is not just a strange coincidence, I think they have picked up a design win or two on that front.
 
Back
Top