Thursday, November 10th 2016

NVIDIA Announces Financial Results for Third Quarter Fiscal 2017

NVIDIA (NASDAQ: NVDA) today reported revenue for the third quarter ended October 30, 2016, of $2.00 billion, up 54 percent from $1.30 billion a year earlier, and up 40 percent from $1.43 billion in the previous quarter. GAAP earnings per diluted share for the quarter were $0.83, up 89 percent from $0.44 a year ago and up 102 percent from $0.41 in the previous quarter. Non-GAAP earnings per diluted share were $0.94, up 104 percent from $0.46 a year earlier and up 77 percent from $0.53 in the previous quarter.

"We had a breakout quarter - record revenue, record margins and record earnings were driven by strength across all product lines," said Jen-Hsun Huang, founder and chief executive officer, NVIDIA. "Our new Pascal GPUs are fully ramped and enjoying great success in gaming, VR, self-driving cars and datacenter AI computing. "We have invested years of work and billions of dollars to advance deep learning. Our GPU deep learning platform runs every AI framework, and is available in cloud services from Amazon, IBM, Microsoft and Alibaba, and in servers from every OEM. GPU deep learning has sparked a wave of innovations that will usher in the next era of computing," he said.
Capital Return
During the first nine months of fiscal 2017, NVIDIA paid $509 million in share repurchases and $185 million in cash dividends. As a result, the company has returned an aggregate of $694 million to shareholders in the first nine months of the fiscal year. The company intends to return $1.0 billion to shareholders in fiscal 2017.

For fiscal 2018, NVIDIA intends to return $1.25 billion to shareholders through ongoing quarterly cash dividends and share repurchases. The company's board of directors has authorized an additional $2.00 billion under the company's stock repurchase program for a total of $2.96 billion available through the end of December 2020. The company announced a 22 percent increase in its quarterly cash dividend to $0.14 per share from $0.115 per share, to be paid with its next quarterly cash dividend on December 19, 2016, to all shareholders of record on November 28, 2016.

NVIDIA's outlook for the fourth quarter of fiscal 2017 is as follows:
  • Revenue is expected to be $2.10 billion, plus or minus two percent.
  • GAAP and non-GAAP gross margins are expected to be 59.0 percent and 59.2 percent, respectively, plus or minus 50 basis points.
  • GAAP operating expenses are expected to be approximately $572 million. Non-GAAP operating expenses are expected to be approximately $500 million.
  • GAAP and non-GAAP tax rates for the fourth quarter of fiscal 2017 are both expected to be 20 percent, plus or minus one percent.
Third Quarter Fiscal 2017 Highlights

Gaming:
  • Announced that NVIDIA gaming technology will power the Nintendo Switch home gaming system.
  • Expanded its line of Pascal GPUs with GeForce GTX 1050 and GTX 1050 Ti, letting new gamers discover the joy of GeForce PC gaming.
  • Introduced GeForce GTX 1080, 1070 and 1060 for notebooks, giving gamers a state-of-the-art gaming platform in beautifully designed notebooks.
Datacenter:
  • Expanded the GPU Technology Conference with a world tour of eight cities that broadened its reach this year to 18,000 developers, researchers, scientists and others.
  • Launched Tesla P40 and P4 GPUs, and the NVIDIA TensorRT deep learning inferencing framework. These expand NVIDIA's deep learning platform beyond training to speed up AI inferencing production workloads in hyperscale datacenters.
  • Began shipping the NVIDIA DGX-1 AI supercomputer to research organizations, including OpenAI, Germany's DFKI and Switzerland's ITSIA; to universities, including Stanford, New York University and UC Berkeley; and to multinationals, such as SAP.
  • Announced a collaboration with Japan's FANUC to implement AI to increase robotics productivity and bring new capabilities to automated factories.
Automotive:
  • Announced that its NVIDIA DRIVE PX 2 platform will power a new AutoPilot system in all of Tesla Motors' factory produced vehicles - the Model S, Model X and upcoming Model 3.
  • Unveiled its next-generation Tegra processor, codenamed Xavier, an AI supercomputer on a chip for self-driving cars.
  • Partnered with China's Baidu to develop a self-driving, artificially intelligent car and mapping system.
  • Announced an AI partnership with Europe's TomTom to create a cloud-to-car mapping system for self-driving cars using NVIDIA DRIVE PX 2.
Add your own comment

39 Comments on NVIDIA Announces Financial Results for Third Quarter Fiscal 2017

#26
Finners
The quote says it all for me "record margins" this is what has stopped me buying a 1070/80

Dont get me wrong Nvidia owe us nothing and are a business but I begrudge paying lots of money for something that is cheap to produce
Posted on Reply
#27
BlueFalcon
SlizzoGoing to trot out my favorite thing to state when people complain about new pricing vs. old.

In 2014, the MSRP of the GTX980 at launch was $549. Adjusted for today, that's just a shade under $560 ($559.88 to be exact).

Launch MSRP of the GTX1080 was $599.

Increase of ~$40. Not exactly earth shattering.
Except this analysis is entirely flawed for 3 reasons.

The first flaw is you are ignoring that NV raised prices 1.5-2X per each caterogy tier by re-labelling lower tiers as higher tiers, and then pushing back the true flagship towards the 2nd half of a generation. This started not during Maxwell, but during Kepler generation. What used to be $199-249 GTX560/560Ti GF114 tier became $399/$499 GTX670/680 GK104 tier. What used to be a cut-down flagship $289 GTX560Ti 448 Core and $349 GTX570 GF110 became $649 GTX780. This trend can be clearly illustrated today with $129 GTS450 GF106 or $149 GTX550Ti GF116 replaced by $199-299 GTX1060 3GB/6GB GP106, and with $249 GTX560Ti replaced by $599-699 GTX1080.

The second flaw in your post is complete ignorance of Kepler generation and the x70 tier by simply comparing 1080 Pascal to 980 Maxwell. Even if we ignore the fact that NV increased prices 1.5-2X starting with Kepler after Fermi, NV still managed to raise prices this round significantly. GTX680 debuted at $499 while GTX1080 costs $599-699. It took at least 6 months for GTX1080 to drop down to $599 as most AIB cards cost $650 or more during the first 4 months of release. Mere months post GTX970's launch, it was easy to find $330-340 AIB GTX970 but 2-3 months post GTX1070 launch, it was not possible to find most AIB 1070 cards below $420.

The comparison of GTX1070's pricing to GTX970 isn't also complete without factoring in that GTX1070 is a more cut-down part compared to the GTX970 or GTX670 during their respective generations. GTX1070's specifications and performance delta compared to GTX1080 more closely parallel the $299 GTX660Ti in the Kepler generation. When looking at it from this perspective, it can be easily argued that GTX1070 is just a $299 GTX1060Ti priced at $379-449. GTX680 was roughly 25% faster than GTX660Ti and GTX1070 is about 22% faster. GTX670 and GTX970 were much closer to the performance of the GTX680 and GTX980. You can try to argue that GTX1080 is now priced much higher than GTX680 but then you precisely fall into your flaw of arguing that NV didn't raise prices. What NV has done is actually raised prices on both the GTX1070 and GTX1080 cards because GTX1070 is not a proper x70 series card, but one of the worst cut-down x70 cards since GTX470. At least the GTX470 was cut-down from a flagship die.

The current Titan X Pascal is not much different from a GTX570 2.56GB (aka a cut-down flagship with double the VRAM of the consumer model). Last generation's GTX980 cost $649 and that was a cut-down flagship, much in the same way $349 GTX570 was a cut-down flagship. That means $649 GTX980's predecessor cost just $349. Should NV price 3328-3456 CUDA core 1080Ti at $699 or above, it will be yet another successor to the GTX275/GTX570 tier, but priced at 2X or greater of those predecessors.

The third flaw in your post is complete ignorance that NV's gross profit margins are approaching 60%. Prior to 2012, NV's gross margins were usually between 30-48%.


www.fool.com/investing/general/2016/03/30/nvidia-corp-stock-in-3-charts.aspx

Looking at NV's gross margins from 1999-2011 reveals an even more startling picture of gross margins. From 1999-2005, NV's gross margins never exceeded much more than 40%, with 4 of 7 of those years not even hitting 40%. From 2005-2011, not once have NV's annual gross margins even reached 50%, with years 2006-2007 and 2009-2011 staying at 47% or lower. Since 2012, NV's gross margins as if by magic went to > 55% and have only risen thereafter:
www.wikinvest.com/stock/NVIDIA_(NVDA)/Data/Gross_Margin

It can be easily concluded that NV has effectively raised GPU prices for every single dGPU tier since 2012 and used marketing names to obfuscate this fact. NV is just doing what a corporation should be doing and that is raising prices and seeing just how much the consumers are willing to pay. The introduction of the FE tax and further raising prices during Pascal generation highlights the ceiling hasn't been reached yet and we should expect even higher prices during Volta generation. The days of NV launching true flagship cards during the 1st half of a generation are over.

As a bonus point, going back to GeForce 2 and until Fermi, nearly every single true NV flagship card has outperformed NV's previous generation flagship by at least 50-80%, sometimes 100%. Neither GTX680, nor GTX980, nor GTX1080 have outperformed the previous generation NV flagship of that era (GTX 580 vs. 680, GTX980 vs. 780Ti, GTX1080 vs. 980Ti), by anywhere close to those numbers. That in itself proves alone that GTX680, 980, 1080 are not real flagships of the Kepler, Maxwell or Pascal generations -- they are simply marketing flagships that NV was able to price at $500 and above due to lack of competition from AMD. NV's 6600GT easily outperformed the last generation's 5900U/5950U but NV didn't call it 6800GT/Ultra and priced it at $499-699 back then. Why is that? Because ATI had flagship cards ready to level 6600GT.

NV's next generation mid-range card almost always either tied or outperformed the previous generation's flagship prior to Kepler. That's why GTX680 easily beating GTX580 was nothing special as all the previous mid-range next gen x04 parts did the same. Surely, people who revisit NV's history can recall GeForce 4 Ti 4200 beating GeForce 3 Ti 500, GeForce 5600U/5700U beating GeForce 4 Ti 4600/4800, GeForce 6600GT beating 5900U/5950U, GeForce 7800GT/7950GT beating 6800U, GeForce 8800GTS and 8800GT beating 7900GTX, GeForce 260 beating 8800GTX/9800GTX, etc.
FluffmeisterYeah cards like the GTX 1080 are expensive but it's clear Nv aren't pricing themselves out of the market (hence another superb quarter).
Just like the high-end audiophile headphone market, NV should keep raising prices even more to test just how much their consumer base and high-end PC gamers are willing to pay. No sense in stopping raising prices when it has worked so well since 2012. NV already priced GTX1060 at $249-299, a historical tier that used to be $129-149. With Volta, they should raise GTX2060 to $349, GTX2070 to $549, GTX2080 to $749. After all, GTX2060 will beat GTX1070, GTX2070 will beat GTX1080 and GTX2080 will beat GTX1080Ti GP102, so those prices will still be a great deal for consumers!!
Posted on Reply
#28
Slizzo
BlueFalconExcept this analysis is entirely flawed for 3 reasons.
I don't believe I ever truly analyzed anything, nor did I state that they haven't increased prices at all. I simply gave a quick comparison of prices from a GPU in the same sector, from the same company, from a generation prior to that which is currently available today.

Three years ago, their GPU in the same exact market segment was only $40 cheaper than the GPU currently on sale in the same exact market segment.

I know all about the GTX680, GTX 780, and GTX 780 Ti. I owned a GTX 680 for about a week before the GTX 780 launched, which I then returned the 680 and bought a 780. I paid the normal $650 for the GTX 780.

Which means that I paid LESS now for my GTX 1080 than I paid for my GTX 780. And I still gained performance.



Look, we can analyze the fact the nVidia is now pushing "mid-range" cards as enthusiast market cards, but when their competitors aren't pushing them to constantly market their most powerful GPUs as their highest tier, then who's really to blame?
Posted on Reply
#29
bug
BlueFalconExcept this analysis is entirely flawed for 3 reasons.

The first flaw is you are ignoring that NV raised prices 1.5-2X per each caterogy tier by re-labelling lower tiers as higher tiers, and then pushing back the true flagship towards the 2nd half of a generation. This started not during Maxwell, but during Kepler generation. What used to be $199-249 GTX560/560Ti GF114 tier became $399/$499 GTX670/680 GK104 tier. What used to be a cut-down flagship $289 GTX560Ti 448 Core and $349 GTX570 GF110 became $649 GTX780. This trend can be clearly illustrated today with $129 GTS450 GF106 or $149 GTX550Ti GF116 replaced by $199-299 GTX1060 3GB/6GB GP106, and with $249 GTX560Ti replaced by $599-699 GTX1080.

The second flaw in your post is complete ignorance of Kepler generation and the x70 tier by simply comparing 1080 Pascal to 980 Maxwell. Even if we ignore the fact that NV increased prices 1.5-2X starting with Kepler after Fermi, NV still managed to raise prices this round significantly. GTX680 debuted at $499 while GTX1080 costs $599-699. It took at least 6 months for GTX1080 to drop down to $599 as most AIB cards cost $650 or more during the first 4 months of release. Mere months post GTX970's launch, it was easy to find $330-340 AIB GTX970 but 2-3 months post GTX1070 launch, it was not possible to find most AIB 1070 cards below $420.

The comparison of GTX1070's pricing to GTX970 isn't also complete without factoring in that GTX1070 is a more cut-down part compared to the GTX970 or GTX670 during their respective generations. GTX1070's specifications and performance delta compared to GTX1080 more closely parallel the $299 GTX660Ti in the Kepler generation. When looking at it from this perspective, it can be easily argued that GTX1070 is just a $299 GTX1060Ti priced at $379-449. GTX680 was roughly 25% faster than GTX660Ti and GTX1070 is about 22% faster. GTX670 and GTX970 were much closer to the performance of the GTX680 and GTX980. You can try to argue that GTX1080 is now priced much higher than GTX680 but then you precisely fall into your flaw of arguing that NV didn't raise prices. What NV has done is actually raised prices on both the GTX1070 and GTX1080 cards because GTX1070 is not a proper x70 series card, but one of the worst cut-down x70 cards since GTX470. At least the GTX470 was cut-down from a flagship die.

The current Titan X Pascal is not much different from a GTX570 2.56GB (aka a cut-down flagship with double the VRAM of the consumer model). Last generation's GTX980 cost $649 and that was a cut-down flagship, much in the same way $349 GTX570 was a cut-down flagship. That means $649 GTX980's predecessor cost just $349. Should NV price 3328-3456 CUDA core 1080Ti at $699 or above, it will be yet another successor to the GTX275/GTX570 tier, but priced at 2X or greater of those predecessors.

The third flaw in your post is complete ignorance that NV's gross profit margins are approaching 60%. Prior to 2012, NV's gross margins were usually between 30-48%.


www.fool.com/investing/general/2016/03/30/nvidia-corp-stock-in-3-charts.aspx

Looking at NV's gross margins from 1999-2011 reveals an even more startling picture of gross margins. From 1999-2005, NV's gross margins never exceeded much more than 40%, with 4 of 7 of those years not even hitting 40%. From 2005-2011, not once have NV's annual gross margins even reached 50%, with years 2006-2007 and 2009-2011 staying at 47% or lower. Since 2012, NV's gross margins as if by magic went to > 55% and have only risen thereafter:
www.wikinvest.com/stock/NVIDIA_(NVDA)/Data/Gross_Margin

It can be easily concluded that NV has effectively raised GPU prices for every single dGPU tier since 2012 and used marketing names to obfuscate this fact. NV is just doing what a corporation should be doing and that is raising prices and seeing just how much the consumers are willing to pay. The introduction of the FE tax and further raising prices during Pascal generation highlights the ceiling hasn't been reached yet and we should expect even higher prices during Volta generation. The days of NV launching true flagship cards during the 1st half of a generation are over.

As a bonus point, going back to GeForce 2 and until Fermi, nearly every single true NV flagship card has outperformed NV's previous generation flagship by at least 50-80%, sometimes 100%. Neither GTX680, nor GTX980, nor GTX1080 have outperformed the previous generation NV flagship of that era (GTX 580 vs. 680, GTX980 vs. 780Ti, GTX1080 vs. 980Ti), by anywhere close to those numbers. That in itself proves alone that GTX680, 980, 1080 are not real flagships of the Kepler, Maxwell or Pascal generations -- they are simply marketing flagships that NV was able to price at $500 and above due to lack of competition from AMD. NV's 6600GT easily outperformed the last generation's 5900U/5950U but NV didn't call it 6800GT/Ultra and priced it at $499-699 back then. Why is that? Because ATI had flagship cards ready to level 6600GT.

NV's next generation mid-range card almost always either tied or outperformed the previous generation's flagship prior to Kepler. That's why GTX680 easily beating GTX580 was nothing special as all the previous mid-range next gen x04 parts did the same. Surely, people who revisit NV's history can recall GeForce 4 Ti 4200 beating GeForce 3 Ti 500, GeForce 5600U/5700U beating GeForce 4 Ti 4600/4800, GeForce 6600GT beating 5900U/5950U, GeForce 7800GT/7950GT beating 6800U, GeForce 8800GTS and 8800GT beating 7900GTX, GeForce 260 beating 8800GTX/9800GTX, etc.



Just like the high-end audiophile headphone market, NV should keep raising prices even more to test just how much their consumer base and high-end PC gamers are willing to pay. No sense in stopping raising prices when it has worked so well since 2012. NV already priced GTX1060 at $249-299, a historical tier that used to be $129-149. With Volta, they should raise GTX2060 to $349, GTX2070 to $549, GTX2080 to $749. After all, GTX2060 will beat GTX1070, GTX2070 will beat GTX1080 and GTX2080 will beat GTX1080Ti GP102, so those prices will still be a great deal for consumers!!
Just so people won't think you're biased or something, will you, please, post the same analysis for ATI/AMD?
Posted on Reply
#30
nguyen
oh wow who would have thought spending big in R&D, increasing performance and efficiency each generation would benefit a company so much as compared to cutting down on R&D and bring out recycled products every generation. It's AMD who shoots itself in the foot, if not for the bitcoin mining community buying millions of AMD cards I suspect sales would be dismal for the AMD camp.
Posted on Reply
#31
m1dg3t
Such disappointment, clearly they need to increase the MSRPs a further $100 per tier. Minimum. I wont be happy until I can buy a gtx 1050 for $600.
Posted on Reply
#32
TheGuruStud
nguyenoh wow who would have thought spending big in R&D, increasing performance and efficiency each generation would benefit a company so much as compared to cutting down on R&D and bring out recycled products every generation. It's AMD who shoots itself in the foot, if not for the bitcoin mining community buying millions of AMD cards I suspect sales would be dismal for the AMD camp.
See, you say this meaning AMD...and I all I can think is, "This is nvidia and amd, wtf is he talking about, especially pascal, it's a cut down joke (so far)."
Posted on Reply
#33
Caring1
nguyen... if not for the bitcoin mining community buying millions of AMD cards I suspect sales would be dismal for the AMD camp.
Pity that hasn't happened for a few years now, miners use Asics, not GPU's.
Posted on Reply
#34
MxPhenom 216
ASIC Engineer
catulitechupThis happend when companies provides competition literally dont show improvements and this provoke see this situations

And when appears amd vega, nvidia dont stay worried because amd gift around 10 months without competition (if vega appears in march as amd said)

For this moment nvidia now sell when them wait sell with big earnings, thanks to pascal inflated prices for lack of competition
Can you say that again?
Posted on Reply
#35
ppn
BlueFalconThe current Titan X Pascal is not much different from a GTX570 2.56GB (aka a cut-down flagship with double the VRAM of the consumer model). Last generation's GTX980 cost $649 and that was a cut-down flagship, much in the same way $349 GTX570 was a cut-down flagship....
GTX 570 cut down 520 sq.mm chip. Now Titan Pascal even smaller 471 sq.mm still gimped and 16nm more expensive to make. Usually it takes 15-18 months for a Titan to become more like 70-tier and affordable at some point. Titan Maxwell is slower than 1070. I wouldn't buy 1070 either or if absolutely necessary will not keep it for more than 6 months since launch, but rather sell while it still retains some value. Or another way Buy SLI and replace it with single card when it costs almost nothing like 970 SLI owners moved to 1070 for free.. 1070 will inevitably transition to GTX 2060 or something like that. It boils down to how much are you willing to spend per year.
Posted on Reply
#36
nguyen
ppnGTX 570 cut down 520 sq.mm chip. Now Titan Pascal even smaller 471 sq.mm still gimped and 16nm more expensive to make. Usually it takes 15-18 months for a Titan to become more like 70-tier and affordable at some point. Titan Maxwell is slower than 1070. I wouldn't buy 1070 either or if absolutely necessary will not keep it for more than 6 months since launch, but rather sell while it still retains some value. Or another way Buy SLI and replace it with single card when it costs almost nothing like 970 SLI owners moved to 1070 for free.. 1070 will inevitably transition to GTX 2060 or something like that. It boils down to how much are you willing to spend per year.
oced vs oced Titan Maxwell still faster than gtx 1070 and 10% behind gtx 1080 (1400mhz vs 2000mhz), even 980ti oced is faster than gtx 1070 oced. 780ti users got screwed with the launch of gtx 970 but 980ti is still in pretty good spot.
Caring1Pity that hasn't happened for a few years now, miners use Asics, not GPU's.
litecoin or some other still use GPU i believe, in my country there are a bunch of used R9 290x that has been through coin mining for some years
Posted on Reply
#37
FYFI13
lZKoce:D or may be, just may be they have a very good pricing calculations. You'd be surprised how many business base their prices on "common sense", "what seems right" and other non-objective factors.
Sounds about right, except the fact that their cards are ridiculously overpriced, which doesn't seem right to me. ~800 euros for a GTX 1080 here in Europe? No, thanks.
Posted on Reply
#38
GC_PaNzerFIN
Couple billions of dollars investment in this generation of cards alone. Gamble is paying off, obviously getting harder to make better and better GPUs and costs need to go to customer.

Expecting everything cost the same although manufacturing & R&D costs go up is a bit silly. That said, competition can result in more aggressive pricing, but both manufacturers seem to agree to increase prices to cover the costs so that is not a long term solution either (unless you want the weaker side go out of business).

I paid 600 euros for ATi X800XT PE years ago, and now I got my GTX 1080 for a bit over 700 euros. Can hardly fault NVIDIA for the pricing. Only thing that has changed is they now offer the superduperexpensive premium range, which nobody really needs.
Posted on Reply
#39
sweet
SlizzoGoing to trot out my favorite thing to state when people complain about new pricing vs. old.

In 2014, the MSRP of the GTX980 at launch was $549. Adjusted for today, that's just a shade under $560 ($559.88 to be exact).

Launch MSRP of the GTX1080 was $599.

Increase of ~$40. Not exactly earth shattering.



I am not going to go into how "NO GTX 1080 WAS EVER AVAILABLE FOR $599 AT LAUNCH!!!!". Truth is you never really can find any card for MSRP at launch as everyone is scrambling for the cards at launch. Currently you can get an MSI Armor 1080 for $599 or less (has an aftermarket dual-fan cooler)
GTX 980 launching price was a totally rip-off though.
Posted on Reply
Add your own comment
Nov 22nd, 2024 18:48 EST change timezone

New Forum Posts

Popular Reviews

Controversial News Posts