NVIDIA’s fiscal year for 2017 ended on January 29, 2017, and for the fourth quarter and full year, NVIDIA reports they had record revenues. For Q4 2017, NVIDIA is reporting revenue of $2.173 billio.n, up 55% from a year ago. Gross margin came in at 60.0%, which is up 3.5% from Q4 2016. Operating income for the quarter was $733 million, up 191%, and net income was up 216% to $655 million. The very high growth in operating income and net income are even more impressive, since Q4 2016 was also a record for them at the time. NVIDIA reported $0.99 per share earnings for this quarter, up 183%.

NVIDIA Q4 2017 Financial Results (GAAP)
  Q4'2017 Q3'2017 Q4'2016 Q/Q Y/Y
Revenue (in millions USD) $2173 $2004 $1401 +8% +55%
Gross Margin 60.0% 59.0% 56.5% +1.0% +3.5%
Operating Income (in millions USD) $733 $639 $252 +15% +191%
Net Income $655 $542 $207 +21% +183%
EPS $0.99 $0.83 $0.35 +19% +183%

Even though 2016 was a record year for NVIDIA, 2017 was even better. NVIDIA reported record revenue of $6.91 billion for the full fiscal year 2017, which is up 38% from a year ago, and had a net income of $1.666 billion. Earnings per share for the year were $2.57. They also repurchased $739 million in shares over the last year, and paid $261 million in dividends.

The majority of NVIDIA’s revenue came from their Gaming segment, which includes their GeForce GPUs. The launch of Pascal has clearly been very successful, and for the quarter NVIDIA’s Gaming segment had revenues of $1.348 billion, up from $810 million a year ago. This is a gain of 66.4% year-over-year, on a market they were already at the top of. This quarter, they also launched the GeForce GTX 1050 and 1050 Ti mobile GPUs, and a new SHIELD TV, along with unveiling the GeForce NOW game streaming service.

Professional Visualization had much more modest growth, being up “only” 10.8% from last year, with revenues of $225 million for the quarter. They just launched the GP100 Quadro though, so it may have even more growth to come.

Datacenter has been a big focus from NVIDIA in diversifying their portfolio, and this quarter they had some announcements here as well, with a collaboration with Microsoft on the Microsoft Cognitive Toolkit, powered by Microsoft Azure cloud running on NVIDIA DGX-1 deep learning system. They unveiled the DGX SATURNV AI supercomputer, and partnered with the National Cancer Institute and US Department of Energy to build CANDLE to “advance cancer research”. The growth of NVIDIA’s Datacenter segment has been almost exponential, and it is now the second largest source of revenue for the company. For Q4 2017, NVIDIA reported $296 million in revenue from this segment, up from $97 million a year ago, or an increase of 205.1%.

NVIDIA has also diversified into Automotive with their Tegra lineup, after originally releasing it as a mobile SoC. This has also been a boon for the company, and they are now powering self-driving cars, and the infotainment systems of major manufacturers like Audi, and Mercedes-Benz, and they’ve partnered with Bosch on self-driving systems as well. This segment saw revenues of $128 million for the quarter, up from $93 million a year ago, or 37.6%.

Finally, the OEM and IP segment had revenues of $176 million, down from $198 million a year ago. Unfortunately NVIDIA has not gone into much detail about just why this is dropping, and how revenue from this category breaks down.

NVIDIA Quarterly Revenue Comparison (GAAP)
In millions Q4'2017 Q3'2017 Q4'2016 Q/Q Y/Y
Gaming $1348 $1244 $810 +8.4% +66.4%
Professional Visualization $225 $207 $203 +8.7% +10.8%
Datacenter $296 $240 $97 +23.3% +205.1%
Automotive $128 $127 $93 +0.8% +37.6%
OEM & IP $176 $186 $198 -5.4 -11.1%

For next quarter, NVIDIA is expecting revenues of $1.9 billion, plus or minus 2%, and gross margins between 59.5 and 59.7%.

The diversification of NVIDIA from a purely GPU company has clearly paid dividends, especially when their core market is doing as well as it is. The datacenter growth is clearly very high margin as well, considering the jump in net income for the quarter.

Source: NVIDIA Investor Relations

 

Comments Locked

21 Comments

View All Comments

  • awehring - Thursday, February 9, 2017 - link

    Grossmargin up 3.5% year-over-year means, that Nvidia could take very high prices on its products. Who doubts, that they did it. Time (and much room) for competition from AMD.
  • Yojimbo - Friday, February 10, 2017 - link

    It has a lot to do with their changing market mix. Their highest margin business is the data center business, which had 3 times the revenue this past quarter than the year ago quarter. They are making less lower margin OEM sales. And since the release of the Pascal cards they've had the high end of the GPU market to themselves, so a greater percentage of their gaming sales is from the higher margin high end (not that the Fury line ever sold all that much). Additionally, the entire market has been moving more towards the high end in the last few years.

    All of that has nothing to do with a greater selling price for equivalent SKUs because of a lack of competition. Such a greater selling price doesn't seem to exist. 10 series prices are pretty much in line with 700 series prices from 3 to 4 years ago, which was a time when there was more competition from AMD.
  • testbug00 - Friday, February 10, 2017 - link

    don't forget that Pascal also has brought FE into the mix, which "steals" margin from AIBs.
  • BurntMyBacon - Friday, February 10, 2017 - link

    Many good points here that I generally agree with. However:
    @Yojimbo: "All of that has nothing to do with a greater selling price for equivalent SKUs because of a lack of competition. Such a greater selling price doesn't seem to exist. 10 series prices are pretty much in line with 700 series prices from 3 to 4 years ago, which was a time when there was more competition from AMD."

    These numbers were compared against last quarter and last year, so why confuse things by comparing to 3 to 4 year old graphics cards. There was in fact a measurable if not significant price bump compared to the 900 series (at least at the upper end). The prices also seem to be staying higher measurably if not significantly longer than they did with the 900 series. More importantly, the die size (significant cost contributor) for the 1000 series is markedly smaller than the last generation (roughly) price equivalent. Most importantly, having higher gross margin suggests pretty directly that they were bringing in more money vs the cost of the products they were selling.

    Now, as you said, there are a lot of good reasons outside higher SKU profits (whether higher price or lower manufacturing cost) for their higher gross margins. Certainly the 205.1% Y/Y ($99M) growth in revenue for the datacenter segment was hugely beneficial here. I just wanted to point out that higher profits on SKUs was still a measurable if not significant factor. Keep in mind, while the gaming segment revenues grew by only a 66.4% Y/Y, nVidia brought in more additional revenue ($538M) Y/Y in this segment than the entire datacenter segment ($296M).
  • Yojimbo - Friday, February 10, 2017 - link

    "These numbers were compared against last quarter and last year, so why confuse things by comparing to 3 to 4 year old graphics cards."

    The comparison to 3 or 4 year old graphics cards had nothing to do with refuting margins numbers, but rather to do with the conclusion the original poster made by looking at the margin numbers. It is independent evidence to refute his conclusion and it doesn't "confuse things".

    "There was in fact a measurable if not significant price bump compared to the 900 series (at least at the upper end)"

    The fact that there was a increase compared to the 900 series, a series that also faced limited competition at the high end, does not refute my evidence that in the long term the claim that decreased competition has lead to more expensive graphics cards is not true.

    "More importantly, the die size (significant cost contributor) for the 1000 series is markedly smaller than the last generation (roughly) price equivalent. "

    What is of overriding importance here is that the node used for the 1000 series was very new, whereas the node used for the 900 series was very old. Older nodes have lower costs and higher yields. That's why the die size of the 1000 series needed to be smaller. Notice that the 900 series, while cheaper than the 700 series, had die sizes significantly larger while they were both manufactured on the 28nm node. And the 700 series had die sizes larger than comparatively positioned SKUs within the 600 series that were manufactured on the 28nm node.

    "Most importantly, having higher gross margin suggests pretty directly that they were bringing in more money vs the cost of the products they were selling."

    That's a circular argument. This is the statement you are meant to be proving, you can't just restate it and say it asserts itself.

    "Keep in mind, while the gaming segment revenues grew by only a 66.4% Y/Y, nVidia brought in more additional revenue ($538M) Y/Y in this segment than the entire datacenter segment ($296M)."

    Yes, and hence the data center only accounts for part of the margin increase. Also accounting for it is the fact that NVIDIA is not selling so many 710s, 720s, 730s, 740s, 830Ms, 840Ms, 850Ms, 910Ms, 930Ms, 940Ms, etc, and a whole lot more 1060s, 1070s, and 1080s, which are the equivalent of 760s, 770s, 780s, 960s, 970s, 980s, 980Ms, 970Ms, etc.
  • Yojimbo - Friday, February 10, 2017 - link

    To be a little more clear about what I said about circular argument. It's circular insomuch as it actually is an argument. In fact it doesn't explicitly say anything at all except a simple definition of what a higher margin is. But you seem to be claiming that simply stating the definition somehow leads to the conclusion you want, i.e., that those higher margins must come from higher margins on similarly positioned SKUs due to less competition, and that is not true. In the post you replied to I already listed alternative ways margins could rise without that claim being true.
  • Yojimbo - Friday, February 10, 2017 - link

    "Finally, the OEM and IP segment had revenues of $176 million, down from $198 million a year ago. Although this dropped, it’s likely due to the loss of the $66 million/quarter they were receiving from Intel as part of the settlement agreement made in 2011, which Intel paid it’s last payment in January 2016."

    No, this reported quarter still includes the same $66M Intel IP payment they've been claiming each quarter. Next quarter will include a partial payment and that will be the last of it.

    NVIDIA's OEM revenue has been decreasing for a number of quarters, by design. The say their OEM sales are lower margin than their other segments and they are focusing on higher margin businesses.
  • Brett Howse - Friday, February 10, 2017 - link

    Thanks I assumed it was over since Intel paid them last in January 2016, but they've been claiming this deferred so there was a bit more then. I'll update the piece.
  • p1esk - Friday, February 10, 2017 - link

    I just bought 4 Titan X cards for neural network simulations. I wonder it they count that as "gaming".
  • Yojimbo - Friday, February 10, 2017 - link

    I was wondering how they account for Titan X or Quadro cards used for training neural networks or other non-graphics, non-visualization tasks. I know remote visualization via GRID is counted as data center and not professional visualization, because someone asked that in the Q3 conference call. My guess is that if you're not someone like Baidu buying the Titan Xs it would be counted under gaming. That Quadro GP100 card is a bit of a conundrum, though. Well a workstation probably isn't part of a data center so I'd guess it does go under Professional Visualization.

Log in

Don't have an account? Sign up now