Nvidia (NVDA) is a controversial name in the semiconductor space, whose long-term viability has been widely questioned on Wall Street. I believe that investors not only drastically overstate the headwinds that the firm will face, but they also overlook the numerous growth avenues that the firm has paved over the last several years, creating a compelling long opportunity in plain sight. With numerous upside catalysts set to hit over the next 6 to 12 months, coupled with the continued broad misunderstanding of the business, shares have potential upside of ~$10/share from the $14.50/share level with downside risk of just $2/share, making it one of the most compelling long opportunities on the market.
A Review Of The Business
Nvidia's primary business is the development of high performance graphics processing units ("GPU"), which are used to accelerate applications such as 3D gaming, visualization, and scientific computing. While in the past, Nvidia has broken down revenues into three major buckets (gaming graphics, professional graphics, and Tegra), the company now merely reports two major segments: GPU (which comprises gaming, HPC, and professional graphics) and Tegra (Nvidia's mobile system on chip).
Within the GPU business, we have the following sub-segments (that are no longer reported explicitly, but are vital to understand)
- GeForce (gaming GPU)
- Quadro (professional/workstation GPU)
- Tesla (high performance computing oriented GPU)
The Tegra business, on the other hand, can be split similarly into the following buckets, although Nvidia has not historically provided this sort of breakdown in its quarterly reports (but is important to understand the moving parts of the business)
My long thesis rests on the following views, which I will justify throughout the remainder of this article:
- Integrated GPU solutions from Intel (INTC) and Advanced Micro Devices (AMD) unlikely to damage current discrete GPU sales, despite broad fears that this is imminent
- Continued market share gains against AMD in the gaming GPU space
- Continued strength in the professional GPU space, particularly as the "Kepler" refresh and improving macroeconomic conditions serve as two important tailwinds
- Market share gains and broader market penetration into the mobile computing space with Tegra 4 and, more importantly, Tegra 4i
Further, while my thesis stands alone without the following potential upside drivers, the following could serve to drive unexpected upside:
- SHIELD (and potential Tegra revenue in other handheld gaming devices)
- Project Denver in HPC/data-center
The remainder of this article is aimed at fleshing out and quantifying the opportunities outlined above, and to ultimately show why Nvidia has ~$10 of upside potential over the next year with downside risk capped to ~$2/share.
GeForce: The Death Of The Discrete GPU Has Been Exaggerated
The most controversial business segment at Nvidia is the "GeForce" business. The argument from the bears (very elegantly outlined by our own Paulo Santos) goes something like this:
Nvidia is set to lose the discrete graphics market, particularly in notebooks, as Intel's integrated solutions become more powerful. Not only will Intel's iGPUs enable nearly the same performance in an integrated package, but this will lead to substantially thinner and lighter designs, further hurting discrete GPU attach rate
On the surface, this seems like a plausible argument, right? Well, not so fast. There are a number of tacit assumptions here that I believe may not be applicable to this market:
- Form factor trumps raw performance
- Power consumption trumps raw performance
- Games will not demand further horsepower, thus eliminating the need for higher performance parts for most mobile gamers
Then, on top of the "Intel Is Coming" argument, we have the "AMD won the game consoles argument" in which AMD bulls/Nvidia bears claim that Nvidia is now at a disadvantage as developers will now tailor their code more towards AMD's GPU architecture and focus less on Nvidia's architecture.
I believe that these arguments do not hold up.
Gaming Is A Moving Target At The High End...
Modern PC games are incredibly demanding, and the fact that Nvidia has been able to sustain/grow its business over the last decade is a rather strong testament to this. Now, many investors are under the impression that Intel/AMD integrated graphics are now "good enough," but at higher quality settings on more demanding games, this could not be further from the truth. Let me present some data, measured by Anandtech, in its review of Intel's latest "Iris Pro" graphics - the fastest integrated GPU currently shipping:
In EA's (EA) Battlefield 3, Intel's best integrated graphics in a 55W envelope can barely hit 28 frames per second in what is one of the more popular first person shooter titles. Keep in mind that in a fast-paced multi-player game, anything under 30 frames per second is unacceptable. This is also at 1600x900, well under "Retina" resolutions.
So, what happens when games start becoming more demanding, and the standard resolution on a gaming notebook is 1920x1080 or, probably a little farther out, 2560x1440? Not only is there a dramatic demand increase for higher power graphics parts to drive these higher resolution displays, but the games themselves will use more sophisticated effects that further tax the graphics hardware.
The point that I am trying to make is rather simple: no matter how "good" integrated graphics gets, the low end of the discrete market will be a step or two ahead, and the demands of the games will similarly demand more from the latest graphics hardware. This is why I fully believe in the following trends:
- Discrete GPUs for high end, high TDP notebooks will not be supplanted by integrated graphics solutions
- The notion of a "good enough" is invalid in the face of dramatically increasing native resolutions on notebooks and desktops, as well as higher image quality within games themselves
So, for the gaming market, highly priced CPU + GPU combinations in a restricted TDP are very clearly unlikely to make a dent in the discrete market, but the issue arises when it comes to the category known as "Premium Notebooks."
The Premium Notebook... What Happens?
Generally speaking, there exists a rather substantial class of buyers who simply buy discrete graphics simply to have something that is "better." These individuals may or may not play games, but it is primarily about having something that is "premium." According to Nvidia, here is how gamers versus premium users contribute to the top and bottom lines:
(click to enlarge)
20% of GPU units come from "special purpose" PCs (gaming, workstation, and HPC), and a whopping 80% come from these "premium" PCs. However, if we look at this from a gross margin dollar perspective, the tables turn and it is clear that most of the firm's bottom line comes from the "special purpose" PCs:
It is clear that the "Premium PCs" are not the largest direct contributors to gross margin dollars, but they are still important in that they allow Nvidia to amortize the fixed costs over a substantially larger number of units to lower average cost per die.
With each successive release of new Intel integrated GPUs, the gaming and media performance gets better; for non-gamers, the improvements from Ivy Bridge to Haswell integrated graphics are likely inconsequential. Despite this, the trend in Nvidia's revenue base has been clear:
My view of why Nvidia's "premium" PC segment is safe is due to the following factors:
- From an OEM standpoint, it is cheaper (and likely delivers a better user experience) to pair a mid-range Intel CPU with an Nvidia GPU than it does to pay for the absolute top of the line Intel part which, in all likelihood, has inferior graphical performance
- From a user standpoint, it is likely that the "GeForce" brand is likely much more associated with "graphics" and "added features" than "HD Graphics" or the very expensive "Iris Pro"
- AMD's integrated graphics are actually quite good from a performance/functionality standpoint, and yet AMD's inexpensive APUs - as well as its discrete notebook GPUs - have been losing share in the marketplace, suggesting that Nvidia's position in premium PCs is much more resilient than what one may initially believe
In addition, it is important to note that ASPs and gross margins have improved substantially as Nvidia's competitive position has strengthened relative to AMD. In the most recent "Kepler" generation, Nvidia was able to build a faster-but-smaller die for gaming, and was able to reserve its large die parts for HPC ($2000+), workstation, and ultra-high end desktop GPUs ($650+). This is, I believe a major factor for Nvidia's recent outperformance on the gross margin front in the most recent quarter and for the strong gross margin guide for the current quarter. The R&D for both lines of GPU is justified by Nvidia's strength in HPC and workstation - something that AMD likely cannot justify as its GPU division is barely turning an operating profit.
Intel's Xeon Phi - A Threat, But Not A Show Stopper
While I do not view Intel's integrated graphics as particularly threatening to Nvidia's GeForce line, especially given the target market(s) for Geforce, I believe the major structural threat to Nvidia's top/bottom line growth is in the HPC accelerator space. In particular, while Advanced Micro Devices has put out solutions, they have failed to gain traction in the marketplace due to a number of factors (developer/software ecosystem is geared towards Nvidia's CUDA, hardware that is better at reaching peak theoretical performance, mind share).
While AMD has neither the resources nor the focus on this space to pose a legitimate threat, Intel certainly does. Intel's "Xeon Phi," which is a custom-designed HPC part that consists of ~60 small X86 cores each married to an extra wide (512bit) vector unit, is a rather potent competitor in HPC, particularly in double precision. That being said, Nvidia's solutions currently have higher peak throughput in both single and double precision, although Intel is likely to become more competitive over time.
The thesis here is the following: while I expect Intel to gain share in this space (and AMD to remain a negligible player as it shifts focus to other areas), I further expect that an increasing TAM, coupled with Nvidia's "head start" in this space to keep the business viable and growing particularly as there are a number of diverse end markets that have only begun to exploit these HPC accelerators.
Quadro: Riding The Macroeconomic Upturn And A Refresh
During FY2013, a sore spot for the company was the Quadro business. This was driven by three major headwinds:
- The Intel "Romley" workstation platform was delayed, which threw off the traditional upgrade cycle
- The macroeconomic environment was weak
- The "Kepler" refresh had not yet occurred
Fortunately, Romley is well into its ramp, with the "Ivy Bridge EP" refresh for the platform coming in September (which could spur further sales), so that headwind becomes a tailwind. The macroeconomic environment, while not particularly great, is improving. Finally, Nvidia has refreshed its workstation parts to be based on the smaller-die (i.e. higher margin), lower power, and faster "Kepler" GPU.
The bottom line is that this business has bottomed, which should drive revenue and margin (since Quadro is higher margin than GeForce or Tegra) upside during the rest of the fiscal year.
Tegra - The Elephant In The Room
Despite the strength of Nvidia's core GPU business, and while the HPC segment represents a rather robust growth opportunity, the real story that grabs the headlines is the mobile system-on-chip business consisting of the firm's Tegra lineup for smartphones, tablets, and automotive.
Nvidia bears typically point to the following negative points regarding the Tegra business:
- It's losing money
- It's a tough competitive landscape
- Nvidia is competing against much larger rivals, thus there is very little chance of long-term success
I would like to now take an in-depth look at this business, the technology, and the competitive landscape to give context to my long-term thesis that Nvidia indeed has a very good chance of running a profitable, high growth business with Tegra.
The Mobile SoC Market - Crowded And In Need Of Differentiation
The mobile system-on-chip market is highly fragmented with many players currently vying for the prize. The ARM Holdings (ARMH) business model is set up so that almost any company can license all of the difficult-to-develop IP and sell customized system-on-chip solutions based on ARM designed processor cores, graphics IP, and so on. Further, ARM has been expanding its business to include physical design IP to further unburden the licensees when it comes to implementing system-on-chip solutions utilizing ARM's processor and graphics cores. The barrier to entry into the mobile SoC world has been lowered substantially and as with any new, secular growth industry, there have been many competitors.
Of course, when nearly every competitor is using the same off-the-shelf core designs, graphics designs, and foundries, there is very little room for differentiation. The ARM licensees have all been slugging it out for design wins, and the winners of this race have been the following:
- Those that could provide the cheapest solution
- Those that could provide a solution with an LTE modem, preferably integrated onto the same silicon die
- The foundry
This means that the winners thus far have been MediaTek and Allwinner at the "cheap" end, Qualcomm on highest performance/highest integration, and then Taiwan Semiconductor (TSM) as it manufactures chips for Qualcomm along with all of its competitors. Of course, ARM gets a cut from all of the licensees, no matter which competitor "wins."
Nvidia, while presenting huge ambitions, plans, and roadmaps, has been - up until now - rather absent from the race in any meaningful way. The Nexus 7/Surface RT wins were positive and drove volume of Tegra 3 significantly, but with the next generation Nexus 7 lost, and with Intel killing Windows RT with its full Windows 8 compatible chips, the future began to look bleak for Nvidia's mobile strategy. However, digging beneath the headlines, we begin to see a longer-term strategy at play here.
Nvidia delayed the launch of its Tegra 4 in order to pull in the launch of its Tegra 4i smartphone platform (integrated LTE, Cortex A9r4 cores, updated same GPU architecture as found in Tegra 4 but scaled down slightly). While this is what led to Nvidia claiming a flat FY2014 for Tegra, I believe that FY2015 - when Tegra 4i is in full throttle - should see not only momentum in the compute space with Tegra 4 (which should still be high end given that 20nm Cortex A15 designs should be a Q3 or later launch at the earliest, and that 20nm Cortex A57 designs are 2015 at the earliest according to ARM's own material), but will also be able - for the first time - to make some serious headway into the high volume smartphone space.
My belief is that with Nvidia's leadership and differentiation in graphics (massive operating and IP leverage here), its use of ARM's rather good A9r4 cores (that aren't power hungry and unsuitable for phones like the A15), and with an integrated LTE modem, Nvidia should be able to win some high volume designs, and this - in tandem with the Tegra 4 in tablets/convertibles - should drive significant sentiment as well as top and bottom line growth. Further, I believe that as Nvidia is one of the few players with the resources to develop its own graphics IP, modem, and - in the 2015 timeframe - its own processor core with "Project Denver," it will be one of the survivors of the initial "race to the bottom" before all of the "me too" undifferentiated players eventually leave for greener pastures.
This is a long-term game, and Nvidia is very well positioned to take advantage of it.
Additional Upside Drivers
In addition to Tegra, Tesla, Quadro, and GeForce, Nvidia is also trying its hand at developing visual computing appliances under the name "GRID." This is a completely untapped market (in fact, the market does not yet exist) which means that it is extraordinarily difficult to model, and as a result, I do not bake in any meaningful upside from these initiatives. As the first numbers come in, it will become much easier to get a read on where this business will go - if it goes anywhere at all.
Additionally, Nvidia's handheld Android gaming device is in a similar boat, although my belief is that the capability of the device to stream games from a PC to the handheld device wirelessly will drive further market share gains as this functionality is available only if the user has a GeForce graphics processor.
Finally, Nvidia's capital allocation policy has become extremely friendly, with the firm intending to return ~$1.2B in dividends/buybacks during FY2014 while shares are still cheap.
Modeling The Upside And Downside
My view is that long term, we are looking at a 50-54% gross margin business as high end GPUs, workstation/professional GPUs, and Tesla continue to drive strength in the GPU business, offset slightly by the Tegra business (depending on how aggressive Nvidia gets about winning designs). Starting in 2014, I expect the Tegra business to grow roughly in-line with projected smartphone tablets, which implies a conservative ~30% CAGR through 2016. I expect the GPU business to grow at mid-to-high single digits, with additional upside possible, should Tesla and Quadro see a dramatic outperformance.
By the end of FY2015, I expect GPU revenue to be $3.6B assuming a 5% CAGR from here. I further expect Tegra revenue to come in at $1B with the assumption of 30% growth. Finally, the Intel royalty payments should continue through 2016, which implies an additional ~$250M. Finally, I assume opex of $1.5B, tax rate of 16%, and gross margin of 53% for net income of $830M, or at a diluted share count of 590M, implying EPS of $1.40. At a 15x multiple, this translates to a $21/share price target exiting FY2015. I expect that this is conservative as the diluted share count should actually be significantly reduced from current levels.
The primary downside risks:
- Nvidia losing market segment share to rival AMD in the notebook/discrete market
- Nvidia's Tegra suffering delays/poor adoption
- Intel's Xeon Phi having a larger-than-expected impact on Tesla market share
I believe that with $3.7B on the balance sheet (after large buyback/dividend, likely to be more along the lines of $3B, but share count is reduced), four years of consistent GPU growth (especially since Nvidia's competitive position is significantly better than it was in 2009/2010), an aggressive buyback, and a worst case fate of either being bought by private equity in an LBO (given the debt-free, cash rich balance sheet coupled with reasonable multiple) or becoming a target for the likes of Samsung aiming to bring graphics expertise in-house, it is tough to see the company trading at less than $12.5/share (7x FY2013e EPS + $5/share in cash).
I further believe that Nvidia will be fairly aggressive in ramping its dividend, which not only puts a floor underneath the stock, but this serves long-term investors well as this floor would consistently be raised. At a constant payout ratio of ~36%, with $1.35 EPS, a dividend of $0.49/share on an annualized basis would not be out of the realm of possibility, and at current levels ($14.2), this would represent a yield of 3.4%.
The risk/reward looks compelling with downside seemingly limited to $12.5/share and upside to $20/share by the end of FY2015, all supported by more aggressive buybacks and potential dividend increases, it seems that Nvidia is one of the best risk/reward plays out in plain sight on the market.
Additional disclosure: I am short ARMH. I may add (but have no plans to reduce) to my long position in NVDA and INTC, or add/reduce my ARMH short position at any time.