Texas Instruments, widely known as the bellwether of the semiconductor space, reported a big miss yesterday on the top and bottom line. With sales down 11% YoY and a downside Q4 forecast of $3.07-$3.3B compared to the consensus of $3.6B, the stock slipped just around 10%.
Semiconductors trade like commodities in that they are highly cyclical, and the timing of the end of the cycle is typically sudden and a sharp move.
I encourage you to review the technical outlook in the Nvidia PDF that I wrote mid-September. The global outlook around the KOSPI as well as the Philadelphia Semiconductor are still in place and have not changed much, since my last writing.
In other words, the KOSPI is in a bear market with weak momentum, suggesting a lower leg, and the Philadelphia Semiconductor Index (SOXX) is still trading in an ascending wedge pattern with weakening momentum.
The US semis are crucial and we’ve now seen two warnings – Texas Instruments and Micron. Xilinx didn’t have the best quarter-over-quarter results even if they did beat analyst estimates. This quarter, Xilinx reported $833 million in revenue down from $850 million last quarter. EPS was flat at $0.94.
To highlight the SOXX ETF, which tracks the Philadelphia Semiconductor Index, you can visually see this pattern unfolding. Notice how the index has been making higher highs while the momentum indicators in the RSI and MACD are decelerating, making lower lows.
The technical evidence coupled with the recent earnings reports of Micron, Texas Instruments and today’s Xilinx suggests that this pattern should resolve to the downside.
Our primary entry target is Nvidia. This is a high conviction, long-term hold.
Also, it’s one of the strongest names in the semi sector, and any cyclical slowdown has not caught up to the sentiment in Nvidia. As you can see, its momentum is also fading as it is failing to break and close above the $200 region.
If you’re bearish on semis, the easiest way to play this position would be either short or buy a put on SMH or SOXX, which are both ETFs. The SMH tracks the VanEck Semiconductor Index while the SOXX tracks the Philadelphia Semiconductor Index.
However, SOXX has a few more names and also caps the allocations allowed, so it’s technically more diversified. Either option should be sufficient. As more earnings flow in, we should get a clear direction for the near future. In the mean time, I’ll leave you with a quote from the recent Texas Instruments (TXN) earnings call.
In the quote below, TXN states their negative results are due to a broad-based macro slowdown on the earnings call.
“I can sense that you collectively are unsatisfied with our answers, and I understand that. We have close to 100,000 different customers, and we sell about 100,000 different products. It’s difficult to pinpoint any one thing, but the sense we get, talking to those customers, getting input from them, from our sales people and all the touchpoints that we have, is that the weakness is broad-based. It’s due to macro events and specifically the trade tensions. And if you think about when there’s tensions in trade and obstacles to trade, what do businesses do? They become more cautious, and they pull back. And we are at the very end of a long supply chain, and when the ones at the very front pull back, it becomes a traffic jam. And so our sense is that is what’s happening in the marketplace. But we’ll see what other companies will report over time and we’ll get a clearer picture over the next several weeks and really quarters, because this thing, we’ve been in it for now four quarters, and it’s going to be longer than that.”
We have a few years before Nvidia will show the market it’s true earnings potential. When a thesis is not reflected in the revenue segments yet, there are typically lower entry points and ongoing volatility. You’ll see in the technical analysis that although I could not be more bullish on this stock long-term, there is weakness in the semiconductor sector and we hope this translates to a lower entry point for our readers.
The market is also in a fierce debate between AMD, Intel, and Nvidia and is also distracted by other chips, such as Micron and NXP. In my analysis, I look for growth. How big is the market relative to how big the company is now?
You can ignore Nvidia’s gaming revenue and other segments for the main trajectory that we are focused on. Gaming is great for stability and earnings reports, but the growth will not be from gaming (a market where Nvidia is already a mature, market leader). I’m also not focused on PC sales or the CPU-powered cloud, as the first is not a growth market and the second is not the piece in the cloud stack that will accelerate future technologies.
I’ve written at length about the Mellanox acquisition and it’s a great reference for Nvidia’s long-term strategy. In this report, I’d like to break down what the GPU-powered cloud is capable of and why it’s important to differentiate Nvidia’s strategy from the competitors (and some competitors to keep an eye on).
I’ve covered Nvidia a few times on my free blog, however, it would be a disservice to my premium members to not formally initiate coverage and provide critical updates to the GPU-powered cloud and AI economy as it’s built out. I believe Nvidia will be on my short list in a decade from now.
To be bold – I believe Nvidia will be one of the world’s most valuable companies by 2030. The research below organizes my investment thesis for the GPU-powered cloud and why I believe Nvidia will emerge as a clear leader.
The question is how long can you have your money in a stock? This is a long-term play that requires a 10-year hold for the full return. Like Amazon, Google and Netflix, it requires ten years before emerging technology goes from a moonshot, to a viable company, to a less volatile company – and finally, to a profitable machine. Nvidia is a viable company that is volatile, proven during the crypto surge. Investors should see the crypto bust as an opportunity as the issues were not based on Nvidia’s core business. The trade war certainly hasn’t helped the stock price, either.
We have a few years before Nvidia will show the market it’s true earnings potential. When a thesis is not reflected in the revenue segments yet, there are typically lower entry points and ongoing volatility. You’ll see in the technical analysis that although I could not be more bullish on this stock long-term, there is weakness in the semiconductor sector and we hope this translates to a lower entry point for our readers.
The market is also in a fierce debate between AMD, Intel, and Nvidia and is also distracted by other chips, such as Micron and NXP. In my analysis, I look for growth. How big is the market relative to how big the company is now?
You can ignore Nvidia’s gaming revenue and other segments for the main trajectory that we are focused on. Gaming is great for stability and earnings reports, but the growth will not be from gaming (a market where Nvidia is already a mature, market leader). I’m also not focused on PC sales or the CPU-powered cloud, as the first is not a growth market and the second is not the piece in the cloud stack that will accelerate future technologies.
I’ve written at length about the Mellanox acquisition and it’s a great reference for Nvidia’s long-term strategy. In this report, I’d like to break down what the GPU-powered cloud is capable of and why it’s important to differentiate Nvidia’s strategy from the competitors (and some competitors to keep an eye on).
SECTION 1: Artificial Intelligence
Artificial intelligence is a collection of categories, including computer vision, natural language, virtual assistants, robotic process automation and advanced machine learning. The AI impact will not be linear, rather adoption will resemble an S-curve pattern with a slow start due to the substantial costs and investment required for applications. The slow start will be followed by an acceleration that is driven by competition across capabilities and innovations. The penalty is steep for laggards, as pointed out below.
The slow start to AI will cause many investors to become complacent, not realizing the artificial intelligence boat will quickly leave the shore, metaphorically speaking, once it is closed to new entrants. If you add in the inevitable recession that will follow this long bull market, we could see many investors rotate back to growth stocks too late to realize the full gains from AI.
As noted in the graph below, we should see an AI acceleration around 2022-2023. The last call for decent gains in AI will be in 2025, although by then, the gains will be somewhat diminished compared to investors who choose their AI stocks by 2021/2022.
Basically, weigh the costs adding artificial intelligence to your portfolio earlier as opposed to later in the tech cycle as it’ll be one of the biggest economic growth drivers in history (more on this below).
Nvidia’s acceleration may happen one or two years earlier as they are the core piece in the stack that is required for the computing power for the front-runners referenced in the graph above. There is a chance Nvidia reflects data center growth as soon as 2020-2021.
Between the years 2025 and 2030, the stage after AI infancy, artificial intelligence is expected to add $13-$15 trillion to global economic activity, or 1.2 percent additional GDP growth per year. Compare this to the spread of information technology (IT) in the 2000s, which added 0.6 percent. Also, compare this to 5G technologies, which are expected to add $2.2 trillion over the next 15 years.
McKinsey points out that front-runners, who are currently investing in artificial intelligence, will reflect an increase in positive cash flow of up to 120% from their AI investments. The capital required to invest in AI, however, is negatively affecting cash flow right now and will continue to do so throughout the next year or two. Laggards are expected to lose at least 20% of their cash flow from the negative impact of investing in AI too late (or not at all).
Most investors today are well aware of what mobile and cloud did for tech stocks. These gains will seem minor in comparison to what artificial intelligence will do for your portfolio by choosing the right companies.
Today, Nvidia is my top pick for AI. I will also be providing more AI stock picks throughout the next 1-3 years to ensure my readers are well prepared for the massive gains AI will deliver by 2030. I believe this report is being delivered before at least one more pullback in Nvidia’s future due to broader semiconductor weakness.
It is too early for the data center to make an impact and this trajectory (data center) is what we are targeting. When I deliver information well before momentum, it helps to be patient with an entry and one of the main benefits of my reports is to give you that time, when possible.
SECTION 2: Competitive Positioning
Desktop GPUs is not the growth category that I am targeting for this investment thesis, which is why many oftcited statistics are irrelevant. For example:
“AMD shipments increased 9.8%, Nvidia was flat and Intel's shipments, decreased -1.4% as indicated in the following chart.” -John Peddie Research
Many financial analysts and authors on Seeking Alpha (etcetera) are quick to think this means AMD is the better investment, whereas this statistic refers to a mature market. To cut through the noise, it’s important to remember this thesis is about the GPU-powered cloud.
I’ve already covered why AMD is not as big of a threat to Nvidia as Wall Street believes. AMD has its hands full competing with Intel on the CPU-powered cloud and does not have the CUDA programming platform (more on this below). The two are competing in other segments (gaming, PCs) but those are less of a concern to the buyand-hold thesis in this report (data center). I can’t stress enough to separate these segments if Nvidia is of interest to you as a growth story.
Competitors to watch for at this layer in the data center stack are:
• Xilinx’s FPGAs,
• Intel’s FPGAs (through the acquisition with Alterra),
• Google’s TPUs (essentially an ASIC on the efficiency/flexibility spectrum).
FPGAs have distinct advantages over GPUs as they offer a higher amount of on-chip cache memory to help reduce the bottlenecks from external memory, and are flexible enough to be reconfigured for various data types, such as binary, ternary, and custom data types, whereas GPUs must be modified at the vendor level.
FPGAs are also known for power efficiency and test at 10x better in power consumption than GPUs and also 4x better than GPUs for general purpose compute. Reconfigurability for FPGAs help provide efficiency beyond deep learning for a large number of end applications and workloads.
The architecture of FPGAs are very adaptable as the chips allow a user to address all of the needs of a workload with the resources provided by FPGAs. Meanwhile, GPUs are restricted as the architecture is a Single Instruction Multiple Thread (SIMT), which provides an advantage over CPUs but can result in lower performance efficiency.
Today, FPGAs require knowledge of machine learning algorithms at the hardware level, in addition to the software development, and this is the barrier to entry for FPGAs.
As readers of mine know, I like Xilinx and this will be a stock I cover in the future with a full-length report. The company operates in a niche, is the inventor of FPGAs and has the ability to attract developers to its ecosystem. The challenge with FPGAs is they are hard to program as most software developers are not able to program hardware. Xilinx is working on becoming more of a platform company to solve this issue, and if the company succeeds, it’ll be a worthwhile investment.
Intel will face headwinds with developers, who are the ultimate decision makers for any ecosystem. Even now, you will be hard pressed to hear much discussion on developer forums and news feeds about Intel/Alterra’s FPGAs. Developers tend to avoid overly-corporate companies and cultures, and Xilinx has a serious shot of overcoming Intel if they execute correctly.
AMD also has a decent chance of eating away at Intel’s market share on the CPU-powered cloud. Overall, I prefer pure play options, when possible, and most of my tech stock coverage focuses on this. In my opinion, Intel is not the growth story in these categories.
This brings us to Google’s TPUs. TensorFlow is rising in popularity as a machine learning language and TPUs primarily run TensorFlow models. This is one of Google’s more successful experiments. They are cheaper and use less power than GPUs and are specifically focused on machine learning.
TPUs train and run machine learning models and power Google Translate, Photos, Search, Assistant and Gmail – i.e. image recognition, language translation, speech recognition and image generation. TPUs do not compete with GPUs in other areas of artificial intelligence.
It’s also important to remember that Nvidia and Xilinx are hardware companies that offer platforms for software developers. This is a distinct advantage compared to software companies (Apple, Google and Facebook) trying to release hardware chips. The market is so valuable, that they will most certainly try, but I think there are a lot of technical hurdles for a software company competing in the chip space other than Google. Workday’s cloud financial management solutions have less traction with 8 companies in the Fortune 500 and 530 customers overall.
SECTION 3: Developer Ecosystem
In November of 2018, I wrote about Nvidia’s developer ecosystem as a primary moat. GPUs are hardware which require software to write applications and utilize GPUs. Nvidia has a special language called CUDA that is universally known due to a first mover advantage in GPUs.
This ecosystem is not apparent to the public markets right now because new technology is developed in waves, and funded by venture capitalists in cycles. We are seeing the last of the mobile era of venture-funded companies with the IPOs of Uber and Lyft, — which began with Twitter, Yelp, Spotify — and was also reflected in Facebook’s epic rise from mobile native app revenue.
We are in the later stages of the venture-funded cycle for cloud software, hence a string of newly public companies over the last two to three years with some runway to go in this category before the majority of use cases are claimed. For artificial intelligence, it is so early that it’s essentially invisible right now to the public markets as development teams are beginning to form.
The strength of the developer ecosystem is what propelled Apple to become a $1 trillion company. While many investors look at iPhone sales, and Mac sales, the ecosystem that created by application developers is why Apple had an impenetrable moat. If the iPhone only had applications from Apple on the device (iTunes, iOS Maps, Safari browser), then many device manufacturers could have competed with Apple. The moat that Apple has enjoyed was created by the third-party developers who created iPhone applications in C and C++ with XCode, which made the device more attractive due to the mobile app ecosystem.
Android then became the second operating system in the mobile duopoly. Due to the friction of learning too many languages, the mobile ecosystem did not entertain any further competitors. This is despite there being 5 billion smartphones globally (i.e. it’s certainly feasible from a consumer supply/demand view point to entertain more operating systems and app stores), yet the limitation came from the number of languages developers are willing to learn. Microsoft Windows failed because it launched too late, and developers had already chosen the two languages they were willing to work with.
This is what is meant by developer ecosystem. Devices themselves do not have moats. The developer ecosystem creates the moat as third-party developers favor developing on certain operating systems and there is a limit to the programming languages they will learn before it impedes progress for the developer and the company the developer works for.
This is what is happening with Nvidia’s CUDA. The chips themselves do not create the moat. The compute platform creates the moat. Due to the need for a universal language to build GPU-accelerated applications, universities are teaching CUDA, and students are graduating knowing Tesla/Volta chips over competing chips, such as AMD’s Radeon or FPGAs or TPUs.
Here’s a quote from Marc Andreessen of Andreessen-Horowitz, one of the most successful venture capitalists in Silicon Valley: “We’ve been investing in a lot of startups applying deep learning to many areas, and every single one effectively comes in building on Nvidia’s platform. It’s like when people were all building on Windows in the ’90s or all building on the iPhone in the late 2000s.”
Here's another quote from a developer on Reddit:
“Nvidia, thanks to the CUDA software stack (which AMD cannot match), has a much more unassailable position than does Intel with Xeon CPUs (where an X86 application just runs on either a Xeon or an Epyc).”
– software developer on Reddit
SECTION 4: Financials
In this case, I began with the investment thesis rather than the financials as the two do not sync up today. Gaming is Nvidia’s strongest revenue segment with $1.3 billion per quarter. Data center revenue has been flat to declining for three straight quarters, ranging between $634 million and $679 million.
The market is encouraged, yet cautious, with revenue of $2.58 billion in the most recent quarter fiscal Q2 2020, up 16.2%, yet down about 17% from $3.12 billion in the year-ago quarter.
GAAP earnings was $0.90, compared to $1.76 a year ago, and $0.64 in the previous quarter. Non-GAAP earnings of $1.24 in the current quarter, compared to $1.94 a year earlier, and $0.88 in the previous quarter.
Similarly, net income was down 50% year-over-year but up 40% quarter-over-quarter at $552 million.
Next quarter, Nvidia is expecting $2.90 billion, plus or minus 2 percent, with gross margins of 62%.
The bigger story this quarter was Nvidia’s gaming growth, which reflected 24% growth sequentially, with revenue of 1.31 billion compared to estimates of 1.29 billion. There is also evidence that inventory is normalizing with an inventory ratio of 50% and on track to lower to around 40% in fiscal Q3 compared to the previous ratio of 71%.
As stated, the current financials do not reflect the growth expected from AI.
SECTION 5: Technical Analysis
By Knox Ridley
5.1 Trend Lines and Internal Strength
Focusing on the black trend lines, you’ll see three sets. The upward trend in price, which coincides with the upward trend in momentum in the RSI, tracks two recent topping patterns in Nvidia’s price action.
The trendlines coincide with each other. When both momentum and price roughly trend together, it can show a healthy trend in place. Using these corresponding trends can also offer reasonable warnings, as well.
By following the approximate time at which both the RSI trend and price trend broke to the downside together, we can see safe and effective exits that allow you to side-step pull backs.
Further information can be found by looking at the set of green and red arrows. Notice that just before the last top in Nvidia, before the May correction, as the price of Nvidia climbed higher, the momentum in the RSI was decreasing. So, when I see a divergence like this coupled with trendlines being violated, it’s a warning of a correction on the horizon.
This exact pattern is unfolding today if we look at the last set of trend lines in the price and RSI. With Nvidia’s RSI closing just below oversold levels and pointing down towards the black trendline, this same negative divergence pattern is unfolding in real time, signaling a weakening of momentum. A break in this trend on both the RSI and price, and we can expect more downside to follow.
5.2 Elliot Wave Analysis
If we are forming a double top pattern, and Nvidia fails to break out above the $200 barrier, we can look at the retrace levels and expected extensions to gauge the likely targets for entry. First, if we zoom into the internal structure of Nvidia’s 3-month drop, which is highlighted by the light blue roman numeral count, you’ll see a very clear 5-wave drop.
It’s hard to see anything other than 5-waves in this move, all of which line up with internal Fibonacci ratios. The first pattern in a correction being a 5-wave drop is a strong indication of a 5-3-5 corrective pattern, which is highlighted in purple.
So, the Purple A was a 5-wave drop, while the purple B was a 3-wave correction, which touched the 50% retrace of Wave A before falling. If this count is correct, we are just now completing the 2nd wave in the final 5-wave pattern.
It’s always worth noting how the stock price reacts to these extensions. The stock found major support at the 61.8% retrace, testing this level before meeting heavy resistance at the 38.2% retrace. If Nvidia cannot break through the 38.2% retrace, the analysis is suggesting that we could see the final C wave play out, which will have Nvidia retest the 61.8% retrace level and likely make new lows.
However, I want to be clear with our convictions in this position. Anywhere between $125 to $160 is a great entry. Even though the analysis in this count is suggesting that we could see lower lows, and correction will be welcomed for a long-term entry. Thus, we will update any entries if the pull-back scenario takes hold.
Due to Nvidia’s fundamental strength, we will now look outside the individual stock for the cause of a sentiment shift. Please also refer to the bullish scenario below.
5.3 Global Semiconductor Sector
Like most semiconductors, Nvidia is a global company that is manufactured overseas (mainly from Taiwan) with 44% of its sales coming from China. Semiconductors are cyclical and sensitive to economic cycles. This is why we must look holistically at this sector when guiding an entry.
The Korean KOSPI Index
South Korea is an economy that is fueled by some of the world’s largest semi-conductor companies, as well as many mid-level players. Companies such as Samsung, and SK Hynix supplied over 60% of the components used in memory chips sold globally in 2018. So, the KOSPI can provide more information about the global health of semiconductors.
Since 2011, the index has been in a long-term uptrend, which it respected until very recently.
The KOSPI broke through this trendline, highlighted in black, which coincides with the 61.8% retrace. This level is now acting as resistance as the KOSPI is showing a negative RSI reversal pattern, which is indicating more downside is likely to follow.
This pattern is highlighted by the blue circles. In short, as the price makes lower highs, the RSI is making higher highs, indicating that the buying pressure is not sufficient to reverse the trend as it reaches oversold levels. This chart is anything but encouraging.
Philadelphia Semiconductor (PHLX)
The Philadelphia Semiconductor index (PHLX) will have some international exposure, such as NXP Semiconductors and Taiwan Semiconductor Manufacturing Company, but it is populated with mostly US companies. Nvidia accounts for nearly 9% of its total value.
Looking at the weekly chart, we can get a glimpse of the bigger patterns at play, which I believe are pertinent for where we are. If we start with the first long term trend from 2013 – 2015, you’ll notice that the price respected the first long-term trend line in magenta.
The black arrows indicate the moment when the RSI and price broke their respective trends. This trend broke first on the RSI, then fell to a lower long-term trend highlighted in black. Once this trend broke along with the black trend in the RSI, the index gave way to a significant correction. When these long-term trends break together, both in momentum and in price, it’s a warning to step aside.
Further evidence can be found by the descending red line in the RSI leading up to the drop. There is a negative divergence with the price and momentum. As the price action increased leading up to the sell-off in 2015, the RSI made significantly lower highs, failing to break out multiple times. When you see negative divergence that is followed by the RSI and price trend lines breaking, this is where technical analysis can help you avoid downside risk.
Today, the price has managed to find a bottom, and has resumed a new trend, which is also highlighted in black, and appears to be trading in a diagonal pattern. The RSI is showing a divergence between the RSI, making lower highs while the price makes higher highs, just like we saw leading up to the 2015 correction. Furthermore, just like in the prior run-up before the correction, the RSI’s momentum keeps failing to break out of the descending trend highlighted in red.
Taking what we are seeing today, the evidence is leaning towards a more cautious stance. The intermarket divergence we are seeing between the KOSPI and the PHLX, coupled with the weakness we are currently seeing in the PHLX, leads me to conclude that the KOSPI is a possible leading indicator of the semiconductor sector.
Nvidia, being a major global player in the semiconductor space, is not immune to this broad market weakness. With the weakness we are seeing with Nvidia’s chart as well, I think letting these patterns play out, will allow a safer and more optimal entry price.
5.4 The Bullish Scenario
I attempt to approach each chart from a blank slate. I let the data and analysis lead me with as little bias as possible. However, especially in a late cycle bull market, I always ask myself where I could be wrong, and what it would take for me to invalidate my primary thesis. The below chart is my alternative invalidation thesis.
From an Elliot Wave Count, the primary Wave 4 correction, which is highlighted in yellow, unfolded in an A-B-C pattern, highlighted in purple, and ended at the December low. This would mean that we are currently in the final 5th Wave push of the larger yellow count. This final 5th Wave will be an impulsive move, so its structure will have its own 5 waves. This means that we have completed the Wave 1 and Wave 2, highlighted in purple and are currently in Wave 3. If we look one degree lower into Wave 3, we have completed waves 1 and 2 of 3, and are about to breakout to the upside.
I have some issues with this count. Specifically, when you zoom into the lower degree structure, we can see a 5wave structure down, who’s ratios line up like we want to see, and we can also see 3-waves up. However, if we see a break above $200, this strongly implies that the Wave 4 in yellow is over, and that the bull trend can continue.
I will need to see the RSI divergence forming to be invalidated by the RSI breaking through the red trend line as well as the price of Nvidia break through the $200 level. At $200, we may have left some minor gains on the table, but it is a much safer entry than where the stock is trading right now and worth the insurance.
Nvidia fell off a cliff last October from a high of $290 to a low of $130. Meanwhile, the challenger Xilinx remained unharmed by the tech rout, and despite unfavorable macro conditions. Nvidia popularized GPUs in 1999 and Xilinx invented FPGAs in 1985, and both are chips that will define the computationally-intensive future.
GPUs originated from the advanced computations required in gaming and FPGAs originated from electronics engineering. There are strengths and weaknesses to both, however, these are the two that will power the artificial intelligence and machine learning-driven economy. The size of this AI and ML economy is expected to reach $15 trillion by 2030 up from $2 trillion this year.
Keep in mind, that long before technologies go public, they are incubating across the startup ecosystem. By the time AI and ML companies reach the public markets, the technology powering and developing this wave of companies was already decided in the years prior. We are in those critical years where startups must quickly design and develop AI if they want to have the first-mover advantage. This is creating a battle between FPGAs and GPUs.
Below, I break down the differences between Xilinx’s FPGAs and Nvidia’s GPUs before analyzing the financials and theories on how the two will perform in the future.
Note: Previously, I discussed how Nvidia stock has two impenetrable moats: the developer ecosystem and GPU-powered cloud. This previous analysis was written during the height of the panic sell-off, which I negated as being overly-pessimistic due to Nvidia’s strong fundamentals.Nvidia stock has two impenetrable moats: the developer ecosystem and GPU-powered cloud. This previous analysis was written during the height of the panic sell-off, which I negated as being overly-pessimistic due to Nvidia’s strong fundamentals.
AI and Machine Learning
On many technical levels, FPGAs (Xilinx) are considered superior to GPUs (Nvidia). They offer a higher amount of on-chip cache memory to help reduce the bottlenecks from external memory, and are flexible enough to be reconfigured for various data types, such as binary, ternary, and custom data types, whereas GPUs must be modified at the vendor level.
FPGAs are also known for power efficiency, and often test at 10x better in power consumption than GPUs and also 4x better than GPUs for general purpose compute[1]. Reconfigurability for FPGAs also helps provide this efficiency beyond deep learning for a large number of end applications and workloads. The architecture of FPGAs is very adaptable as the chips allow a user to address all of the needs of a workload with the resources provided by FPGAs, such as reconfiguring the data path during run time and with partial reconfiguration. Meanwhile, GPUs are restricted as the architecture is a Single Instruction Multiple Thread (SIMT), which provides an advantage over CPUs but can result in lower performance efficiency when enough parallels cannot be found while mapping the workload.
As pointed out in my previous analysis on Nvidia, software developers prefer GPUs as their frameworks are easier to develop on. Nvidia’s CUDA architecture, for instance, does not require an in-depth understanding of underlying hardware. FPGAs require knowledge of machine learning algorithms at the hardware level, in addition to the software development, and this has been a barrier to entry for FPGAs. FPGAs are a reconfigurable integrated circuit (hence the strengths on being easily reconfigured), which requires specifying a hardware circuit, whereas GPUs are configured via software[2].
“Nvidia, thanks to the CUDA software stack (which AMD cannot match), has a much more unassailable position than does Intel with Xeon CPUs (where an X86 application just runs on either a Xeon or an Epyc).”
– software developer on Reddit
Section takeaway: FPGAs result in faster and more efficient compute but are harder to program due to hardware circuit configurations when compared to GPUs for machine learning, which are more universal and require less engineering resources.
Financials
Nvidia and Xilinx power more than data centers, of course. Nvidia’s top revenue segment is gaming, the origin of GPUs, and this drives about $1 billion per quarter in revenue. Xilinx’s top segment is Communications with many investors using Xilinx as a global bet on 5G with communications revenue increasing 41% year-over-year as reported in the most recent quarter. Xilinx also was not as affected by crypto as the Broadcast, Consumer & Automotive category was 17% of revenue compared to 15% of revenue in the same quarter YoY. (Xilinx classifies crypto as consumer in this 10-K).
Xilinx has a direct competitor with Intel, who acquired Alterra for $16.7 billion. Intel is keen to solve the development uptake issues with FPGAs with the release of Stratix 10 hardware, which has a software layer to simplify development. Microsoft Azure is partnered with both Xilinx and Intel/Alterra on FPGAs although there is some indication that MS is leaning more towards Xilinx in the near future after announcing they will replace Intel chips with Xilinx in over half of their servers.
Developers favor Xilinx over Intel as a brand, and Microsoft is doing quite a bit to court developers right now including the acquisition of Github – read more tech stock analysis here. Therefore, the shift towards Xilinx was not unexpected.tech stock analysis here. Therefore, the shift towards Xilinx was not unexpected.
Nvidia:
While Xilinx reported double digit increases, Nvidia reported double digit declines with revenue down 24 percent, earnings per share down 48 percent to $0.92 and operating income down a shocking 73 percent year-over-year in fiscal Q4. The annual numbers ended on a better note with revenue increasing 21 percent to $11.72 billion, and GAAP earnings per share increasing 38 percent to $6.63. Of Nvidia’s revenue segments, gaming was hit the hardest due to the crypto bust flooding the market with GPUs, which in turn, caused reduced unit shipments overall. In addition, the new Turing architecture and real-time ray tracing, while impressive from a graphics perspective, are ahead of their time and are seeing slow adoption (At release, I had originally put Q3 2019 for these to find early adopters and this timing still looks accurate or maybe Q4).
This upcoming quarter is not likely to be the comeback quarter for Nvidia with guidance of $2.20 billion, which is flat from last quarter and represents a 31 percent decline year-over-year. As you’ll see in the takeaway paragraph below, I am very bullish on Nvidia in the long term as crypto causing temporary GPU saturation offered an opportunity to enter the stock below its value.
Gaming is a foundation for Nvidia, but most certainly, this is not the growth story. The GPU-powered cloud is the future due to AI and ML. If you can get Nvidia below a $100 billion market cap, then my prediction is you will be resting easy by 2022 and 2023 with a stellar return as it’s understated presence across cloud data centers and AI applications should have a firm hold on the market.
Xilinx:
Xilinx’s revenue growth is at 34% year-over-year in Q3 2019, with 63% growth in operating income YoY in the same quarter, and 42% net income growth. It’s important to mention that Xilinx is a small fish in a big pond and this quarterly growth of 34% and 42% equals $200 million to the top line and less than $100 million to the bottom line. Meanwhile, Xilinx commands a PE ratio of 38, at time of writing.
Guidance for the upcoming quarter is revenue of $815 to $835 million compared to $800 million in the previous quarter. One reason Xilinx’s stock price continued to climb, while Nvidia fell off a cliff, is that the smaller fish did not have enough market share to reflect a big impact, whereas Nvidia’s crypto business alone exceeded Xilinx’s net income for the entire year (at around $500 million per quarter). In addition, one year ago Xilinx posted negative net income of $12 million but is now at a net income in the range of $200-$250 million the last two quarters.
In other words, Xilinx is more of a trout than a tuna, but is a pure play option that is likely to see very solid returns as the AI economy is built out. (This is why I don’t invest in Intel; I prefer pure plays when possible).
Nvidia is one of my favorite companies from a fundamental standpoint, and it is worth repeating that I was not fair weathered during the crypto bust, rather encouraged readers to look at the developer moat and GPU-powered cloud as future drivers of growth. As I stated to a reader over email two days before the Mellanox acquisition: “Can Xilinx’s FPGA disrupt Nvidia GPU’s at 4x faster? My best guess (and it’s only a guess) is that Nvidia will continue to release the right chips that the market demands.” In this case, Nvidia is acquiring the right company that the market demands. You can read my analysis on Mellanox acquisition published on FATRADER here.
I want to point out that Xilinx will make a solid investment, as well. Xilinx is priced a minimum of 25-30% higher than Nvidia when looking at PE ratio, Price to Sales, and EPS. Quarter-over-quarter growth for Xilinx right now is in the single digits, and for this reason, I’d like to see Xilinx priced 20% cheaper before I build a position or I’d like to see more than single digit QoQ revenue growth in a highly competitive market for a 30+ PE ratio. Due to Nvidia’s upcoming flat quarter (per guidance), Nvidia is also likely to trade sideways for a quarter or two. I bought Nvidia in 2017 and cost averaged down to $160, and am comfortable here for the long term.
Tech stocks are getting slammed right now, and Nvidia may be one of Wall Street’s biggest losers in the sell-off that began last month and continued into this week. Nvidia’s stock has seen a 30-day high of $292 and a whiplash low of $176 – equaling a 40% plunge in the matter of four weeks. Today, it stands at $197.60.
Economic indicators and earnings from tech companies have not exactly warranted this reaction from the market. Fears the semi-conductor industry is slowing down based off Advanced Micro Devices earnings report were negated when Intel reported strong Q3 earnings. And while Apple may be on the precipice of a capped out $1 trillion-dollar market cap due to possible iPhone saturation, Nvidia’s outlook is quite the opposite in regards to public-company growth trajectory. The market may continue to have volatility, but Nvidia investors who are patient will be rewarded due to competitive advantages in GPU-powered cloud performance and developer adoption of Nvidia’s platform.
Brief Overview of Nvidia’s Revenue Segments
To summarize, gaming claims the majority of Nvidia’s revenue at $1.81 billion, up 52% YoY. Gaming will get a nice boost in 6-12 months from the new GeForce RTX 2070, RTX 2080 and 2080 Ti chips, which introduce the possibility of hybrid rendering through ray-tracing. In layman’s terms, ray-tracing mimics how light behaves in the real world by mapping out rays from 3D illumination sources. The imagery is much more realistic as a result. Electronic Arts released the first raytracing game today (November 14th) whereas 6 months ago, the gaming industry did not think raytracing would even be possible. Companies who have signed up for the new Turing architecture include Adobe, Pixar, Siemens, Black Magic, Weta Digital, Epic Games (maker of Fortnite) and Autodesk.
Data center revenue has been picking up speed at 83% YoY, or $760 million, as GPU chips are powering more of the cloud for machine learning and artificial intelligence applications. Data center revenue, once a small blip, claims 24% of the company’s total sales. This will continue to grow steadily into the near future due to the computing power and flexibility GPUs provide over CPUs, which is what Intel sells, or TPUs and FPGA, which are custom machine-learning chips by Google and used by Microsoft that are too specific to one platform for widespread adoption – more on these points below.
Source: TechCrunch
Smaller segments by Nvidia include professional visualization and automotive, which grew to $281 million and $161 million, respectively, up 20% and 13% year over year.
Two Impenetrable Moats: GPU-Cloud and Developer Adoption
Revenue segments are your typical Nvidia stock coverage. But can Nvidia take market share from Intel? Will Google, Microsoft, Facebook and Apple design their own custom chips to compete with Nvidia? This is what investors need to answer for themselves especially if we continue into correction territory.
Regarding Intel, the cloud is too competitive to forego the performance and efficiency that Nvidia delivers. Recently, the Turing T4 GPU became the fastest adopted server GPU of all time in just two short months of hitting the market. Prior to the release of the Turing T4 GPU, Nvidia’s data center growth was 3x compared to Intel. Intel posted 26% growth YoY whereas Nvidia posted 83% YoY. However, Nvidia’s data center revenue is 1/6th compared to Intel’s at $760 million vs. $6.1 billion. This revenue segment will continue to grow as the GPU-powered cloud is built out. Unfortunately for Intel, GPUs are the better choice for cloud customers as the usage pattern is constantly in flux, demanding a wide variety of models and different software frameworks. Intel’s CPU Xeon Processor cannot compete with the performance-per-watt of what Nvidia offers in the cloud. Per the announcement on September 13th, 2018, Microsoft, Google, Cisco, Dell EMC, Fujitsu, HPE, IBM, Oracle and Supermicro plan to release servers with Nvidia’s T4 GPU on board.
My newsletter subscribers get this information first. Sign up here.My newsletter subscribers get this information first. Sign up here.here.
Google and Microsoft have both made chips for their data centers. Microsoft adopted the field-programmable gate array (FPGA) which is used for AI apps. And Google has built a custom chip called the Tensor Processor Unit (TPU) for Google’s TensorFlow deep learning framework. Competing, customized chips will become the new norm as tech giants prefer to use proprietary tech. The biggest weakness that competing customized chips face like TPUs, from Google, and FPGA, used by Microsoft, is that they may be too specialized for developers to adopt. The drawbacks will continue to be price and difficulty, as programming for FPGA is an area not many engineers have expertise in. The same goes for Google Cloud Platform (GCP). They’ll have to get developers to adopt GCP and keep them locked into TensorFlow. Even so, there are alternate frameworks such as PyTorch from Facebook which add further to the fragmentation of developer frameworks. In addition, even if Google uses TPUs for inferencing, it may still use Nvidia’s GPU for training neural networks.
Let’s use mobile application development as an example. One of the reasons mobile is a duopoly between Android and iOS is that developers can only learn so many tools and development environments before the process becomes inefficient. In order to truly excel at a language, it has to be universal. For instance, Microsoft attempted to launch a Windows phone, which was met by resistance as developers did not care to learn a new operating system that could not prove itself with user adoption. In turn, mobile users did not buy the Windows phone because their favorite applications were not available to download. iPhone’s success was due to iOS developers who learned tools like XCode to create applications. Android became the competing universal language for the remaining manufacturers, such as Samsung, LG, Sony, Pixel, etcetera. The next wave of AI applications and machine learning inferences will follow the same path of limited competition due to development bandwidth. Developers will self-regulate the number of competitors for processing units due to a need for a universal platform that supports all frameworks.
Here’s a quote from Marc Andreessen of Andreessen-Horowitz, one of the most successful venture capitalists in Silicon Valley:
“We’ve been investing in a lot of startups applying deep learning to many areas, and every single one effectively comes in building on Nvidia’s platform. It’s like when people were all building on Windows in the ’90s or all building on the iPhone in the late 2000s.”
There is an even greater need to simplify artificial intelligence and machine learning than exists for mobile standards. There are thousands of variants emerging each year in AI as neural networks evolve and expand in depth, complexity and architecture. There are multiple frameworks supported by major industry players and Nvidia’s GPUs are flexible enough to accelerate all of these frameworks and workflows including Caffe2, Cognitive Toolkit, Kaldi, MXNet, PaddlePaddle, Pytorch and TensorFlow.
In addition, AI occurs beyond the cloud and Nvidia’s GPUs are available in what is called edge devices, such as self-driving cars, desktops, workstations, data centers and across all major cloud providers.
Conclusion
Nvidia is already the universal platform for development, but this won’t become obvious until innovation in artificial intelligence matures. Developers are programming the future of artificial intelligence applications on Nvidia because GPUs are easier and more flexible than customized TPU chips from Google or FGPA chips used by Microsoft. Meanwhile, Intel’s CPU chips will struggle to compete as artificial intelligence applications and machine learning inferencing move to the cloud. Intel is trying to catch-up but Nvidia continues to release more powerful GPUs – and cloud providers such as Amazon, Microsoft and Google cannot risk losing the competitive advantage that comes with Nvidia’s technology.
The Turing T4 GPU from Nvidia should start to show up in earnings soon, and the real-time ray-tracing RTX chips will keep gaming revenue strong when there is more adoption in 6-12 months. Nvidia is a company that has reported big earnings beats, with average upside potential of 33.35 percent to estimates in the last four quarters. Data center revenue stands at 24% and is rapidly growing. When artificial intelligence matures, you can expect data center revenue to be Nvidia’s top revenue segment. Despite the corrections we’ve seen in the technology sector, and with Nvidia stock specifically, investors who remain patient will have a sizeable return in the future.
