Introduction AMD (NASDAQ:AMD) shares have skyrocketed a stunning 134% in the last year, 48% YTD, and 595% in the last five years. When Lisa Su, AMD's
Tech News: Introduction
AMD (NASDAQ:AMD) shares have skyrocketed a stunning 134% in the last year, 48% YTD, and 595% in the last five years. When Lisa Su, AMD’s current CEO took over the company in October of 2014, the company was on the verge of bankruptcy, had no product roadmap to combat larger Intel (NASDAQ:INTC) and Nvidia (NASDAQ:NVDA), and was reliant on Intel to stay afloat. In October of 2014, AMD shares traded at ~$3. Since then, AMD shares have done this:
The company has transitioned itself out of near bankruptcy, and has focused on investing in strong growth opportunities and quality products. Such moves included investing into their Radeon GPU segment, and creating a whole new “Zen” CPU architecture to compete with Intel’s high performance offerings.
The result has been extraordinary, with AMD being one of the greater comeback stories of the last few years. Long-term, I can see continued price appreciation as AMD gains significant market share in both CPUs and GPUs.
Tech News: AMD’s Greatest Opportunity: CPUs/Ryzen
AMD’s phenomenal success over recent years all started really with the overhaul of previous x86 architectures in favor of a new start-from-scratch Zen architecture. For those that did not know, when AMD was nearing insolvency, competitor Intel (valued at ~$200 billion today) poured money into keeping AMD afloat, potentially to avoid anti-trust concerns from regulators.
In my view, AMD’s single largest opportunity is in the CPU market, competing with Intel. While Intel did keep AMD afloat in the early stages of the company’s turnaround, AMD is probably going to take significant share from Intel in the coming years.
Before Lisa Su took the helm at AMD, the company’s product strategy was to deliver lower performance, budget processors at cheaper price. While the margin was far lower, it helped AMD gain decent market share on Intel. So we got AMD’s process architectures like Bulldozer and Athlon. However, with AMD’s Zen architecture, the company transformed the technological story away from budget products and towards high price performance CPUs that, for a few years were competitive with Intel. Now however, I believe that not only is AMD competitive with Intel’s offerings, they surpass Intel’s offerings.
Let’s compare AMD’s recent R7 2700X offering against Intel’s i7 8700K CPU.
|Price||Speed (in GHz)||TDP (in watts)||Release Date|
|Ryzen 7 2700X||$270||4.3 GHz||105W||CQ2’18|
|Intel Core i7 8700K||$350||4.7 GHz||95W||CQ4’17|
Based on this table, it appears Intel’s i7 8700K processor outperforms AMD’s 2700X processor, thus deserving the higher price. And in general, Intel processors have outperformed AMD’s for some time. But lets look at the same comparison when it comes to specific tasks.
In every task, the 2700X is meaningfully faster than the 8700K. That is except for gaming, where the two processors are basically equal to each other. All of this for about three quarters of the price.
With AMD’s lower price Ryzen 3 series CPUs, we see a similar result, with AMD matching or surpassing Intel in speed, efficiency, and keeping power use low.
Outside of professionals and content creators however, the largest segment by far for CPUs is PC gaming. With the wildfire like spread of PC based battle royale games like Fortnite and PUBG, gaming and deciding which CPU is best has never been more relevant. Lets compare this same Ryzen 7 2700X to a higher priced, greater performance i7 9700K, the next generation above 8700K. Here is a screenshot of a video I found on YouTube showing the differences in speed, energy consumption, and work that each CPU is doing during gameplay of a car racing game, Forza Horizon.
(source: YouTube video)
The video is about 12 minutes long, and shows the performance differences between the two CPUs across a variety of games. There are a few things that remain consistent across all of the games.
First of all, the 9700K is always faster than the 2700X. In this case, the 9700K operates at 173 FPS (frames per second) versus the 2700X’s 149 FPS, a staggering 16% performance difference. We see similar spreads in other games with different digital environments and workloads. However, there are three other factors that skew a positive result from the 9700K.
These factors have to do with the work rate of CPU and GPU, and the temperature of the CPU. As you see, the GPU temperatures are roughly the same, at 57 degrees celsius and 58 degrees for the 2700X and 9700K respectively.
For the CPU however, the 9700K is operating at 51 degrees, while the 2700X is operating at 46 degrees. Because the 9700K is working so hard to produce this result, the CPUs temperature is far higher, making the processor less efficient.
On top of the temperature difference, there is a much more staggering difference. Look at the work rates for the CPU and GPU. The GPUs in this case are the exact same for both CPUs. However, the GPU associated with the 2700X is working at 84% of its peak potential processing power, while the 9700K’s is working at 96%. The CPU’s spread is even more staggering. The 2700X is working at 36% of its potential processing power, while the 9700K is working at 62%.
Because the Intel CPU is working so much harder to process the same gameplay, it is far less efficient, regardless of FPS numbers. For gamers considering competing options, FPS isn’t the only metric. Temperature, work rates, price, and memory all play into the equation.
Considering all this, and understanding Intel’s 9700K comes at a $400 price point, we can see the value proposition for gamers for AMD’s Ryzen chips. The 2700X sells for just two thirds of the 9700K.
My conclusion on AMD’s processors is the following. AMD makes lower price, faster, and more efficient processors than Intel.
If this is the case, then why is it that AMD’s market share in x86 notebooks is only 13.3%? Wouldn’t the better performing technology have greater market share?
The answer is, AMD is making impressive strides in catching up to Intel. Just a year ago, AMD was at 8.6% market share. Today, it is at 13.3% share. The reason AMD is not the leader in the market has to do with one key word: marketing.
In 2018, Intel spent billions of dollars on marketing its products. During football games, I remember personally seeing multiple Intel ads for their next generation CPUs. Intel’s brand identity and trust, and a lack of consumer knowledge regarding AMD is what has allowed Intel to stay ahead in market share. However, as AMD sees increases revenues and profits, they are more likely to invest into marketing, increasing customer knowledge.
As important, if not more so, is AMD’s lead in process node technology. Intel’s 10nm struggles were well publicized in 2018, with the struggles being a large cause of the stock’s decline in value. And while Intel has been recovering recently, I still believe AMD is multiple quarters, if not years ahead of Intel. As a matter of fact, AMD ditched GlobalFoundries their long-time manufacturing partner in favor of TSMC. Why? GlobalFoundries announced they were pulling back on plans to manufacture on the 7nm node, in favor of the 12nm and 14nm nodes. TSMC on the other hand has been full steam ahead on 7nm.
AMD’s 10nm processors are already on the market, while Intel’s 10nm offerings are expected to hit shelves in 2H’19. These 10nm Intel products are expected to enter volume production by 2020. Meanwhile, AMD’s 7nm Ryzen 3000 chips are expected to hit the market sometime this quarter. So, I believe purely from a process node standpoint, AMD is far ahead of Intel.
As a matter of fact, let’s bring up that same 9900K CPU I talked about earlier, and compare it to the Ryzen 3000 chip demoed at CES 2019 that will launch this quarter.
This Ryzen 3000 chip had a Cinebench score of 2057, whereas the 9900K, Intel’s best chip, ran at 2040. In addition, the Ryzen processor ran on 133.4 watts of power versus the 9900K’s 179.8 watts. The Ryzen 3000 delivered slightly better speeds at three quarters the level of power consumption. This alone is extremely impressive, showing AMD’s lead in performance processing.
In addition, lets look at what TSMC, AMD’s new manufacturing partner has had to say about 7nm vs 10nm. TSMC manufactures on a manufacturing process known as FinFET. Compared to 10nm, TSMC expects 1.6X logic density, 20% speed improvement, and most importantly, a 40% reduction in power.
So, 7nm is a big deal. The question is, how much of an impact will it have on AMD’s business. Since Ryzen 3000, the 7nm CPU is expected to begin sales this quarter, we should see strong back-to-school demand in Q3, and strong gaming/holiday sales in Q4. Right now, I can see AMD market share climbing to 15-16% as Intel struggles to get 10nm offerings on the market until 2020.
However, AMD is not stopping at 15-16%. If AMD is ahead of Intel by a couple of years, we could see a long-term market share gain towards a quarter, a third, maybe even half of the market. For the last several years, AMD has been a small, budget, non-competitive player in CPUs. Now however, with AMD taking a multi-year lead, the company should see significant market share grabs. First, we have to look at the TAM of the PC CPU market.
This is a slide from AMD’s Q2’2018 earnings presentation.
This came from their Q2 presentation, and doesn’t include an estimate for 2019 TAM and hasn’t been updated in most recent earnings reports. Since then, it seems the PC market has cooled slightly mostly because of Intel’s inventory overhang. So, I’ll revise this number down to $29.5 billion by 2020. And I’ll guess that the 2019 market size is $28 billion, growing ~5.3% from 2019 to 2020.
On $28 billion, assuming 15-16% market share, we get revenues between $4.2-$4.48 billion for this year.
|Ryzen’s Value||TAM||Market Share||Revenue|
|2019||$28 billion||15% (up from 13.3% currently)||$4.2 billion|
|2020||$29.5 billion||17.4% (7nm, 7nm+)||$5.133 billion|
|2021||$30.5 billion||18.6%||$5.673 billion|
|2022||$31.5 billion||19.5%||$6.142 billion|
|2023||$32 billion||20%||$6.4 billion|
Tech News: AMD In The x86 Datacenter:
While AMD’s PC performance should be impressive over the coming years, even more impressive than that is their performance in the server CPU space over the coming years.
Years ago, when AMD was a major player in the server CPU space, their market share hit ~26%. Since then, Intel has reclaimed basically the whole server market, as AMD abandoned their server push. The continuously growing server market was Intel’s for the taking. That was until AMD announced EPYC in June 2017, nearly two whole years ago.
Just like the x86 PC industry before it, the x86 server market was extremely uncompetitive, with Intel offering high-end products and AMD occupying the low-end. Actually, it was worse than that. AMD didn’t operate in the market at all. Intel was 100% of the market. Instead of operating in a market niche, Lisa Su has made clear from day one of EPYC that the company’s goal was to deliver high quality products. EPYC was going to attack the high end. EPYC was going to attack Intel.
Before we talk about the arguably most important part of the EPYC story, Rome and Milan (7nm and 7nm+), lets talk about first generation EPYC, aka Naples.
Naples started at 32 cores and 64 threads. Comparing Naples to Skylake Xeon however, we see some interesting results. EPYC is deemed a more flexible alternative to Skylake in terms of PCIe utilization (Infinity fabric design). In addition, AMD’s EPYC has a far higher memory capacity than Intel’s Skylake offerings.
Something interesting I noticed watching a demo from EPYC’s initial 2017 launch was that EPYC saw strong memory capacity relative to Intel’s offerings. AMD’s EPYC offerings could process a dataset at the demo with three servers. It took Intel Broadwell four servers to process the data set.
In the demo, we see three AMD Naples servers go head to head against four Intel Broadwell servers. Despite the fact that Broadwell had an entire additional server, AMD’s six sockets outperformed Broadwell by ~12%. This is all impressive work. As a result, AMD has claimed multiple Super 7 cloud computing customers (the largest public cloud providers on the planet) for its Naples offerings alone.
In addition, AMD is powering the Frontier system, the world’s fastest supercomputer. AMD is flexing their technological muscles with this deal, one-upping Intel.
All of this technical progress from AMD has lead the company to gain ~3% of the server market. This is an extremely minuscule portion of the market, with Intel, despite all its slip-ups and AMD’s better technology, commanding 97%. This is where it gets really interesting though.
The next generation EPYC, built on the Zen 2 architecture previously talked about, should accelerate AMD’s market share gains. This next generation architecture is called Rome, and is manufactured on the 7nm process node.
The gap between Rome and Intel’s recent best in class Skylake server processors is even larger than the gap between Naples and Intel’s previous Broadwell architecture. Let’s look at AMD’s CES 2019 EPYC presentation. This was the same presentation in which AMD unveiled the Radeon VII.
In the presentation, we see two of the best of the best Intel Xeon processors (dual socket architecture) against one AMD EPYC Rome processor.
It took 28.1 seconds to render the image using the single AMD Rome processor, while the two Intel Xeon chips took 30.2 seconds, a ~7% speed difference. Consider the fact that there is only one Rome chip, versus the two Skylake chips and we can see, regardless of price, that raw power alone is a significant value proposition for datacenters. Then, considering the cost savings associated with having only one processor, and the value proposition for Rome is further enhanced.
Not only that, but AMD’s time to market for their 7nm chips is much better than Intel’s. AMD is expected to begin producing and selling their 7nm Rome processors in Q2, with the ramp really kicking in 2H. Meanwhile, Intel will not push out a 7nm server processor until the beginning of 2021. If AMD sticks to their current roadmap, we will be on to EPYC’s third generation Milan processors by the time Intel gets 7nm to market.
Because of the faster speeds, solid memory capacity, lower latency, and lowered pricing, AMD should find its way into other major Hyperscalers. With the hyperscalers AMD is already in, AMD should take a significant share of the x86 server market.
|EPYC’s Value||TAM||Market Share||Revenue|
|2019||$21 billion||5.4% (up from 2.9% currently)||$1.134 billion|
|2020||$26.5 billion||8.2% (Rome ramp, Milan ramp)||$2.173 billion|
|2021||$29 billion||10.8%||$3.132 billion|
|2022||$31 billion||13.2%||$4.092 billion|
|2023||$32 billion||16.0%||$5.12 billion|
This assumes AMD still gets nowhere close to their all-time high market share of ~26% achieved in 2006. It assumes we get continued steady growth of the total market.
Tech News: Radeon: Currently A Headwind, But A Tailwind Long-Term
Recently, volatility in AMD shares has been based around weaker than expected business from their Radeon GPU business segment. While I believe they managed their inventory levels far better than competitor Nvidia, 1H of 2019 was a weak point for Radeon because of crashing cryptocurrency demand. While this story has been told many a time, management teams at both Nvidia and AMD agree that these issues will be put past the market by Q3. So that begs the question: Where is the GPU market, and AMD within it heading from here?
Lets start with a little bit of history. As with the CPU market, AMD and ATI before it has long established itself as a budget player in the GPU market. Nvidia on the other hand, with their first mover advantage has always been the higher performing company. Now however, the story is quickly changing. As with Intel, AMD is positioning themselves technologically to gain market share on Nvidia.
Lets start with AMD’s offerings versus Nvidia’s offerings pre-Turing and Radeon VII. Most notably, I am going to look at the Vega 64 GPU from AMD against the GTX 1080 from Nvidia.
|Nvidia GeForce GTX 1080||AMD Radeon Vega 64|
|GTA V FPS||94||88|
|League of Legends FPS||221||197|
|The Witcher 3: Wild Hunt FPS||94||86|
|World of Tanks FPS||135||126|
As you can see, Nvidia’s 1080, a GPU that is more than a year older than AMD’s Vega 64, still delivered solid gaming performance. However, this is the reason for AMD’s lower pricing. The Vega 64 is a staggering 24% cheaper than the 1080. This has been a reason for AMD’s lower margins, and lack of high performance products. However, it seems AMD’s recent 7nm GPU offering, Radeon VII, will bring AMD to the table as a competitive offering in the huge gaming market. Coupled with AMD’s next generation offerings, like Navi in Q3, we could see AMD take share from Nvidia in the next few years.
However, we need to look at AMD’s offerings against Nvidia’s most recent offerings. In particular, I am comparing AMD’s Radeon VII with Nvidia’s recent RTX 2080 offering.
|Nvidia GeForce RTX 2080||AMD Radeon VII|
|Shadow of the Tomb Raider FPS (4K)||45||45|
|Rise of the Tomb Raider FPS (4K)||63||55|
|Ghost Recon Wildlands FPS (4K)||55||52|
|Hitman FPS (4K)||80||79|
|Far Cry 5 FPS (4K)||59||60|
|Metro: Last Light FPS (4K)||67||61|
As you can see, Nvidia’s 2080 offering, released in mid 2018 is higher performance than AMD’s newest Radeon VII offering, even though they are at the same price point of $700. While it is clear that the RTX is far ahead of Radeon VII, it is clear that AMD is making progress at the high end of the performance spectrum.
Despite AMD’s clear performance improvements, AMD has a lot of catching up to do in the high end segment. In general, it is more likely that AMD occupies the low to mid end of the performance spectrum. So, it is unlikely that AMD hits the high end of the GPU market. So AMD should probably stick to the low-end, lower margin end of the market. At the end of the day, Nvidia has the first mover advantage, and while AMD is making significant progress, will the tides really turn in the favor of AMD? I doubt it.
Now, we need to quantify the impact Radeon’s gaming segment has on AMD’s overall business. First lets, look at my estimate for what AMD’s GPU business did in Q1.
In Q1, competitor Nvidia saw gaming related GPU revenues totaling $954 million. A market share report pegged Nvidia’s market share at 81.2%, a huge step up from Nvidia’s Q4 number of 74.3%. AMD took the other 18.8% of the market. This means, if you do the math, AMD delivered GPU revenues of ~$220.8 million. If you combine AMD’s revenues with Nvidia’s, you get a Q1 market size of $1.175 billion. Annualizing these trough Q1 numbers, we get a 2019 market size of $4.7 billion. However, the 2H reset should yield solid growth in the overall market.
Let’s look at my estimate for AMD’s GPU business in the year ago period. In the year ago period, Nvidia reported gaming revenues of $1.739 billion and 66.3% market share. This sizes the Q1 gaming GPU market at $2.623 billion. This brings AMD’s Q1 revenues to ~$883.9 million. This means the GPU market has decreased by ~55%. AMD’s GPU business, because of its staggering market share decrease, has seen a staggering decrease of ~75%.
Now, lets use this data to estimate the 2019 GPU TAM. We know from previously that AMD estimates a 2020 TAM of $10 billion, but lets work out the math behind AMD’s 2019 TAM. I expect Nvidia to report 2019 gaming revenues of $4.872 billion. However, I expect AMD to report incremental market share gains throughout the year from Navi rollout and increased Radeon VII sales. These combined factors lead me to a market share estimate of 21%. This sizes the overall GPU market at $6.167 billion for 2019.
|Radeon’s Value||TAM||Market Share||Revenue|
|2019||$6.167 billion||21% (up from 18.8% currently)||$1.295 billion|
|2020||$8.5 billion (below $10 billion mgmt guide)||23.2%||$1.972 billion|
|2021||$10 billion||25%||$2.5 billion|
|2022||$11 billion||26.3%||$2.893 billion|
|2023||$12 billion||26%||$3.12 billion|
Tech News: Gross Margins & Operating Leverage
Another key piece to the AMD story is the company’s gross margins and operating leverage. For the first time in many years, AMD delivered a full year profit in 2018. While operating expenses like SG&A and R&D skyrocketed in 2018, the keys were AMD solid revenue growth and gross margin expansion. Gross margins increased from 34.1% in 2017 to 37.8% in 2018. However, compared to competitors Nvidia and Intel, AMD is still far away from adapting a similar margin profile.
|Margin Profile For FY’18||AMD||Nvidia||Intel|
|Gross Margin||37.8% (up 370 bps y/y)||61.6% (up 150 bps y/y)||61.5% (down 60 bps y/y)|
|Operating Margin||5.1% (up 510 bps y/y)||33.2% (up 30 bps y/y)||32.9% (up 50 bps y/y)|
|OpEx as a % of Revenue||30.7%||28.7%||28.7%|
As you can see, Nvidia and Intel sport far higher gross and operating margins than AMD. This is because of Intel and Nvidia’s focus on high performance, high price point products, leaving AMD to take on low-end low performance products. However, if this story is changing, with AMD taking on the high-end of the market, then AMD should see significantly higher gross margins. To be conservative however, I have projected that AMD only gets to the 50% margin threshold, rather than breaking into the 60% profile.
In addition, as AMD’s revenue continues to skyrocket, operating costs as a percentage of revenue should normalize to this ~29% level. This still allows AMD to take on high levels of SG&A as well as invest in future product R&D.
The bedrock of this thesis is that AMD gets to (a.) revenue levels that allow for operating leverage to kick in and (b.) a focus on taking share in high-price, high-margin products that will enable this operating leverage. A and B are intertwined. In order to see, high levels of revenue, AMD needs to take significant share in the high-end. As this article has pointed out however, I believe this is more than possible. And AMD has done it before.
Tech News: Cash Flow and The Balance Sheet
One of the key struggles AMD has had in the last several years is getting a solid financial foundation in place. However, under the leadership of Dr. Lisa Su, AMD has seen reduced leverage, higher cash levels, and profitability for the first time in years.
One of the key tenets of the AMD bear thesis has been, despite all of AMD’s progress when it comes to financial footing, a lack of cash flow and solid cash position. Meanwhile, bears still believe AMD’s debt levels are bloated.
While it is true that AMD’s financial position is not as good as other chipmakers like Intel or Nvidia, I believe that eventually, AMD will have solid financial footing.
Back in 2014, when Lisa Su took over the company, the business was on the verge of financial collapse. Now, the company could probably weather a recession.
An improving balance sheet and cash flows will help back AMD’s bull thesis and valuation.
Tech News: Valuation: $28 Price Target, Hold Rating
Anybody that reads my reports knows that I enjoy the process of valuing stocks, and frequently put ratings and price targets on those stocks. AMD is no exception. As with the other stocks I cover, I use a discounted cash flow model to value AMD shares.
Let’s start by calculating AMD’s cost of equity:
This cost of equity factors in an average beta of 1.77, taking an average of the unlevered and levered betas of the 1, 2, and 3 year periods. In addition, it factors in a 10 year treasury yield of 2.398%. In addition, it assumes a 499 basis point equity risk premium, per Aswath Damodaran.
Now, lets look at my WACC calculation.
You have already seen my general breakdown for AMD’s revenue profile. Now, lets look at the margin story.
The bloated operating income adjustments line for 2019 is because of the company’s high GPU inventory levels in the first half.
Despite my general optimism on the AMD business model, I believe much of the business’s upside has already been priced into the stock. However, if AMD can pullback, the stock could become a buy.
Tech News: Conclusion
AMD offers one of the most exciting opportunities in the chip space. It appears to be taking significant market share in significant markets from larger chipmakers like Intel and Nvidia. The only material problem I find with AMD is its valuation. At this point, all the potential upside may be priced in. Waiting for a pullback to initiate a position may be the most prudent move.
Disclosure: I/we have no positions in any stocks mentioned, and no plans to initiate any positions within the next 72 hours. I wrote this article myself, and it expresses my own opinions. I am not receiving compensation for it (other than from Seeking Alpha). I have no business relationship with any company whose stock is mentioned in this article.
Additional disclosure: This is not financial advice. I am not a financial advisor. Please do your own due diligence before investing in any of the aforementioned securities.