Advanced Micro Devices (AMD) is undeniably a rocket in flight with a 1100% run up in the last 3 years. Such a large move has a lot people worried – just because the price has gone up a lot. This article is for new investors and existing investors thinking about taking profits – to explain that AMD might be nowhere near the peak of it’s flight. There’s a chance AMD could be a many bagger from today’s prices over the next 2.5 years. With the average analyst one year price target currently at $18.41 – how could such a thing be possible?
AMD Is Still Undervalued
AMD has a 40 year history and for the most part deserved a lot of negative sentiment in the last decade prior to 2017. AMD was a distant second place to Intel in the PC market. More recently Nvidia had an advantage in gaming GPUs and AI processing. AMD lost money and was in precarious financial condition in 2016, so much so bankruptcy was a real concern and stock price went as low as 1.81. AMD stock stayed sub-16 for more than a decade.
Still today AMD is under-owned by institutional investors relative to it’s main competitors(data from Yahoo finance):
- AMD owned by 692 institutions
- INTC owned by 2861 institutions
- NVDA owned by 1759 institutions
Look at Market Cap/Revenue for Q2 2018:
Nvidia revenue is being valued more than 4X higher than AMD revenue.
Look at AMD market cap vs. Intel + Nvidia market Cap:
Intel + Nvidia is being valued 17.5 times higher than AMD.
Clearly AMD is currently valued at a small fraction of it’s competition’s market value. But does this tiny valuation make sense based on future prospects? Let me give you 21 reasons why AMD deserves more respect:
1) Great Leadership – Dr. Lisa Su and team
One of the major reasons AMD is a different company today than it was several years ago is the CEO – Dr. Lisa Su is a brilliant MIT trained engineer and has brought an incredible level of smarts, focus, leadership, and accountability to the company. It starts at the top, but clearly AMD has an incredible team up and down the ranks. In the last couple of years AMD just keeps doing what they say they are going to do – and if there are surprises – they are to the upside. I have gained a high level of trust in what the company says. To manage and execute as a company with a tiny fraction of the resources of it’s two major competitors, and attain either close to parity or product advantages across so many products – is a truly remarkable accomplishment.
In contrast, INTC is CEO-less in the middle of a CEO transition that could go on for months.
2) Architectural Design Brilliance
Sometimes being the underdog forces you to think out of the box. In the 2011 timeframe, AMDs architects started designing their hardware using modular interfaces that enabled the CPU and GPU components to relatively easily be combined and reused in different products. And the most brilliant part of the design was something they call Infinity Fabric – which is a very flexible high performance communication system, that allows all these different hardware components to communicate either on the same Chip, but even more importantly across chip boundaries.
3) Chiplets/Modularity/Infinity Fabric for Moore Performance
As Moore’s Law slows down and drags down the ability of semiconductor makers to develop more complex chips at low cost, computing demand just keeps increasing, To satisfy demand, AMDs competitors have been forced to continually increase the size of silicon chips they produce to increase the performance of their high end products. For example – Nvidia just released it’s latest RTX 2080 GPU with a massive 780 mm2 die size. Intel server chips are of similar large size.
It is a law of semiconductor manufacturing that the larger the silicon chip is, the lower the yield, the higher the cost to manufacture, and the harder it is to find high binning chips. The genius of Infinity Fabric is that it enables AMD to fabricate high yield, small chiplets, and then combine one or more chiplets in a single composite chip to create to create a range of products that can scale from low end products to extremely high end products.
The smaller die size chiplets have higher yields relative to larger monolithic chips so manufacturing costs are far lower. And being able to select from a high number of smaller chiplets, it exponentially easier to produce higher speed products because it’s far easier to find some chiplets with a small number of fast cores, than it is for a large monolithic chip to have all fast cores. AMD can cherry pick the best of the chiplets and produce a high performance chip more efficiently than Intel and Nvidia can with their large monolithic chips.
It is also true that new manufacturing process nodes start out by yielding much lower rates than mature yields. New 7nm (or 10nm) manufacturing yield will be far worse than mature 14nm yields. It is not a coincidence that AMD is moving to 7nm before either Intel or Nvidia is making this move. Chiplets and infinity fabric make this possible because the much smaller chiplet dies yield exponentially better on low yielding immature manufacturing nodes. Getting a large monolithic design onto 7nm or 10nm may take years to get the process mature enough to be financially viable. This could be a major factor in Intel’s 10nm multi year delays.
4) Modularity+Infinity Fabric for Less Engineering Cost
AMD’s modular architecture and Infinity Fabric also massively lowers the demand on the engineering teams to create different products – because the same chiplets can be mixed and matched to create many products of different capabilities. This factor allows the much smaller AMD to trade punches with two much larger competitors. Threadripper is a powerhouse high end desktop part that was created by a very small team inside AMD – because it is reusing the same chiplets designed to power Ryzen and Epyc.
5) Modularity/Infinity Fabric Enables Unique Products
Not only does Infinity Fabric work to connect multiple CPU chiplets together, but it was also designed into the Vega GPU architecture. The brilliance of this design allows AMD to combine one or more CPU chiplets with one or more GPU chiplets and create their game changing high performance APU products the dramatically outperform Intels comparable integrated graphics products.
AMD is leveraging their brilliant modular architecture with all of their future semi-custom products. AMDs unique ability to combine super high performance X86 cpu power with cutting edge GPU technology and custom AI hardware is a capability that no other competitor currently has. Such devices will be perfect for providing the ‘brains’ for assisted driving and self driving cars that are going to be an enormous market with millions of units produced annually in the near future.
- Zhongshan Subor gaming console in Chinese market
- Microsoft Xbox
- Sony Playstation
- AMDs new generation of laptops and desktop APUs
- Tesla has discarded Nvidia GPUs for its new autonomous driving hardware. It is rumored that the new hardware is using AMD hardware with custom Tesla designed elements
These APU and semi-custom devices will create millions of units of demand for AMD chips in the coming years.
AMD is shipping products this year across a larger TAM than any of it’s competitors:
7) Semiconductor design is like driving a semi-trailer
That means slow. AMD started designing the Zen cpu architecture and Infinity Fabric in 2011. It took all the way until 2017 for the products to be released. The massive competitive advantage AMD has today took 5 years to develop. There is no way for a competitor to match the technology that AMD has already mastered quickly – it will take a minimum of 3 to 5 years – and that’s assuming they started working on it a couple of years ago. Yes, Intel is working matching AMD’s current capabilities, but they road maps they have published show their first GPU products and their 10 nm server products will appear in 2020. That’s 2 full years from now – and AMD is not standing still.
8) Process Advantage
AMD is a fab-less company and they will be fabricating chips using TSMC and Global Foundries – and they could potentially use other foundries in the future if necessary. This is a huge advantage because building new generations of semiconductor technology has become outlandishly difficult and expensive. AMD can save huge capital expenditures and take advantage the largest most advanced fabs in the world – that can amortize capital expenses across many companies products. AMD will be shipping high end 7nm GPU products this year (2018). Intel has delayed it’s roughly equivalent 10nm CPU technology for 3 years and Intel road maps show 10nm server chips will not ship until mid-2020. AMD has historically always been behind Intel in process technology – being in the lead is unprecedented.
9) 7nm Rome Epyc “Monster”
First generation Epyc has put Intel Xeon on it’s heels with a multitude of competitive advantages in more PCI lanes, 32/64 cores/threads per chip, and single socket servers than can outperform dual socket Intel servers in many ways.
And just wait until legions of 7nm “Rome” Zen 2 Epyc chips show up in 2019. These chips are rumored to support 64/128 cores/threads. All other aspects of the chip will be improved as well. This chip could have a large advantage against Xeon on older monolithic architecture with 14 nm process.
Intel’s ex-CEO has publicly stated it “hopes” to hold AMD to 15-20% market share in the data center. From the likely performance advantage AMD will have with 7nm Rome in 2019, and AMD’s better pricing, it’s possible AMD will take a lot more than 20% datacenter share.
10) 7nm Vega X2 “Monster”
- 7nm Vega has Infinity Fabric and can be run with MULTIPLE Vega chiplets in one GPU.
- The pain that Intel Xeon is feeling from Infinity Fabric inside Epyc, could soon to be felt by NVDA Volta from Infinity Fabric inside 7nm Vega 20 X2.
- Not only are 12 nm Volta at a process disadvantage vs 7nm Vega – buts it’s going to have to go against 2 or more GPUS. I won’t be surprised if 7nm Vega X2 or X4 tops Volta in AI processing and Datacenter compute loads.
11) AMD Deeper
NVDA has had a free run in the deep learning/AI market for the last 4 years. This has been a major factor in the NVDA massive stock run up and 162 Billion market cap. AMD has been a minor-factor in this market.
This all changes this year (2018):
- AMD has developed open source ROCm platform that enables easy use of upcoming AMD GPU hardware on all the major AI software frameworks such as TensorFlow, Caffe2, etc. NVDA proprietary CUDA programs can be compiled to run on AMD hardware.
- 7nm Vega has powerful AI specific hardware that was not present in previous AMD GPUs
AMD is working on multiple future generations of future products with multiple leapfrogging design teams. They are not being complacent with their new found competitiveness. We will see improved new generation products coming out for the foreseeable future. By the time Intel ships 10nm server chips in 2020 AMD will already be bringing out Zen 3.
There seems to be no end to the series of security flaws that impact Intel chips. Some flaws have been found in AMD chips, but they have been fewer than Intel, less severe, and more easily remediated. These Intel flaws have been a nightmare for the major cloud providers like AWS and Google – because their systems require servers to be able to run virtual machines from multple untrusted customers on the same hardware. Intel security remediations have a significant performance impact on many workloads. Enterprise customers really care about security.
AMD Epyc chips also have powerful chip level encryption features that encrypt virtual machines on servers at almost no performance penalty.
14) Datacenter Latency = Acceleration Coming
The datacenter markets are notoriously slow to sell into, because even though you have a superior product, it takes a year or more for customers to qualify, test, deploy systems. From recent conference calls it is clear AMD is getting traction. AMD’s product superiority is likely getting greater with 7nm Rome – so the gain of high margin market share could continue and even accelerate next year. However, even though these gains are likely coming – they haven’t hit the bottom line yet. Profits could not only grow, they could accelerate in the next couple of years.
15) Financial Metrics
All these abstract advantages AMD has in the first wave of new products like Ryzen and Vega, has already resulted in tangible financial results improvements:
- AMD is now profitable at an accelerating pace
- AMD revenue growth was 50% year over year in Q2
- AMD data center increased from virtually zero to 1% in Q1 and doubled from Q1 to Q2
- AMD data center revenue is “on track” to accelerate to “mid-single digits” by end of year
And the full weight of new products in new markets has not even hit yet.
16) TAM Expansion + Higher Margins
Prior to 2017 – AMD primarily competed in the low end CPU/laptop market, low end to mid range GPU market, and semi custom chips. All of these low end markets are low margin and AMDs primary point of differentiation was low price.
NEW Markets AMD has/will enter 2017-2019
- mid and high end desktop CPU market
- mid and high end laptop market
- Entire workstation market
- Entire professional desktop market
- Entire professional laptop market
- high end GPU market
- Entire datacenter server CPU market low middle and high end
- AI market
- Embedded Market
AMDs TAM is growing 5X 2016 to 2020 – and most of that gain is backloaded into 2019 and 2020. And much of that higher TAM is in HIGHER MARGIN product lines.
17) Momentum/Technical Factors
- AMD technical outlook is massively positive:
- AMD stock has tremendous upward momentum
- AMD is the top performing stock in the SP500 for 2018
- longer term it is already a 11 bagger off of it’s 2016 lows.
- Shorter term AMD is up 150+% over it’s lows in the last 4 months
- AMD has broken key resistance points and is at or near 12 year highs
18) Institutional Demand
Only 692 institutions own AMD vs 2789 who own INTC. All the major funds are going to have to get an AMD position soon – because what investor will want to own a fund that isn’t smart enough to own the best performing SP500 stock? There could be a massive amount of institutional buying to come.
19) Massive Short interest
AMD ain’t going out of business. For all the reasons we’ve already covered – the numbers are only going to improve from here. The stock has massive upward momentum. 167 million shares are still short as of a few weeks ago. Are the shorts feeling a bit tight?
20) Timid Analysts
As of right now the average analyst price target is $18.41 (from yahoo finance). The stock is at $23.95 as of today. The analysts could be way behind on this stock – and if AMD continues it’s great execution, this means there’s a whole lot of analysts that are going to have to issue a whole lot of upgrades again and again that will chase the stock price higher. If you looked at the analyst projections from a couple months back, they were at $14.60 – with AMD it seems like their projections are trailing indicators.
21) Sentiment Will Swing to Positive
If AMD keeps piling up quarter after quarter of improving results and keeps successfully launching great products, eventually many of the doubters and trailing-metric followers will come around and see past the history. Excitement could grow and momentum could build. This move up could happen far faster than people think.
- I present 4 models below with differing levels of optimism. Notes on the models:These models are for the end of 2020
- 2020 TAM Estimates are from https://ir.amd.com “AMD Investor Presentation” Slide 5
- I estimated gross margins for each market component, however I did make sure that the calculated aggregate gross margin resulting from my estimates fall within the AMD projected future range of 40-44%, which is from https://ir.amd.com “AMD Investor Presentation” Slide 36
- Model “What if” inputs are in blue text – calculated fields are black text
- AMD’s Q2 2018 operating expenses are $467 million ($1.8 billion annualized) at 27% but dropping rapidly from prior years 34% (see AMD presentation Slide 32). I am assuming AMD’s operating expense percentage will continue this drop going forward as revenues increase over the next 2.5 years – however due to much higher revenues going forward, the lower percentages result in much higher dollar amounts.
- With AMD’s growth rate going forward likely strong, I believe PE ratios in the 35-45 range are justified. Also, Intel has ~100K+ employees. AMD has ~12K employees – AMD being a far leaner company, any market share that AMD takes from Intel will result in higher profits for AMD relatively.
- In gaming GPUs AMD does not compete in the high end against Nvidia’s products today. All of the gaming graphics estimates are conservative. Though, maybe I am way off here with all the gaming consoles coming out with AMD chips in the next few years…
- NOTE: there’s no guarantee that any of these Models will turn out to be accurate. Do not make any investment that you cannot afford the downside if things don’t turn out as you hope.
Model 1 – Conservatively Optimistic
This model uses ex-Intel CEO’s lower ‘hope’ estimate for AMD datacenter market share. With AMD in 2018 – a qualification year – going from 0 to 5%, only reaching 15% in 2 additional post-qualification years would have to be a fairly mediocre effort by AMD.
Model 2 – Moderately Optimistic
Model 3 – Optimistic
Model 4 – Insanely Optimistic
If anyone is interested in the spreadsheet let me know I’ll send a copy.
- AMD going forward is an entirely different company from the AMD of a couple of years ago. AMD in the coming years will be addressing a far larger, higher end market with higher margins.
- AMD is attacking the highest end of the market with the highest growth and biggest margins – the datacenter. They have the products coming that the competition probably cannot match for years to come.
- Looking at current or next quarter numbers completely misses the long term potential of AMD as quality products in many new market segments ramp up.
- AMD’s current stock price ($23.98) is quite undervalued relative to future potential.
- In the next 2.5 years, an AMD stock price of $100+ is possible.
- There is no such thing as a sure thing in investing. Future projections, especially those made out 2.5 years, could go wrong in a wide variety of ways. Competition could be stronger than currently projected, or there could be product or manufacturing problems. The stock market could crash at any time. Invest responsibly and do not invest more than you can afford to lose. See the companies 10-Q publications for all the risk factors each of these companies face.
Disclaimer: The author has a Long position in AMD. The author is not receiving compensation for this article. This article is intended for informational and entertainment use only, and should not be construed as professional investment advice.