Investor Call Notes: 7th July
Market Dynamics and Future Roadmaps
As you may know, I do advice calls with investment firms on Wall Street, as well as investment groups with a multitude of clients. They reach out to me to get the broad strokes view of what I’m seeing, particularly in the AMD vs Intel space, regarding competitiveness, performance, manufacturing, markets, and applying that to short and mid-term roadmaps.
For these 1hr group calls, the host sends me some broad topics and questions in advance, I do some mild prep, and its styled as an enhanced Q&A session, with some questions coming in through email during the call itself.
In these ‘Investor Call Notes’ posts, I’ll be summarizing my notes for the call. These will be brief and abridged, so won’t be full in detail and content, and most of it behind the paywall. These aren’t public calls, so if you want to get on them you either have to be a client of one of the hosts, or organize a call with me 1-on-1. For anyone interested with a 1-on-1 call, please reach out and we can discuss, or I’m accessible through Guidepoint.
This call was on 7th July with Guidepoint, and was done before Intel’s recent quarterly financials, but still very relevant as we move into Q3/Q4.
Disclaimer: None of the content underneath offers direct investment advice. It is intended to help direct your understanding of these companies. The content contains predictions and statements that are at the best of my knowledge at the time given.
Q: Let’s start with your thoughts on macro impacts to the group, and specifically the markets these competitors touch within the CPU market – there’s a looming fear of recession and PC/Smartphone units have been guided down already, what is your sense of where we are?
A: There are lots of talks about a recession, not just in Semis. In light of moving out of working from home 5 days a week, and even as a response to the Crypto markets, what we’re seeing is a significant bifurcation in the various semi markets. PC is guiding down, after a record year of almost a million units per day during 2021, either flat or down single digits, and this seems to be offering opportunity for AMD but reduced margins for Intel, but the beyond PC markets such as embedded and automotive are continuing to see strong multi-year growth on the back of limited production capabilities. We’re seeing foundries like TSMC invest heavily in the legacy process node technologies that drive the embedded and automotive markets as a result, going through with a multi-year commitment and pushing partners who have been ordering the same chips for 10+ years onto those more profitable process nodes. On the flip side we’re seeing companies such as Micron recently downscaling their DDR5 production ramp in light of delays in semiconductor leading edge roadmaps. Certain companies are poised to take advantage here, while others are not as agile as they wish they were.
Q: And what’s ahead for INTC and AMD – can either excel in a recessionary environment? (Let’s talk about the impact to each segment specifically):
Today in the consumer market, Intel and AMD have their strengths. Intel has the best processors at the low-and-mid range, and they are priced very competitively in line with Gelsinger’s reduced gross margin targets. Performance per dollar, in both desktop and mobile, Intel has a lead here, and can also seem to match the volume. The downside however is that reduced GM – the process node they’re using is still seemingly expensive and not as ideal as 14nm was in previous years, and at the high-end they’re struggling on power consumption, again due to manufacturing. In a recessionary environment, I see Intel actually increasing market share, but ASPs declining as the market becomes more frugal.
AMD by contrast has the more competitive product in the mid-high end of the market. Their product is also expensive to make, however they have a performance per watt advantage at the top end, and are actually driving ASPs higher. We’ve already seen that in a depressive notebook market, analysts are showcasing that AMD’s market share is decreasing, but ASPs are increasing, and because AMD can’t produce in sufficient volumes to meet demand, the overall uptick is a revenue lift for AMD: just as many units, at higher prices. This applies to desktop as well, which has another couple of factors: high-end CPUs are often paired with high-end GPUs, which if rumor is anything to go by, will be surpassing 400W with the next generation coming later this year. Users at that end of the spectrum are cognizant of simply the thermal environment of what they’re buying, and so if an AMD CPU can deliver 98% of the performance of Intel at half the power for around the same price, paired with a strong NVIDIA GPU, it’s an easier choice to make. How this changes moving forward with AMD will depend on their pricing with TSMC: later this year they will come out with the next generation product, built on TSMC 5nm, which is a more expensive process node for AMD. To keep GMs increasing, prices will have to increase, and we’re still waiting to hear on performance metrics to see if performance per $ is on the same lines.
Intel retains the commercial crown, simply due to longevity, volume, and agreements. Last year AMD stated that their two main goals were Enterprise and Commercial, and part of the issue here is commercial markets are slow to build trust, even 3 or 4 generations into AMD’s new high performance product line. Intel has long term agreements in place as well, making penetration difficult. Ultimately AMD will still progress after this market, but I see in a limited capacity where big contracts can be had. Commercial silicon is usually the same as consumer silicon, just with more verification – AMD has the finances to do that verification, but as overall units drop and belts tighten, it’s going to be tough to compete with Intel purely on volume.
The enterprise discussion is one that’s going to linger until the middle of next year. AMD holds the perf/$ and perf/W crown in enterprise, but due to the market and volumes, is seeing slow market share increases. AMD is selling everything they make, but the volumes aren’t there. I once asked Forrest Norrod, about 3 years ago now, if AMD had unlimited volume and a leading offering, where could market share go, and he said around 30%. Both AMD and Intel are set to launch next generation parts this year (or next), with AMD showcasing significant performance uplifts while Intel is showcasing delays and a limited understanding of what exactly they’re offering. In a recessive environment however, I still expect AMD to sell all they make, simply because the demand is there with current contracts. However in that environment, the willingness for customers to make a switch becomes depressed – why refactor your chain when it’s easier just to rebuy the same you’ve always had. This becomes less of an issue as cloud is growing, and it has been a long-term problem as AMD becomes the leader in this market in performance, if not market share. AMD will also drive ASPs here, especially with the next generation promising good gains, while Intel seem to have an expensive product to manufacture on their hands.
Let’s also discuss covid/WFH – as relates to CPU’s and workloads – how do you see the roadmap shifting as workflows have shifted back and forth?
Aside from learning to deal with remote working, and workloads such as video encode/decode (which is still pretty terrible), we’re not seeing much change in CPUs and workloads for consumers. AMD and Intel are both committed to including ‘AI’ in their hardware, but realistically that’s more of a software problem these days, and while AI seems easy to put into a smartphone, exactly how to use it (and use new CPU/GPU features for it) in a WFH environment is limited. We’ve seen features such as presence detection be demoed, features such as AI changing your camera so that it looks like you’re always looking into the screen on a video call, but it feels like the companies producing those solutions want fixed contracts with specific laptop manufacturers or cameras, and as we all know, if there’s one thing the PC/laptop community have failed in a WFH environment is installing decent cameras on their devices.
WFH has affected the enterprise markets however, especially as businesses are relying on a form of cloud infrastructure to support them. This has led to both AMD and Intel to announce upcoming ‘cloud focused’ processors: AMD has Bergamo in 2023, Intel has Sierra Forest in 2024. These are high core count cloud designs, trying to match Ampere’s offering and Amazon’s Graviton. How those agreements look is a bit tough to say right now – a more expensive processor for a focused workload, or will CSPs get a good deal with better performance and an overall lower spend? We won’t know until later next year.
Questions below the fold:
Q - PC: Intel's Alder Lake has completely redefined x86 desktop PC chips with a new hybrid architecture that delivers amazing levels of performance. AMD released its Ryzen 7 5800X3D, a new CPU with 3D V-Cache – how does the performance of the two stack up?
Q - Server: Ice Lake increased the core count from the top end from 28 to 40, but AMD is still offering 64 cores at the high end, and an advantage when it comes to performance/core and performance/watt?
Q: Can AMD compete on a fairly level playing field in the server market vs. Intel? How do CIO’s compare the two at this point and how long might they be willing to wait for Sapphire Rapids?
Q: Is there any reason AMD can’t take more than 15% market share in your view?
Q: Now that XLNX has closed – how do you view their FPGA’s within AMD’s portfolio?
Q: How has the continued delay of Intel's Sapphire Rapids launch impacted the competitive dynamics?
Q: Sapphire Rapids: When might this product launch?
Q: Sapphire Rapids: What are your expectations about when it can ship in volume?
Q: If Sapphire Rapids can’t hit its latest targets what are the implications across the space? Will AMD gain more share if it continues to be delayed?? What are the chokepoints to its release?
Q: AMD next gen Genoa: Perspectives on AMD's recently announced Genoa - product road map through 2024, which contains plans for 4nm and 5nm CPUs based on Zen 4 architecture, which will include Ryzen chips.
Q: RISC-V -- will emerging semiconductor alternatives become more prominent? What are the limitations of Risc-V based technology?