Open Source Semiconductors
Open Source Semiconductors and the Accelerating Pace of Change in the Information Age
7/27/19
(The following is an update to the original blog posted on February 22nd, 2019)
Click here to SIGN UP for SITALWeek’s Sunday EMAIL (please note some ad blocking software may disrupt the signup form; if you have any issues or questions please email sitalweek@nzscapital.com)
What happens when the low-cost, high-value supplier in an industry is disrupted by an even cheaper alternative like open source? This is a growing trend in the Information Age where early disruptors of legacy Industrial Age capitalism are now finding themselves disrupted by even cheaper alternatives at an accelerating pace. One of the takeaways from my writing on lessons learned investing in ARM (here for those that didn’t read it) was this: predicting the future is hard and getting harder in an age of disruption, so you want to make as few predictions as you can and instead focus on being adaptable. Whether you are running a company or building a portfolio, often there are unexpected developments, like a new competitor with a new business model. In this age of transparency, we find that the business tactics of the 1900s don’t necessarily work anymore. In 20 years of analyzing public companies, the majority of which ultimately fail (and these failures are happening at an accelerating rate), there are only 2 key indicators of potential longevity we’ve come across, and these two cultural traits are highly related - humility and adaptability - the humility to know what you don’t know, and a culture that facilitates constant adaptation within the complex “ecosystem” a company operates in. There are also many smaller pillars of humility and adaptability, such as decentralized organizational structure or pricing products well below what the market will bear. This latter trait we’ve called a “low tariff extractor” by which we mean the company is generating much more value for its customers and partners than it is for itself, which ultimately compounds and creates positive feedback loops, platforms, and network effects that become less exposed to disruption (this is the foundation of our Complexity Investing framework, and we expanded on this topic in the NZS white paper). While most 20th century MBAs and investors (or even worse - investors with MBAs!) believe pricing power - the ability of a company to steadily raise prices - is a key indicator of long term value creation, in most cases, we think the opposite is true. Instead of steadily raising prices, companies that regularly provide more value for the same (or less) cost are more likely to survive long term.
How ARM disrupted Intel to take the mobile market
And this brings us to ARM Semiconductor. Why should we care about ARM? ARM is the semiconductor technology that powers every smart phone in use today and hundreds of millions of other consumer, enterprise, industrial, and healthcare devices. Without ARM there is no iPhone, and without the iPhone or other smart phones most of the innovation you see around you today wouldn’t exist. ARM traditionally made money 2 ways: 1) ARM charges a meaningful upfront license to companies that want to incorporate their intellectual property (which consists of low-power processors, graphics, and other building blocks for things like smart phones, networking gear, and much more); and 2) ARM then charges a very low royalty rate for each chip that is made containing that licensed intellectual property. In most cases, this per-chip license fee of tens of cents is well below the value it creates for the customers. For example, a company like Apple (the A-series processors and many other chips in iPhones, iPads, and their watches are all ARM-based) might pay tens of millions up front to license the latest semi building blocks from ARM, and then pay a nominal fee that might add up to $1-2 per device. Effectively, the vast majority of Apple products today can only run on ARM, and lot of Apple’s value is driven off this nominal fee. We loved that - ARM is a go-to example of a company that creates far more value for its ecosystem than it does for itself - ARM’s customers have combined market caps in the trillions while ARM itself was sold to Softbank for only $30 billion dollars. As a matter of public record we were decades long holders of ARM in the tech fund we used to manager. But, now, ARM is facing a threat from an even lower cost alternative: open source. Much like we saw in the IT world with Linux eating away at Sun Microsystems, Microsoft, IBM, and others in the server market (and Linux is also the core of Android, the dominant smartphone operating system globally despite many efforts by others to win in mobile), ARM now faces direct competition from open source competitor RISC-V.
Open Source and RISC-V
Open source solutions have traditionally been software based - communities of software engineers develop and maintain large repositories of programming code for anyone to use and modify as long as they do so in agreement with the goals and rules of the project. Recently, this ideology moved into hardware with projects like Open Compute (started by Facebook), which is focused on design of servers, networking gear and other data center devices. Open source semiconductors are a relatively new phenomenon. In December 2018, the chief technology office of hard disk drive and flash memory giant Western Digital made the following comments:
...we announced our first RISC-V processor core. And we announced that it would be completely open source. So in effect, think about this as doing to the processor world what Linux did to operating systems...Now let me give you a little interesting data point. When we started this with RISC-V and developing our own cores rather than acquiring them from the outside, we said, "Okay, we're on a marathon. We're not on a sprint. And so we need to have modest goals to get going." Right? So we said, "Let's just try to achieve parity with the cores we're using today." Right? So we shipped 1 billion cores a year. And so our first modest goal say, "We're learning something new by putting this together. Let's just basically achieve parity to what we have." That was essentially the goal we set for ourselves. Here's what happened: 30% improvement in power consumption, 40% improvement in performance and 25% reduction in footprint. Version 1.0. Not bad, right?
Wow, that’s quite a statement from a company that ships products with over 1 billion processors every year! A search of conference call transcripts over the last year shows a rise in references to RISC-V, including Microchip shipping the first RISC-V-based FPGA and Marvell Semiconductor potentially losing business at Western Digital. Google is also releasing a version of their machine learning software Tensorflow on a RISC-V chip and is an early supporter of the architecture. Alibaba’s semiconductor subsidiary PingTouGe (Mandarin for honey badger!) announced the world’s most advanced RISC-V processor design (not an actual production chip) in July 2019 and is making the FPGA chip available as open source for anyone to iterate on and improve. In fact, all of China is a major driving force behind RISC-V development as this article goes into more detail on. With rising trade tensions globally putting semiconductors at the heart of the battle between the US and China, open source represents a major way to circumvent export bans. (It’s worth noting that despite China’s efforts to lead the world in RISC-V designs, the country itself lacks the infrastructure to produce advanced chips. Instead they would need TSM in Taiwan or Samsung in South Korea. And, they would in turn rely heavily on many technologies exclusive to US and European countries.)
Softbank, who now owns ARM, was asked on their September 2018 earnings call about RISC-V, and a company representative responded with the same thing we heard from software companies fending off Linux 20 years ago - it’s an eerie echo from the past:
RISC-V. It's an open source, and they're working hard and we respect that. However, when they try to develop responsible products, IP issue as -- and security has to be insured. Hacking or virus or various issues are emerging these days, so who will ensure that? This is another point that needs to be considered, so I think that will be a bigger problem or a bigger issue. RISC-V is countering the IoT chip. So Arm's royalty is 0.01 or 0.02, to begin with, so 0.01 or 0.02. To save 0.01 or 0.02, the free, open-source companies, will they go to the OS or go for something that is not responsible or where security is not insured, just to save 0.01 or 0.02?
Well, it seems like many of ARM’s customers, at least in the case of Western Digital and Microchip, are exploring RISC-V. And, there appears to be broad and growing support for the project despite ARM being a low tariff extractor and platform enabler. So, what’s going on here - why did this happen? One of the problems is the rather large upfront license fee, which probably ranges from $1M to tens of millions depending on the product and the customer. In one of the last quarter’s ARM reported as a standalone public company license revenues were $150M, or about 40% of total revenues. On an annualized basis that’s $600M of expenses that don’t exist in the RISC-V ecosystem. However, even though that may seem like a lot of money, it’s a drop in the bucket compared to $10B’s in revenue these chips ultimately garner in the marketplace; so, the upfront cost barrier seems like a strawman argument. And, certainly, that 1-2% royalty Softbank mentions above is enough to keep the lights on and earn ARM a decent profit, but, again, ARM’s value at $30B (based on that royalty cash flow) barely measures compared to the value it creates for their customers. In response to RISC-V, ARM recently introduced a program that postponed the upfront license fee until a chip goes into production. This would allow a startup chip company access the ARM library to design a chip, but then they would still have to pay both licenses and royalties if that chip goes into production. It’s an incremental barrier removal, but likely not the right response to RISC-V for a variety of reasons that underlie the real motivations behind open source’s rising popularity.
Let’s look in more detail at the reasons open source now makes sense for semiconductors
The semiconductor industry is following the inevitable trend toward free and flexible, community supported solutions for the long tail of new applications. When there were a handful of large, concentrated markets with more homogeneous architectures like cell phones, PCs, servers, hard drives, etc., a centralized IP provider like ARM seemed to make sense. And, certainly, ARM was cheaper and more power efficient than the CISC-based x86 alternatives from Intel. Without ARM, it’s safe to say we wouldn’t yet have the multi-billion unit smartphone market - the alternatives prior to ARM simply didn’t have the power efficiency needed. But, now we are entering a renaissance of semiconductors like nothing the industry has ever seen. We have gone from millions of early computers to hundreds of millions of PCs to billions of phones to tens of trillions of connected IoT devices - anything that can be connected and anything that benefits from being smarter with AI, machine learning, and software will be connected and will get smarter. The massive, virtuous circle of centralized cloud computing and decentralized, intelligent, connected devices working together to make technology more useful, combined with new interfaces (such as voice input) represent a tsunami of new opportunities - all powered by new semiconductors. These new semis need a hugely diverse set of capabilities and appear much more like the Linux analogy as opposed to monolithic, dominant operating systems that preceded it like Sun, Unix, and Windows Server. Western Digital’s CTO made this very same point last December:
The other reason we were able to do this goes back to the very foundation of all of this, special purpose, general purpose. When we are acquiring cores, great cores, nothing wrong with them, from third parties, they are general purpose. They are trying to solve a problem for many of their customers using one set of IP. When we set out to design our core, we were solving our problem, or more precisely, our customers' problem, how do we optimize the data for our customers. And so because we weren't constrained by putting all of this extra superfluous stuff that we didn't need, this was the end result.
Of course the word “free” above is misleading. RISC-V users will spend hundreds of millions of dollars developing new chips, and there will be at least that much spent on software, security, and other higher level services for the chips. Linux is not free either. But, it is more adaptable and moldable to a broader set of diverse use cases. So, shifting costs from licensing software to maintaining and improving capabilities of that software is key.
The explosion of new semiconductor demand represents a heterogeneous mixture of a huge variety of use cases. Intel’s dominant x86 processors is a dinosaur, ill equipped to handle the new workloads in a power efficient manner and is rife with security issues. GPUs such as those made by NVIDIA have risen to take on machine learning and AI workloads as have FPGAs, or programmable chips such as those made by Xilinx. But these GPU and FPGA chips also don’t meet the Cambrian explosion of new chip use cases. Google has developed their TPU ASIC to run their custom AI tensor flow workloads and many other companies such as Facebook, Amazon, Tencent, and Alibaba are working on custom chips. Tesla recently introduced their own custom chip for autonomous driving. All of this is happening because the old Intel CPU just doesn’t cut it, and increasingly ARM may not be flexible enough either. Further, the ending of Moore’s law (which was several years ago now) has precipitated a move toward multi chip packaging and cutting edge EUV technology raising the bar significantly for advanced chip design and manufacturing. In other words, it’s getting complex. Really complex. And that complexity favors open source, flexible and adaptable designs. Here is just one cool example: a new processor called Morpheus, based on RISC-V, which can change its internal code every 50ms to make hacking at the chip level impossible - this is a need that x86 and ARM would likely never address.
It’s not about product risk, it’s about business model risk
So, what do you do if your company or industry faces a challenge like this from a “free” or better value proposition alternative? Companies have traditionally responded in different ways to open-source competitors. One way is to double down on vertical integration - Sun arguably did this with their servers and storage systems (spoiler: it didn’t work). Another not-so-successful route is litigation and fear - that’s what Microsoft did in the early 2000s referring to open-source software as a cancer and restricting the ability for it to work with Microsoft Software. Microsoft also supported lawsuits trying to bring Linux down. There is also an analogy of the current Oracle lawsuit against Google over Sun’s Java that is ongoing now. So, the first 2 options seem to directly backfire historically. What then might work if faced with such a challenge? Well, in a plot twist, we have to look no further than Linux-hating Microsoft for the best example of what actually works. Under Satya Nadella’s leadership over the last 5 years, Microsoft has become one of the biggest contributors to open-source software including Linux. They also acquired Github, a key repository and collaboration platform for software engineers, for $7.5B. Microsoft realized they needed to add value, not in the operating system, but in the services, tools, and applications that customers use. Instead of minting money on Windows Server software, they have emphasized compute-as-a-service on Azure, which also runs Linux workloads. They’ve emphasized Office 365, SQL databases, management and security tools for IT administrators, etc. In essence, they’ve shifted up the value stack from the operating system layer to provide their customers with a higher level of tools and services necessary for transition to the cloud/subscription software and de-emphasized selling licenses to Windows Server. This is what you need to do when lower cost alternatives come at you: provide even more value to your customers.
Embracing Change
How might a company like ARM respond to a threat like RISC-V? ARM is a phenomenal company filled with amazing engineers and managers that are deeply focused on the success of their customers. As an outsider, it’s unfair for me to speculate and judge, but, with the limited knowledge I have, here are some things I might consider if I were steering the company as an armchair CEO: I would embrace RISC-V for heterogeneous uses while still supporting development in the core roadmap of ARM’s general purpose products. I’d double my engineers and put as many on RISC-V as I have on internal products. I’d create a series of new intellectual property for IoT, security, data, AI, machine learning, etc. I’d completely eliminate the upfront license fee to work with core ARM products and, to offset, you could raise the per-chip fee, thereby lowering the initial hurdle but creating more value and, consequently, make as much (or more) over time. This strategy is all very risky, but, perhaps not as risky as doing nothing at all. Indeed, if you look at the hiring ARM is doing, I think they understand the threat and there is a good chance they are implementing an adaptive strategy. Bloomberg profiled ARM earlier this year, indicating much of what I discuss here might already be in progress. Such a transition would be difficult to implement as a public company, especially when ARM’s core market in smartphones have slowed dramatically. So, ARM is lucky to have the cover of Softbank if ARM can focus both on their core business and embrace value-added software, services, and IP blocks for RISC-V. But, that’s still risky, as it’s near impossible to shift a culture from a proprietary mentality to an open-source one like Satya did at Microsoft.
There are many large semiconductor companies that rely on the ARM processor for many of their products. They too need to be highly aware that this innovation in open source could create competitors, especially in China, which can then attack markets with lower cost chips. ARM’s current licensees thus need to also aggressively engage with open source semiconductors and move up the value stack to provide more security and adaptability in their products to meet the explosion of diverse customer needs.
So, here’s the main takeaway: be adaptable and focus on customer success. Answer this question and work backward: what are the fundamental needs of your customers, and how is a competitive threat meeting those better than you might be? Answering this might require a change in business model, not just a change in your products or services, and that’s the hardest shift for any corporate culture. Rules are changing under the transparency of the Information age, and hiding under high priced, proprietary products when an open source or free alternative is brewing is no longer viable. Companies need to constantly disrupt themselves to survive in this age of increased innovation even if that means a wholesale change in how business is done.
For investors there are a few important things to keep in mind. First, the next 2 decades will see an enormous explosion of demand for semiconductors as the industry moves from billions to trillions, and possibly tens of trillions of units. This follows a decade of consolidation and rationalization across many sectors of the semiconductor industry. And, it happens at a time of rising complexity and costs along with increased security, intelligence, connectivity with the perfect storm of cloud, AI, IoT, and 5G. While the industry remains somewhat cyclical with general economic exposure, and there are key geo-political risks with semi’s at the heart of the conflict between the US and China (not to mention the disputed status of Taiwan which sees nearly 70% of all semi’s made or pass through its borders), the long term growth drivers appear to be largely ignored by investors today. This creates a lot of opportunities as there are many areas of the chip supply chain to explore. Design software makers like Cadence and Synopsys are vital and seeing an expansion of potential customers. Further, equipment companies like ASM Lithography are indispensable. Taiwan Semi and Samsung are leading the world in complex, cutting edge chip manufacturing. Memory has become more important and more rational lead by Samsung and Micron. NVIDIA and Xilinx are leading the industry in solutions for high performance computing. A host of companies are creating an array of solutions and components for customers including Texas Instruments, Microchip, and Amphenol. There is a lot of area to hunt in this sector, and if you can look to the long term trends of heterogeneous chips and end markets, you might find some very interesting investments.
Disclaimers:
The content of this newsletter is my personal opinion as of the date published and are subject to change without notice and may not reflect the opinion of NZS Capital, LLC (“NZS”). This newsletter is simply an informal gathering of topics I’ve recently read and thought about. It generally covers topics related to the digitization of the global economy, technology and innovation, macro and geopolitics, as well as scientific progress, especially in the fields of cosmology and the brain. I will frequently state things in the newsletter that contradict my own views in order to be provocative. I often I try to make jokes, and they aren’t very funny – sorry.
I may include links to third-party websites as a convenience, and the inclusion of such links does not imply any endorsement, approval, investigation, verification or monitoring by NZS Capital, LLC (“NZS”). If you choose to visit the linked sites, you do so at your own risk, and you will be subject to such sites' terms of use and privacy policies, over which NZS Capital has no control. In no event will NZS be responsible for any information or content within the linked sites or your use of the linked sites.
Nothing in this newsletter should be construed as investment advice. The information contained herein is only as current as of the date indicated and may be superseded by subsequent market events or for other reasons. There is no guarantee that the information supplied is accurate, complete, or timely. Past performance is not a guarantee of future results.
Investing involves risk, including the possible loss of principal and fluctuation of value. Nothing contained in this newsletter is an offer to sell or solicit any investment services or securities. Initial Public Offerings (IPOs) are highly speculative investments and may be subject to lower liquidity and greater volatility. Special risks associated with IPOs include limited operating history, unseasoned trading, high turnover and non-repeatable performance.