De-Mystifying Design Wins
WARNING: this is much longer than a traditional blog post – 3,500 words. Think of it as a primer or reference article, only for those interested in a full discussion of a complicated subject.]
Last week, a publicly listed Canadian semiconductor company released their year-end results. As part of that, they stopped disclosing ‘design wins’, and there ensued a spirited debate in the various Bay Street analyst research reports about the size of the company’s design wins, how fast they were growing and what that all meant for the future of the stock.
My friend Brian Piccioni and I started an email conversation about the analyst reports, and we quickly realised something. At least some of the analysts don’t seem to really understand what a design win is, how it works, and the correlation (if any!) between design wins and future revenues. Even those analysts who do understand the topic weren’t disclosing their knowledge in their morning research notes, nor did the company get into any detail in its press release.
This is a typical structural failure of both Bay Street and Wall Street technology research: not everyone knows as much as they should, and if the analysts and the company were forced to do an in-depth briefing on all possible topics, morning notes wouldn’t get published until next month, and prospectuses would be 3,000 pages long!
So as a public service, I am adapting our email exchange into a question and answer interview format – we hope it proves useful to some of you. If you are wondering why you should read further, don’t think of it as an email exchange between two friends.
Instead, think of being present at a conversation between two people with over 50 years of tech investing experience between them, both with CFA designations, and both having served on the boards of multiple semiconductor and semi equipment companies. One is a former buy-side money manager, responsible for billions of dollars of investments over the years, and former portfolio manager of the Canadian Science and Technology Fund of the Year. The other is a sell side analyst who was the top semiconductor analyst in North America for three years, and the #1 Canadian technology analyst for six straight years. Oh…and he also worked in the tech industry as a designer for 13 years before coming to Bay Street.
Duncan: We need to get started at ground level. When a new technology box is being invented (let’s say it is a smartphone) the designer and manufacturer need to deal with a lot of various components that will go into the finished product…months or even years before production starts! Some of them will be fairly standard: memory chips, maybe the displays, and things like connectors and so on. There will also be a set of chips that do things in very standard ways, and the designers know that they can just go onto the merchant semiconductor market and buy as many of these parts as they need at widely agreed upon specifications and prices.
But then there will be other chips. These are not standard, but in some way are novel. Perhaps they use less power, are smaller, more integrated or do something clever that isn’t available in off-the-shelf chips. They may not even exist yet. The designer (if it is a big enough company) may be working on designing a chip that fills this role, it may decide not to, or it may decide to seek an outside supplier in case its internal design team doesn’t come up with a viable part in time.
Meanwhile, there are potential external companies that could supply the part. They might be an IDM (integrated device manufacturer such as Intel, Freescale, Samsung, Texas Instruments, etc.) or a ‘fabless’ semiconductor company (they do not own their own fabrication plant, but rely on a 3rd party like TSMC to make the chips for them.) It costs them tens of millions of dollars to do a design these days, but if they build the right part, the smartphone manufacturer might award them a “design win” (also called a socket.) If ten million phones sell, and it’s a $10 part, that is $100 million in revenues to the company, whether a fabless player or an IDM.
Both IDMs and fabless businesses track design wins. But when analysing a given company, sometimes design wins are the only metric that outside analysts have to assess early stage fabless players. A large IDM or major fabless company has existing revenues and profits, and can usually be valued by traditional metrics such as earnings, cash flows and growth. In contrast, an early stage fabless company (or one transitioning between products or generations) may have minimal revenues, negative earnings and be burning cash. In that situation, the number and dollar value of design wins can often be the two most important numbers discussed!
Although the company may not be profitable yet, a pipeline of design wins could indicate significant potential. The gross margins in the space can be 40-60%, so the fabless semi company can pay for their design costs, cover other expenses, maybe generate a profit, and make their investors’ money. And if they can get other design wins, in other products, then the money can really start rolling in!
Brian: Not so fast! While everything you say above is accurate at a high level, the story is much more complicated. As an overview: 1) design wins are rarely contractual arrangements; 2) the companies developing new products often source competitive chips concurrently; 3) many products are never brought to market; 4) many products which are commercially released do not sell as many units as initially hoped or forecasted; and 5) the companies with the best track records for product introduction rarely design mass market systems solely around the chips of small fabless semiconductor companies.
Duncan: Let’s break that down. You used to design new products for large companies. When someone wanted you (or your company) to give them a design win, how did that work?
Brian: When I was a design engineer we had frequent visits from representatives of various electronic component manufacturers: these were the people who fabless semi companies paid to sell their new products. My favorite rep was a guy called Dave who knew his product line cold. Despite a lack of engineering expertise, Dave could triangulate your requirements and find you a chip which met your needs. Plus, since we were all hobbyists, he had a knack for scoring ‘engineering samples’ (like $400 microprocessors) which eventually ended up in our home-made computers. Not quite bribery, but just part of the selling process.
As good as Dave was, there were a dozen or so manufacturer’s reps who were almost as good and they were all positioning their solutions for our next designs. Sometimes word would come down from management that we were required to use a particular component: a fate we would ascribe to a less benign form of bribery, since many of our managers rarely understood even basic electricity.
Design is a multi-step process: the designer has to understand what the system is expected to do, choose what devices are needed to get it to do that, create a circuit diagram which is hoped will work, have that circuit diagram converted into an actual system, debug it, modify the design to correct any errors, and iterate until there is a salable, manufacturable product that works.
Component selection sounds easy, but it is not. Thanks to Moore’s Law, most of the time you are forced to choose between an existing-but-soon-to-be-obsolete product, and a new product you hope will be available in production volumes around the time your product is expected to enter volume production. Price and availability are always important factors and the vendors are not going to quote on either unless they have a good handle on how many units you want and when you want them. So it was customary for our purchasing department to ask for quotes on tens of thousands, or hundreds of thousands of units…even though, if we were being honest, we really had no idea how many units would actually be needed.
In other words, Dave, and the other reps, would be asked to provide us with all the information we needed, including samples, etc., based on the expectation their device would be selected and, once production began, they would be selling hundreds of thousands of units. Once the engineers decided what components we were going to use and those decisions were accepted by purchasing and management, we would inform the lucky reps their device had made the grade and they had a ‘design win.’
Duncan: Great news for the component maker. They could probably announce that win, and take it to the bank?
Brian: I’m sure guys like Dave were happy to hear about the design win, but I am equally certain neither Dave nor the manufacturer went out and spent the money, or did much else based on that information, aside from adding it to some spreadsheet that tracked design wins. Maybe a press release if it was a public company and the design win was big enough. After all, it would be months or maybe even a year or two before we finished our design, tested the prototypes, corrected the design, and readied it for commercial release.
Not only that, but we undertook no contractual agreement with the manufacturer – we could, and sometimes did, change our minds as the design progressed, requirements changed, or new solutions became available. After all, electronics is not a static environment and everything is changing all the time. So perhaps Dave would have let the factory know he had just scored a $10 million design win, but nobody would have taken that to mean they were going to sell us $10 million in parts.
Duncan: One of the fabless semi companies I was an investor in (and was on the Board) got a design win for a very specialised chip. We were thrilled, but we also knew that the buyer had their own internal design team, plus there was another small fabless company that thought they could make the same chip. We were never sure what percent of the business we would end up getting, if all three teams were successful. As it happened, ours was the only chip that worked to specification, so we got all of the business: about $25 million if I recall. But we could have ended up with only a third of that, or even less if the buyer wanted to keep most of the business in-house. Was that normal?
Brian: Absolutely. One other thing Dave and everybody else in the electronic business understood is that the attrition rate for designs tends to be quite high. Depending on the environment less than half – often much less than half – of designs ever make it to production. In fact, in many companies, multiple ‘competitive’ designs may be underway at the same time with the expectation that only one of those is expected to be released. This hedges the impact of project delays, errors (basing a design on a component which ends up being discontinued), changes in requirements, and so on. In order to confuse the competition, some companies run multiple competitive designs and don’t even let the design teams know they are in competition with each other.
If you think about it, the same day Dave might be informed he had a design win; two or three other competitive reps would also be informed they had design wins, even though it might happen that only one, if any, would ever sell anything, or they might split the order. Did they or their manufacturers each announce an additional $10 million in future revenue? I hope not.
Duncan: I have seen press releases where fabless semi companies not only announce a design win, but that they are ‘sole sourced.’ I assume that lack of competition reduces almost all the uncertainty?
Brian: Not even close. As I’ve already noted, only some designs ever make it to commercial release, but that isn’t the biggest source of uncertainty when it comes to how many devices actually end up being sold. Until a product is actually commercially released you have no idea, just aspirations, as to the number of units which are likely to sell. It is all very well and good to look at a company like Apple and conclude every new high tech product flies off the shelves but it simply isn’t so. The market is harsh, consumers are fickle, distributors can be downright stupid, and marketing can be inept. The numbers that are quoted in the design win are not entirely fiction, but in most cases they are much larger than the eventual sales.
Duncan: Your design engineer days are a while ago now. I know a lot of things have changed in the industry. Back in the 1990s there were a lot of well-funded fabless semi companies, and they might be able to design a hot new chip for $2-5 million dollars. The costs have soared, and it is usually tens or even hundreds of millions of dollars to develop a series of products. Plus the buyers of chips have learned some tough lessons from the bubble, right?
Brian: In the late 1990s when everybody with a website, or an idea, or even an idea for a website, found ready financing, the market was moving very fast. So fast that companies needed the hottest silicon and fabless semiconductor companies were popping up like mushrooms. Large companies, including household names like Cisco, embarked on the design of systems based on these emerging companies. There was just one problem: it takes a lot of time and money to design a chip and, just as with system designs, many innovative chips never make it to market, especially when those devices are being developed by start-ups.
Imagine you’ve spent two years developing a router around a chip only to find that the chip supplier can’t supply the chips you need with the price and functionality you require. Most products have dozens of different chips and if one is not available you have no choice but to cancel the project and write off whatever investment you’ve made in it. This was a harsh lesson for manufacturers and as a result, almost all large companies will not even consider designing devices from small semiconductor companies, regardless of the apparent benefits of those devices. This is particularly true of ODMs (original design manufacturers) like Foxconn who make the vast majority of electronic gadgets and who subsist off negligible margins. It just doesn’t happen. In other words, the companies most likely to buy large volumes of chips are the least likely to buy them from small companies. Further, very small fabless players have less leverage: a larger company might be aiming to have multiple design wins or sockets on a given device, and they can play around with price and so on for individual components that the smaller player cannot.
Duncan: That makes sense. But I know that the big buyers do sometimes give a design win to the ‘little guy’, although that tends to be to de-risk a specific project. And if other, bigger semi suppliers (who already have multiple other parts with the manufacturer and a long commercial relationship) succeed in making the chip, the smaller player gets less than their fair share?
Brian: That happens all the time. But what it means, for our smaller fabless players, is that their most likely customers are the emerging vendors who are themselves financially weak, and are often keen to get a jump on their much larger competition. That means they end up making the mistakes their competitors made in the 1990s by going for the hottest parts out there. Needless to say, this is an approach which often ends in tears: after all, when was the last time you saw a TV or any other high tech product from a new vendor which wasn’t cranked out by an ODM? The companies most likely to award a design win to a small company are those least likely to be around by the time the product hits the shelves, least likely to have a distribution channel, and least like to sell in large volumes. I call these guys Happy Smiling Panda Co.
Duncan: I’ve known you for 20 years, and I think sometimes you can be too sceptical. Let’s talk about Happy Smiling Panda. Yes, maybe they are a small company now, or one with minimal market power in the space they are targeting. But that can change: I remember when a semi company received a design win from Huawei. At the time, no one had heard of Huawei, and they are now the largest telecom manufacturer in the world! Equally, there were chip companies that got designed into the first generation iPhone. Back in 2007, no one imagined how many units that device would sell, and the companies that had those design wins made a lot of money.
Brian: Yep – I’m often too skeptical for my own good, but ask yourself how many profitable fabless semiconductor companies have emerged since the Dot Com meltdown? Off hand, I can’t think of any.
You also need to remember the numbers of “lottery ticket” wins like Huawei or the iPhone are a tiny fraction of 1% of all design wins. Next, there is an asymmetric risk going on for the fabless semi company. If you get a sole-source design win from Happy Smiling Panda worth a nominal $10 million, and the product never launches or doesn’t sell well, you get nothing or not very much. If it does sell $10 million, then you get to book what you hoped for. But if the lottery ticket works and HSPCo needs $100 million worth of your chip? If you’re lucky, they only second source: they have a blockbuster product and they can’t take the risk that you fail to deliver enough product on time, so you will participate in only some of the upside. If you’re less lucky? They decide the chip is now core to their business, reverse engineer your product, and replace you entirely.
Duncan: I have seen that movie and it doesn’t end well. The tiny fabless semi company can sue the manufacturer, but 1) suing your customers seldom enhances your rep with other buyers; and 2) the company doing the suing is running out of money, just lost their main revenue source, and has little ability to fight extended battles, while the manufacturer has teams of lawyers, billions of dollars, and can wait it out.
So let’s try to summarise: if a given fabless semi company announces that it has design wins worth $150 million, what does that likely mean for revenues over time?
Brian: That’s the key question, of course. Design wins aren’t enforceable, and there aren’t even any agreed upon accounting definitions under IFRS. Further, as an outsider, we lack good enough information to properly evaluate a pipeline of design wins. Some wins are sole-sourced and some are shared; some are with better or worse customers, for better or worse products, and in better or worse markets. But looking at history, I would state that I am unaware of a single company that has booked 100% of their design wins as revenues within five years.
That’s just the starting point: I would go further, and say that on average, most companies book less than 50% of design win dollars as revenues. If you wanted my best estimate of the appropriate discount? Perhaps 66%: only a third of design win dollars get converted to revenues, and even less to profits, of course. And that is over time, so the present value of a design win that is a few years out needs to be discounted further. But as a rule of thumb, a small fabless semi company that says it has $150 million of design wins is likely to book only about $50 million of those as revenues.
Duncan: That feels about right to me. When we read about a design win, we need to know that it is NOT the same as ‘backlog.’ Backlog represents firm orders that have not yet shipped, enforced by contract and with cancellation and take-or-pay provisions, and it is a term which does have an accounting definition.
Then again, we need to not be too negative. For any given semiconductor company: 1) getting a design win is always a good thing; 2) having the total dollar value of design wins going up is always a good thing; 3) but there is not a one-to-one relationship between design wins and revenues.
Thanks to Brian for participating, and if you want to hear more of his informed thinking, check out The Geeks Reading List, and then subscribe. Every week he puts together ~20 articles on issues of interest to the tech community, with his own perspectives.
I would also like to thank John La Bouff, a former Deloitte colleague who is a semiconductor industry expert and advisor, and who reviewed this article and made some suggestions. All responsibility for the final content is ours of course. You can find John on LinkedIn.
[Edited to add:] By the way, nothing in this post except the intro is a comment on the Canadian public company that reported. I don’t know them, haven’t met them in years, and have no idea how good or bad their design win pipeline is, or what it means. They were simply the starting point for an in-depth discussion of design wins, and nothing we say should be construed as praise, criticism, endorsement, investment advice or anything.