Good morning. My name is Jamie, and I will be your conference operator today. At this time, I would like to welcome everyone to the Teradyne fourth quarter 2017 earnings conference call. All lines have been placed on mute to prevent any background noise. After the speaker's remarks, there will be a question and answer session. If you would like to ask a question during this time, simply press star, then the number 1 on your telephone keypad. If you would like to withdraw your question, press the pound key. Thank you. Mr. Blanchard, you may begin your conference.
Thank you, Jamie. Good morning, everyone. Welcome to our discussion of Teradyne's most recent financial results. I'm joined this morning by our CEO, Mark Jagiela, and CFO, Greg Beecher. Following our opening remarks, we'll provide details of our performance for 2017's fourth quarter and full year, along with outlook for the first quarter of 2018. The press release containing our fourth quarter results was issued last evening. We're providing slides on the investor page of the website that may be helpful to you in following the discussion. Replays of this call will be available via the same page after the call ends. The matters that we discuss today will include forward-looking statements that include risk factors that could cause Teradyne's results to differ materially from management's current expectations.
We encourage you to review the safe harbor statement contained in the earnings release, as well as our most recent SEC filings. Additionally, those forward-looking statements that are made as of today, we take no obligation to update them as a result of developments occurring after this call. During today's call, we'll make reference to non-GAAP financial measures. We've posted additional information concerning these non-GAAP financial measures, including reconciliation to the most directly comparable GAAP financial measure, were available on the investor page of our website. Between now and our next earnings call, Teradyne will be participating in investor conferences hosted by Goldman Sachs, Morgan Stanley, Citi, and Susquehanna. Let's get on with the rest of the agenda. Mark will comment on our recent results and the market conditions as we enter the new year.
Greg will offer more details on our quarterly and full-year financial results, along with our guidance for the first quarter. We'll answer your questions. This call is scheduled for one hour. Mark?
Good morning, everyone, and thanks for joining us. In my remarks today, I'll provide a brief summary of 2017, share additional insights on our strategy, describe our 2018 outlook for the markets we serve, and close with comments on our capital allocation plans. Greg will take us through the financial details. The fourth quarter capped off a strong year for Teradyne. Sales were up 26% compared with the year-ago quarter, and non-GAAP EPS was up 44%. All business segments showed double-digit sales growth over the fourth quarter of 2016. In SemiTest, our memory test shipments hit record levels in the quarter, which we are set to exceed again in Q1. SoC sales in the quarter were also strong, with EagleTest and service revenue up 29% and 20% respectively.
Universal Robots sales were up 61% compared to the year-ago quarter, as shipments to the Asia-Pacific region, in particular, exceeded our forecast. For the full year of 2017, Teradyne sales grew 22%, and non-GAAP earnings per share grew 55%. A healthy, growing test business combined with an exciting hypergrowth collaborative robot business resulted in Teradyne breaking the $2 billion revenue mark and exceeding our $2 EPS target three years ahead of plan. Market leadership with differentiated products and a disciplined financial model have allowed us to grow with high leverage to the bottom line. Over the last 10 years, we've methodically grown our share of the combined SoC and memory test market from 24% to about 50%, while improving gross margins. We've taken a disciplined approach of targeting the attractive growth subsegments of the market.
Focusing on technology inflections and market shifts in mobility, automotive, and flash memory paved the way for these gains. This will continue to be our game plan going forward. The ongoing increase in device complexity from advancing transistor count, multifunction cores, and advanced packaging, combined with shorter time to market and shorter device life cycles, have been a balloon for the SoC test market. Here, the diminishing value of increasing parallelism in test has changed the trajectory of the overall market to sustained growth. This doesn't suggest that smooth monotonic market growth in tests. We expect there will continue to be tooling cycles yielding both quarterly and annual volatility. The overall trend line is positive. In addition to complexity growth, new technologies are making their way into the market that create additional opportunities for tests.
One example is millimeter wave technologies, which are in development for mainstream deployment in a variety of communications applications as well as vehicular radar. Another example are emerging LIDAR and time-of-flight sensors and various positioning technologies. These technologies require a new class of test equipment that will create growth segments well into the next decade. In memory, on top of bit growth, we are seeing increased diversification of application-specific memories. This diversification tends to center around a variety of interface protocols that both up the test intensity and provide the opportunity for Teradyne to differentiate our products. Over the past three years, our Magnum memory tester has capitalized on the shift to high-speed interfaces for the final testing of flash memory devices, driving us to the number one position in this segment.
While we continue to ride that wave, in 2018, we are also expanding our served market by extending the Magnum platform to wafer test. While Teradyne's R&D teams are focused on creating instrumentation to test these interesting new technologies, a large part of our investment goes into the software for our testers. Importantly, our software focuses on reducing the complexity of testing these new technologies. Over time, our software gets stitched into the fabric of our customers' development environment to allow their design teams a seamless, efficient time to market advantage. With design cycles ever shorter and ramps ever steeper, having the software toolkit to allow global teams to develop and release reliable test programs is another key advantage of Teradyne. At Universal Robots, we continue to expand our applications and ease of use through our open architecture Universal Robots+ program.
With over 55 certified product enhancements and more added every month, customers are enjoying a growing toolkit of options to put our UR cobot into widening applications. At UR, we do what we do best, focus on creating the world's best universal cobot arm and the best intuitive user interface. We then open that architecture to invite the creativity of a global ecosystem of developers to create plug-and-play add-ons. Additionally, while rising labor costs and labor scarcity improve the economics of automation, the value proposition of collaborative robots is much more than just lowering costs. For example, in December, we established operations in Bangladesh, aiming to leverage the success we've seen in the Indian, Sri Lankan, and other low-cost markets, where customers value the repeatability and accuracy of our cobot, resulting in higher quality manufacturing and less rework.
At the other end of the spectrum, the value of repeatability, accuracy, and speed is also evident in a project underway at NASA's Langley Research Center. To speed up and improve the inspection of aircraft composite structures, our cobot is used to create an inspection pattern that doesn't miss any areas. Using our UR10 cobot with a third-party infrared inspection system and software from a Universal Robots+ partner, a single operator can oversee the inspection of complex composite structures such as aircraft fuselages. Additionally, the UR10's built-in safety features allow people to work closely alongside the cobot while it's inspecting the aircraft. That allows other inspections of manufacturing processes to take place during the infrared inspection, which improves the efficiency and saves time.
Closer to home, not only are UR cobots used to make new UR cobots, our SemiTest contract manufacturing partner has extended that practice to some of the assembly tasks in building our own UltraFLEX semiconductor test product. In wireless test, LitePoint had a strong year in a weak production test market. We returned to model profitability, began early production ramp of our 802.11ax product, and delivered initial millimeter wave and sub-6GHz testers for 5G cellular. While we are about a year away from the impact of 802.11ax and about two years away from the beginning of the 5G ramp, we've been active in positioning our next generation of production-focused testers for the 5G era. Turning to market outlook. As a baseline, in 2017, we saw a combined SoC and memory test market of about $3.35 billion, up from $2.8 billion in 2016.
The 2017 SoC test market was about $2.65 billion, and the memory test market was about $700 million. On top of that, we added about two points of share to reach 50% share of the overall market. Our view of 2018 has strengthened slightly since our October preview. We now see the SoC test market in the range of $2.4 billion-$2.8 billion and the memory market in the range of $700 million-$800 million. Additionally, the annual spring and summer ramp for mobility tooling looks to be about four weeks later than recent years, resulting in a shift of some revenue to later in 2018. Unlike the past two years, where the first half sales were about 55% of annual sales, this year, we expect first half sales to be more like 2015, when we shipped 52% in the first half.
At Universal Robots, our aggressive strategy has yielded accelerating annual growth, 58% in 2015, 62% in 2016, 72% in 2017. We are delighted in the progress we are making, and we continue to add staff, distributors, integrators, and ecosystem partners to give us an even broader reach across more end markets in more regions of the world. While we'll invest more in operating expenses, we also expect operating profit to remain in the mid to high teens. Our goal remains to grow at 50% or better in 2018. In wireless test and system test, we see 2018 as essentially flat with 2017. Finally, I'd like to give you a high-level view of our capital allocation plans for 2018. As we've said, we plan to maintain a balanced approach to capital allocation, consistent of dividends, buybacks, and disciplined M&A.
The strength of our test business over the past several years has enabled a consistent approach. A growing industrial automation business gives us a platform for investment. Tax reform gives us some additional flexibility in how we deploy our capital. The board has authorized a $1.5 billion repurchase program, of which we plan to buy back at least $750 million in 2018. Furthermore, a strong outlook in our core test businesses gives us the conviction to increase our quarterly dividend 29% to $0.09. Finally, we remain committed to fully exploit and expand the leadership position we have in test and industrial automation. Our M&A pipeline is focused on that mission. With that, I'll turn things over to Greg for the financial details.
Thanks, Mark. Good morning, everyone. I'll start with the highlights of 2017. I'll then offer comments and perspective on 2018, including our capital allocation plans and the impact of the newly enacted tax laws. I'll also offer perspective on our strategic position and market trends at the business segment level, outline our new midterm financial model. I'll close with the fourth quarter results and first quarter outlook. On the financial highlights front, as Mark noted, our $2.14 billion of sales in 2017 was up 22% from 2016, driven by a strong ATE market, SemiTest share gains, 72% sales growth at Universal Robots. Non-GAAP EPS was $2.34 for the year, up 55% from 2016 and exceeding our 2020 $2 EPS target three years early. In 2017, we performed quite nicely against our key objectives.
We gained two points of share in SemiTest, making 2017 the sixth consecutive year of ATE share gains. We grew Universal Robots' top line 72%, well above the 50% threshold target. We had model operating performance in wireless test, MiR and production board tests. We also bought back $200 million of our stock at an average price of $34.30. Since the beginning of 2014, our balanced approach to capital allocation has returned $839 million to shareholders in buybacks and dividends and invested upwards of $350 million to acquire Universal Robots. When we bought Universal Robots in 2015, its sales for its most recently completed year of 2014 were only $39 million. Over the subsequent three years, they've grown over four-fold inside of Teradyne, with sales of $170 million in 2017.
With the new $1.5 billion share repurchase authorization, we plan to repurchase a minimum of $750 million of our stock in 2018. The increased quarterly dividend of 29% to $0.09 a quarter is effective immediately, subject, of course, to ongoing board approval. Our strong operating model and balance sheet comfortably support these moves. On the operating model, over the last three years, we've averaged a 23% non-GAAP operating profit and $400 million of annual free cash flow. Over the next few years, we plan to steadily migrate our cash and marketable security balance of $1.9 billion down, while at the same time maintaining dry powder for our attractive M&A, subject, of course, to comparison against even more aggressive buybacks. We do see a number of attractive candidates in automation, particularly in software, aimed at extending cobots into new applications. The newly enacted tax laws shift the U.S.
from a global tax system to a territorial tax system, subject to a foreign minimum tax rate of 13.2%. As our foreign rate is estimated to be just above this threshold, we do not expect to be subject to this new U.S. tax. The new toll tax on our accumulated foreign earnings will cost $150 million, payable over eight years. Due to the lower U.S. federal tax rate, we also wrote down our deferred tax assets by $34 million. These two one-time charges are included as discrete items in our Q4 GAAP results, but not in our non-GAAP results. Looking at the segment level, annual highlights, and trends.
Starting with SemiTest, we had another year of ATE market share gain, adding two points to bring us to an estimated 50% share in 2017. These gains were in SoC tests and more than offset the few points of decline in memory test share due to short-term buying shifts. We also scored the number one spot in the VLSIresearch survey for semiconductor equipment suppliers. Shifting to Universal Robots, we grew sales 72%, from $99 million in 2016 to $170 million in 2017. Universal Robots also had an operating profit rate of 19% in 2017, more than double the 9% rate in 2016. Teradyne supply chain group delivered several gross margin points of improvement, and we can see further improvements as we move towards our target rate of 20% or better by 2020.
Apart from the strong financial numbers, we strengthened Universal Robots' strategic position, as Mark highlighted, by growing our UR+ ecosystem platform with more grippers, vision systems, and other cobot accessories. This online platform is another key step in making it even easier to deploy our cobots and deliver a faster ROI. This easy access to proven solution reduces deployment time and project risk for our customers. We also opened up UR Academy, which is an online interactive training platform offered in multiple languages. Training is free, and in less than two hours, anyone can learn to train a UR cobot. The pattern that you no doubt see is that we're making it easier and easier to deploy our cobots.
We, along with third-party research firms, expect 50%+ robust growth in the market for years to come, as the applications being served by cobots today are just scratching the surface of what's possible. There are hundreds of highly repetitive tasks that are not well-suited for people. For example, some repetitive tasks require very high precision, such as applying the exact amount of glue, or are ergonomically challenging, such as driving screws with force, or are simply too tedious, such as machine tending or pick and place material handling. There are also new emerging technologies that should further extend the cobot served market, such as lower cost vision systems for more advanced pick-and-place applications. The picture we see is that undesirable tedious tasks will be increasingly automated, and the human workers will be performing the higher skilled and safer tasks.
While we expect capable competitors to emerge, today, the challenges are much more about building end user awareness. To that end, we'll have more robust digital and print awareness pieces, along with a new digital lead generation program in 2018. Shifting to system tests, our defense and aero and space and production board test groups together grew 2017 sales 4% over 2016 and operated above model profitability for the second consecutive year. We expect similar performance in 2018 as defense and aero continues to benefit from the deployment of new weapon systems and new aircraft like the F-35, while in production board tests, our high throughput tester designed for high panel count applications continues to gain traction in the growing automotive space.
In storage tests, we return to model profitability in the second half of 2017 after incurring losses in the first half, bringing to market a new product aimed at system level test of semiconductor devices. Storage test should operate at around $60 million in sales and model profitability in 2018. Turning to wireless tests at LitePoint, the group rebounded very nicely from 2016 with sales growth of 16% and above model profitability in 2017. We expect 2018 to be similar to 2017 as we position for a near-term market uptick, driven by the new 802.11ax wireless standard in 2019 and 5G cellular thereafter. For the full year at the company level, non-GAAP gross margins of 57% were at record levels and improvement of two points from 2016 due to higher volume and favorable product mix.
Before I move into the more granular fourth quarter numbers, I will note that after this call, we plan to no longer provide bookings information as the short-term timing of order placement adds unnecessary noise with little added insight to our business. I also won't cover bookings in my prepared remarks, but the numbers can be found in the press release and investor decks. SemiTest sales were $317 million in the fourth quarter, with SoC making up $250 million and memory test at balance at $67 million. SemiTest service revenue totaled $70 million. Universal Robots had sales of $54 million, a quarterly record. Geographically, fourth quarter UR sales were 47% in Europe, 24% in Asia, 23% in North America, and 6% in the rest of the world. Our largest channel partner was only 3% of our total 2017 sales, so again, we have very wide diversity.
System test sales were $80 million, up 125% from the third quarter due to revenue recognition of our new system level test product. At Wireless, sales were $28 million. At the company level, our fourth quarter sales were $479 million. The non-GAAP operating profit rate was 23%, and non-GAAP EPS was $0.46. We had one 10% customer in the fourth quarter and one for the full year. Non-GAAP gross margins were 57% in the quarter with favorable product mix. You'll see our non-GAAP operating expenses were down $3 million to $160 million compared to the third quarter due to lower variable compensation accruals on lower profits. Regarding our earnings power, when we set our original 2020 $2 EPS target, we conservatively estimated the ATE market would grow on average 1% annually from the 2014-2015 average market size of $2.7 billion after many years of sequential decline.
Through 2017, the actual annual growth rate has been much stronger at 11% and is a key reason why we achieved our $2 EPS target three years early. With that recent history, combined with our latest outlook, we've updated the earnings model target to $3.50-$4 of non-GAAP EPS by 2021, driven by some very familiar themes. First is strong annual growth at Universal Robots. We've used 45%-50% for modeling through 2021, which would bring Universal Robots to $750 million-$860 million in sales in the midterm. Next is further SemiTest growth at a more modest rate than the last few years. We've modeled 3%-4% growth rate from the 2014 average market size and a few points of share gain through 2021. Finally, repurchases, model performance, and modest growth elsewhere will contribute to the EPS goal.
We've detailed the new model in our investor deck. Looking a bit closer in at 2018, we expect gross margins to be approximately 56%-57%. In OpEx in 2017, non-GAAP operating expenses were up $49 million versus 2016, driven principally by distribution investments in Universal Robots and higher variable compensation tied to increased profitability. In our test businesses, apart from normal changes in variable compensation tied to profit levels, our total OpEx spending in 2017 was down slightly from 2016. Looking ahead to 2018, we plan to keep test OpEx flat, apart from normal changes in variable compensation, and to increase UR OpEx from $16 million a quarter exiting 2017 to upwards of $30 million a quarter by the end of 2018. These investments should strengthen both distribution and product development and help fuel another 50%+ growth year in 2018 and beyond.
Shifting to our outlook for the first quarter, sales are expected to be between $460 million and $490 million, a 4% increase at the midpoint from the year-ago first quarter sales of $457 million. The non-GAAP EPS range is $0.38 to $0.45 on 200 million diluted shares. Please note that while we are confident of our full year 50%+ growth plan at UR, the first quarter revenue growth rate compared with last year will likely be lower than 50% due to the pricing related changes that drove very strong demand in the first quarter of 2016. The first quarter guidance excludes the amortization of acquired intangibles and the non-cash imputed interest on the convertible debt. First quarter gross margins are estimated at 55%, down two points from the fourth quarter due to product mix.
The first quarter OpEx running at 34% to 36% of first quarter sales is up about $5 million from the fourth quarter due to further distribution and product development investments at Universal Robots and some NRE project spending in SemiTest concentrated in the first quarter. The non-GAAP operating profit rate at the midpoint of our first quarter guidance is about 20%. Non-GAAP interest income, excluding the non-cash imputed interest from the convert, is expected to be about $2 million quarter in 2018, factoring an interest income on our cash balances, partially offset by the 1.25% annual coupon on the convertible debt. In 2018, we've earmarked $80 million to $100 million for CapEx. We start 2018 with very strong momentum after achieving our $2.20 EPS target three years early and are forging a path to $3.50 to $4 of EPS by 2021.
Rising device complexity, IC unit growth, and increasing quality requirements are driving SemiTest, while Universal Robots is expected to grow at 50%+ for years to come. Our wireless test business is back to healthy profits and positioning for the next demand wave, and we expect healthy performance in system tests. Layer on top of that, increased shareholder returns with our new $1.5 billion share repurchase program, and a 29% increase to our quarterly dividend, and the future of Teradyne looks quite bright. With that, I'll turn the call back to Andy.
Thanks, Greg. Jamie would now like to take some questions. As a reminder, please limit yourself to one question and a follow-up.
Thank you. At this time, I would like to remind everyone, in order to ask a question, please press star, then the number one on your telephone keypad. Our first question is from Toshiya Hari with Goldman Sachs.
Good morning. Thanks for taking the question. Mark, you made a slight change to your 2018 SemiTest market outlook today. I guess, what were some of the developments you saw over the past three months that led to that change? If you can talk a little bit about some of the sub-segments within SoC, mobility, automotive, image sensors, et cetera, that would be helpful. Thank you.
Okay, sure. Since November, we ended up in fourth quarter of 2017 above the high end of the guidance we had for revenue. We've seen throughout November, December into January, a steady increasing demand for SoC capacity, more so than we expected. It's pretty broad-based across everything except mobility, where mobility, as I mentioned in my remarks, has shifted maybe four weeks further into the future. Automotive, image sensor, linear, all of these areas were getting pulled for reasonably quick buy delivery cycles. Typically, our lead times are 8-12 weeks, so we get a reasonable amount of turns within the quarter sometime. It's that sort of across the board demand that has brightened the outlook looking forward.
Within mobility itself, although there's been a bit of a, I'd say the peak tooling installations this year about four weeks later than last year or the year before, the demand there also looks very strong. On the SoC side, I think it's just a broad-based rising of demand. I would point out that with about an 8-12 week visibility, we don't get to see the back end of the year with any precision, until we're sort of into the back end of the year. A lot of what customers are talking about as a general budgetary number, is kind of flat with 2017. On the memory front, just to turn to that for a minute, things are very busy and hot there. I mentioned we'll exceed our Q4 shipments in Q1 in memory. Similar kind of lead times.
The $700 million-$800 million range hasn't changed since November. If anything, there's pressure there on the upside. I could see that going up, especially if in the indigenous fabs in China, we see that the yields get to a sufficient level that they begin to turn on production. We haven't assumed that for 2018. We've assumed that's more of a 2019 phenomena, but that could be upside on the memory discussion.
Great. Thank you for the details there. My second question is on Universal Robots. In your long-term model, you've assumed 45%-50% top-line growth for the business. I guess I was curious what the rationale was for that number. Why not 30? Why not 70? Then, I guess related to that, if my math is correct, I think UR gross margins improved three, maybe four percentage points in 2017 versus 2016. What's the path forward there as you look out to 2021? Thank you.
Okay. This is Greg. I'll start, and maybe Mark will add to it. I'll start with the gross margins. I'll go back further. When we acquired Universal Robots, the gross margins were about 51%. As we end 2017, we're in the upper 50s. We see a path to increase margins a few more points. Now, this doesn't factor in what does the competitive field eventually unfold and look like. Today, as I think I mentioned in my prepared remarks, it's much more about stimulating awareness versus we're going up against competitors. We do see opportunities for further margin improvement, but again, at some point there'll be some competitors that we'll bump into, particularly at large accounts. On the first part of the question, why not 30? Why not 70? Well, we look at third-party reports. There's a number of them out there.
We speak to our distributors, our channel partners. We see that our existing channel partners grow their business about 50% a year. They sell 50% more cobots each year. There will come a point where we'll probably slow down or stop adding new distributors or integrators in terms of net increases. There might always be some adjustments for quality. If I say, what do existing ones sell today? It's about 50%. We're constantly improving in terms of some of the offerings we're giving them, as well as on the UR+, what's on the online website, the Academy. We've seen sales grow at a higher rate each year, which is unusual to see the growth rate increase. We all know as numbers get larger, that gets harder to continue.
In the end, it's a judgment, we come at it from many different perspectives, we also look at bottoms-up, the types of jobs that can be automated, that leads you to very astoundingly big numbers, then we kind of back off of that because the numbers just look too big looking at it that way. It's judgmental, we look at it many different ways, our track record to date has been well above 50%, we know as numbers get bigger, that'll get harder to be well above 50%.
Thanks so much.
Our next question is from Mehdi Hosseini with SIG.
Thanks for taking the question. Just back to your ATE forecast. Is there any impact, or are you anticipating any incremental market impact from some of the new ASIC designs that are coming to the market? These new ASICs are designed for blockchain computing, like a cryptocurrency. I'm just wondering if the market is big enough for the ATE market and Teradyne specifically to see that in your forecast. I have a follow-up.
I think cryptocurrency processing as a sub-segment of the ATE market is really de minimis. It's not significant in the grand scheme of things. On a broader scale, though, I think the idea of array processing, whether it's used in edge applications like AI for things like facial recognition or voice recognition, that's becoming a discernible segment and growing. I'd say $several hundred million of that TAM, you could attribute to the combination of all the cryptocurrency processing plus GPU, plus AI-related processing. Our participation, Teradyne's participation in that space, in the traditional GPU side has been low, below our average market share. In the emerging edge space of applications, whether that's in automotive or in cell phones or in infrastructure areas, is kind of right at the norm, right at that 50% point.
A couple of hundred million in TAM, growing faster than average, we are participating outside of GPU.
Is this what's driving your market share target specifically for 2021?
No, not specifically. I think we will grow with the market there. I think our market share gains for 2021 will come from the digitization of automotive, where the combination of our strength in traditional automotive electronics, and our strength in mobility will help us in gain share in that area.
Got you.
That's probably more-
Sure. Got it. Thank you. A quick follow-up for Greg. I appreciate for updating us on the longer-term earning power. It seems to me there isn't really much of a leverage left in the P&L, and these ranges that you're targeting, $3.50-$4, to a large extent depends on your ability to execute on a share gain. In that context, what is it that gives you as a rather conservative company to step up and discuss these share gains? B, what if we hit an air pocket, and what is your flexibility in managing cost, given the fact that you're still investing in some of the growth areas?
Got it, Mehdi. The slide we've put forth, which we hope will give you a very clear path, we've put the assumptions on the upper half, and frankly, the biggest assumption is Universal Robots growth rate. That's driving most of the sales and earnings growth. Let me go to the SemiTest question on market share. We have steadily gained market share in ATE, in both SoC and memory over many years. We've shown that on this slide here, the last four-year period, we grew from 44%-50% market share. For modeling, we're using anywhere from four to six points. We're modeling what we've done in the past, and we've been a bit conservative saying it could be as low as four versus six points. If you go back further, we've gained much more than six point of share.
We do have the ability to carefully target the right segments, grow with those waves, as well as find areas for differentiation and not compete against capital cost. It's a long-running formula that we've had at Teradyne that I think has proven that it's sustainable. There's always levers, Mary, Mehdi, in the model. In here, there's a bunch of variable compensation tied to profits. If the profits are lower, the variable compensation comes out. There's a lot of spending that goes with the growth. If we see the growth of any business isn't what we think it is, then we'll pull some of the spending back. There certainly is flexibility.
Got it. Thank you very much.
Our next question is from Patrick Ho with Stifel Nicolaus.
Very much. Mark, maybe first off, in terms of the SemiTest market that you talked about and the outlook you provided. With some of the timing, I would say, push outs on the mobility end, what are your concerns there that maybe that pushes out potentially into 2019 or so? Do you feel that it'll come that four weeks later that you're talking about right now, and do you have that visibility, or is this something that could potentially extend a little further?
Yeah, I think that's one area where you have pretty good visibility. No, it will not extend into 2019. What will happen is that some shipments that would normally be in second quarter will leak over into third, and it's about that four-week window is the delta, but not a lot of ambiguity in that.
Great. That's helpful. Maybe Greg, in terms of some of the commentary you made about UR costs going up, it clearly makes sense because obviously you're growing it at a very rapid rate. Can you give a little more detail? Is it going into more on the distribution side of things and particularly Asia, given the rapid growth there? Or is it a mix of both the distributors as well as integrators?
By far, it's distribution, and that's the single biggest item is more salespeople that can then work with distributors and channel partners in new regions that we're not covering adequately, as well as existing regions where there's some accounts we're just not getting to. Then the next is engineering to continue the innovation that we deliver to step that up even further. Those are the two big areas with distribution by far the largest.
Great. Thank you very much.
Our next question is from the line of C.J. Muse with Evercore.
Thank you for taking my question. I guess first question, going back to some of the prior questions around your overall SemiTest guide. In the prepared slide deck, you talk about growth there, but in terms of your outlined expectations, you're guiding to flat. Curious how we should interpret that, and as part of that, how you are thinking mobility spend on the SoC side year-over-year, and is that sort of part of what's driving that flat guide?
Yeah. One thing I'll start out with as a caution. A year ago in this very call, we gave a guide for the SoC market that was in a range of $2.2 billion-$2.5 billion, and we ended up at $2.65 billion. With that caveat, we're not completely able to see what's going to happen. In mobility, I would say year-over-year, our expectations is that it will be flat to very slightly up. That market is one that typically the majority of the installations come to conclusion by the end of third quarter. We have a little bit more visibility there. I think what's less clear to us and what surprised us last year is automotive. How much longer will that have legs? It's been going now for about a year and a half.
If automotive continues at the rate that it did last year or even a little bit of acceleration, that moves us toward the high side of the SoC market. Industrial linear. Those are two that are very kind of quick turn businesses that we get very little visibility beyond the current quarter. Mobility, everything else being equal, roughly flat, and then industrial, linear, automotive, the wild card that swings the rest of it. I'll just make a comment about memory. Memory, as I said before, the range we gave of $700 million-$800 million could easily have another $100 million or [$200 million] on top of that if the indigenous Chinese fabs really decide to or are able to go into a production ramp. That's something we've modeled into 2019, but that could pull in.
Very helpful. As my follow-up, as you think about UR and building out your infrastructure to support that, curious if you've seen sort of any momentum year-to-date related to tax reform, accelerated depreciation, and impending industrial accelerated spend, whether or not that's changing perhaps your nearer term outlook for perhaps faster growth on that business.
C.J., we haven't seen that yet in North America, but we will have our antenna up to see if that's possible. As you know, the cost is fairly low, but I certainly could see that might make its way into the ROI calculation. Most of the ROIs are very quick without a further advantage, a tax advantage. Often the cobots are improving quality or doing tasks that aren't well-suited for human. Sure, it might help a little, but we haven't seen it yet.
Great. Thank you.
Our next question is from Richard Eastman with Baird.
Yes, just a quick question on UR. Could you just go back a little bit to how for calendar 2017, how UR's business model did shake out? I think you referenced high 50s gross margin at year end. I'm curious, what does the model look like as we go into 2018 with the growth here? Are we kind of expecting flattish operating margins close to 20%? How are we modeling that out for 2018?
Okay. I'll start that one. There's actually a range because the growth rates are so high, there's a stretch growth rate, and then there's the growth rate 50% plus. There's a range of scenarios we could be in, and we have to put some OpEx in early, and then we try to meter it where we can. This year we hit 19% operating profit with end of the year with about 58% gross margin. We actually wanted to bring more people on quicker, frankly, to position for another wave of high growth. If we end up at 19% or 20% profit, we're delighted. If we end up at 17% or 18%, that might be a great story too if we've done the thing strategically. We're not driving it with a sharp eye on profitability.
It's much more on the growth and are we getting further ahead is what we're monitoring and pushing hard on. The market has many years of high growth thereafter, and we want to extend our lead through with the best well-trained distribution partners, the best third-party ecosystem accessories on our website, easy to do training, and then development that makes it even easier to program. The more we can get ahead, the more the numbers naturally in the subsequent years will certainly meet our midterm model.
Okay. Understood. Just a question. Mark, we had spoke maybe off of the third quarter results as well, and we've ticked up our SemiTest market size twice now in the past, whatever, 4 or 5 months. I'm curious, as we look into 2018, given the puts and takes of Teradyne's share in memory versus SoC test, would you expect to expand market share in 2018? Your slide makes it clear that you're looking for similar share gains out to 2021. I'm curious, how do you think the Teradyne share, does it hold in 2018 or does it expand in 2018?
I think we would be delighted if we held our plans or to gain a bit of share. If you look at the two segments, we're 50% of the overall market, but in SoC, we're more toward the mid-50s and in memory, closer to 30. To the extent the memory business hits the upside or exceeds the upside, we will participate in that. We might gain a little bit of share, but optically, when you roll the whole year together, it looks like the share could be flat when that all gets said and done, even though we might pick up share in SoC. It really depends on the math around the size of the memory market, where our share tends to be below the combined average.
I see. Okay. Very good. Can I just ask one last question? What share count are we looking at for the first quarter guide from an EPS perspective?
200.
200.
200 shares.
Okay. Thank you.
Our next question is from Farhan Ahmad with Credit Suisse.
Thanks for taking my question. My first question is just on the deceleration in growth that you're forecasting for March. Can you just talk about why are we seeing growth decelerating from something like 20% for last three quarters to 4% in March quarter? Is there a risk that that further decelerates as we go later in the year?
I think that the growth in the quarter, there's a variety of factors. One thing we talked about was Universal Robots being up less in Q1 over last year's Q1 due to the price increase we instituted a year ago that drove an extraordinary amount of buying into the Q1 2016 period. UR will likely be up less than it's been growing at. If it's been moving at, say, last year, 70%, Q1 could be closer to 25% up numbers. That's one factor. I think the other thing is with mobility having moved out four weeks, what you see is that Even last year or the year before, we began shipments in Q1 for the tooling for mobility. There's less of that in this Q1 because of that four-week shift.
Got it. My second question is on SoC test market. You kind of talked about the impact of memory tests becoming larger, and that might weigh down on your overall share in the combined ATE market. I wanted to dive a little bit on the SoC test market. My understanding is that you don't play in the CPU, GPU, in server markets, for example, where things are pretty strong this year. Is that going to be a headwind for your market share in SoC test overall this year? Or do you think you have some products in that area as well that will help you gain share?
Yeah, I think we have participation in segments of that'll help us. That market was relatively strong last year, but as I mentioned to an earlier question, the crypto GPU straightforward application sub-segment, it's sub-$100 million. When you roll GPUs together in the whole thing, it's a couple hundred million USD. Outside of the strict GPU market, we do participate in some of these applications. I don't think the market itself is of the size, and our share position in that market would be below the 50% average. I don't think that's going to be a factor in moving our share up or down.
Got it. Thank you. That's all I have.
Our next question is from Atif Malik with Citi.
Hi, thanks for taking my question. Congratulations in meeting your long-term EPS target three years earlier. Hopefully, you can get to 2021 target earlier as well. Mark, I have a question on system test business. Your sales and orders grew strongly in December quarter. Can you just talk about what are some of the new products driving the momentum here, and how big the opportunity is for these new products? I have a follow-up for Greg.
Okay. Well, as I mentioned, as we went through last year, inside our system test group, we've had a storage test business that's been primarily focused on testing hard disk drives. That business has diminished over the years as the hard disk drive growth has slowed, and we repurposed that platform last year for more of a semiconductor testing application called system level test. We recognized most of the revenue for that initial project with the initial customer in the fourth quarter. The fourth quarter bump there that you saw was primarily getting through deliveries and acceptance of that first instantiation. Frankly, looking forward, we haven't put a lot of growth into that product in 2019. We're at the phase now where we can go beyond this initial customer and look for some other opportunities in the market.
That's what we'll be doing in the first half of this year. We have a sort of a steady business now with one account, and we'll be exploring whether we can expand that over the next six months. Sometime by the summer, we should have a verdict on whether this can be a growth business or not.
Great. Greg, with higher accessibility to offshore cash, you touched on the M&A strategy, but you talked about software being an area of interest, but you're using third parties for infrared sensors. Is your strategy on UR fundamentally different from Rethink Robotics, using sensors and AI? Can you just talk about or elaborate on your approach on UR?
Okay. I think Mark will pitch in on this one, too. Our approach has been to distinguish our cobot based upon ease of use. That is the very different approach we took to automation so that a shop floor operator and not an engineer is able to train our cobot and repurpose it when demand changes in their factory or at a small manufacturer. That's why somewhere about half of our customers are small and medium enterprises. It's a different approach than what I believe others have originally taken. There may be others trying to do what we're doing, but it's not easy to do because you also have to build up a distribution and an integrator network that can be good selling one or two cobots at a time. Most companies we find like chasing, the very large orders, which tend to be multi-year sales.
We've tended to have a different focus and a different go-to-market approach. We're going to continue to do things that make it easy to use the cobot, and there'll be new enabling technologies. I mentioned vision modules coming down that'll make using vision more likely for bin picking or advanced applications where vision can help you getting deeper into a tray or a bin. I think there's a lot of innovation yet to come in areas around the cobot that the cobot can take advantage of.
Thanks.
Our next question is from the line of Edwin Mok with Needham.
Great, thanks for taking my question. My first question is in UR margin. If I take your OpEx increase on face value is growing at a faster rate than the 50% plus top line growth. Does that mean that we might see some operating margin pressure this year on UR?
UR and the gross margin line you're referring to?
On OpEx, you said that you will grow from $16 million exiting last year to $30 million exiting this year, right? That's growing at a faster than 50% rate, right? Wouldn't that have some pressure on your operating margin this year?
It is possible that we may not end up at 19% or above where we ended up this year, because as I mentioned, we didn't bring out as many people as we wanted to. We actually weren't targeting 19%, we were targeting 15%, it's a challenge bringing out as many people as we wanted to, and that got somewhat delayed. I don't consider it a challenge. If we end up next year at 17% versus growing to 19%, to me, that's not a challenge. It depends on what we look like. If we get to 17% as a number and we're growing at a healthy rate, and we see the next year as another strong year because we've put in the investments necessary to continue this 50% plus healthy growth rate, then I think we'd all be very pleased.
We're much more focused on maintaining the high growth because we know the opportunities are out there. It's a matter of stimulating the awareness, lowering the transaction cost to put a cobot in by making them easier to use. That's where our overwhelming focus is. We have really no worries about the operating profitability. Frankly, we're way ahead of where you might expect us to be at 19%, given the growth prospects of this business.
Great. That's helpful. I have a question on the 2021 model. If I did my math correctly, your other test part, your system plus wireless test, you're actually modeling to grow around 4.5% CAGR from 2017 to 2021. What's driving that? Is it market growth, like 5G turning on, or is it new products? Specifically on the new product, I think on a last question regarding system-level tests. My understanding that's more for testing semi on a system application, right? Historically in semi test, especially mobile semi, pretty much each chip maker select ATE and stick with them, and there's a lot of stickiness and very hard to gain share against your competitor because customer qualified. Is that the same kind of dynamic going on on system-level tests?
Okay. I'll take the first part of the question. The growth rate in the other test, you are correct, it's in 5G wireless test. That's the one element of those other businesses where we see credible, meaningful growth. The other businesses are more 2% or 3% GDP type growth. 5G is 2020, 2021, thereabout, and we believe we'll be well-positioned for it. We have an integrated 5G product that is out with a chipset company or two now. We're in a good position, albeit it's very early and 5G would be new to us because we were in 4G, but we didn't have meaningful share because we got to that market late. We don't plan on getting to 5G late. As you might recall, we're the leaders in connectivity, but the cellular side has been more difficult for us to break into.
On your other question about is it sticky with system-level test, I don't see it that way. What we're seeing is that there may be more applications that need system-level tests that didn't have it before. Some of these more advanced semiconductor devices may need an additional level of test to weed out the harder to find faults, and there may not have been any system-level test in the past, so it's a green field. The challenge for us is it a good fit for us? Meaning, can we leverage our existing platform? If we can't and it's a whole new design, it probably isn't worth it to us. That's what Mark was referring to. We're going to look to see if there's other opportunities, but they have to fit with our platform.
Great. That's all I have. Thank you for squeezing two in one on that one.
Operator, we have time for just one more quick question, please.
Thank you. Our final question is from the line of David Duley with Steelhead.
Thanks for slipping me in. I appreciate it. I was wondering on the robot side of things, initially you talked about being able to take that into some of your larger electronics customers, and I think you've been highlighting on conference calls that the average sale is still two or three units. Could you talk about what the average sale is currently per customer on the robots, and if you've had success moving into larger electronic customers?
We have a small number of customers that we call on from other divisions, LitePoint and our production board test group. One is an automotive tier 1 supplier we're in, and another one is an Asian production PCB board supplier, and they do assembly as well. There's two that are over 100 cobots. There tends to be coordination that we're able to help the Universal Robots distribution team get to the right person. There hasn't been a list of 15 accounts that I could point to you and say, "Here's the leverage." Where we've given Universal Robots significant leverage, frankly, is cost down on the material pipeline. We've gotten them several gross margin points. We're able to help them hire people much faster around the world, open up new offices. There's a host of things that we're able to help them move faster.
We help them with engineering quality. We bring some people over there to learn best practices and pick and choose what fits their culture. We're helping them in a number of areas. On the customer facing, there has been some help, there's not a list of 10 or 15 accounts that I have.
Okay. One final thing from me is on the high-performance computing segment, is there any sort of crossover point in complexity or speed where you might be able to insert yourself in a larger way in the GPU market?
Yeah. I think what's happening in the GPU space or the array processing space is right now a lot of diversification. There are a lot of companies that do not traditionally make silicon, developing custom silicon for their applications, whether these are automotive companies or consumer goods companies, web infrastructure companies. I think like we're seeing with memory, a diversification of task-specific, let's say GPUs or array processors. That's what's going on right now. When we get into next year, I think you'll see this array of suppliers doing custom ASICs for their own end markets. That's frankly where we see the ability to participate and where we're focused.
Thank you.
Okay, folks. Thanks so much for joining us. Sorry to be going over just a minute or two. Look forward to talking to you in the days and weeks ahead. Bye-bye.
Thank you.
Ladies and gentlemen, this concludes today's teleconference. You may now disconnect.