Good morning. My name is Ginger, and I will be your conference operator today. At this time, I would like to welcome everyone to the Teradyne Quarter Four 2016 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. Andy Blanchard, you may begin your conference.
Thank you, Ginger. Good morning, everyone, and welcome to our discussion of Teradyne's most recent financial results. I'm joined this morning by our CEO, Mark Jagiela, and our Chief Financial Officer, Greg Beecher. Following our opening remarks, we'll provide details of our performance for the fourth quarter and full year 2016 and our outlook for the first quarter of this year. The press release containing our fourth quarter and full-year 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 involve 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 are made as of today, and 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. Also, between now and our next earnings call, Teradyne will be participating in investor conferences hosted by Goldman Sachs and Susquehanna. Let's get on with the rest of the agenda. First, Mark will comment on our recent results and the market conditions as we enter the first quarter.
Greg will offer more details on our quarterly financial results, along with our guidance for the first quarter. We'll then answer your questions, and this call is scheduled for one hour. Mark?
Thanks, Andy. Good morning, everyone, and thank you for joining us today. I'll be providing a summary of our Q4 and 2016 results and describe the trends we see driving our markets in 2017 and beyond. Greg will provide additional details on our financial results, followed by our outlook for the first quarter of 2017. 2016 was a very good year for Teradyne. We finished the year with our strongest Q4 sales in over 15 years and set a new record for Q4 orders. Our $380 million of sales in Q4 puts our full year sales at $1.75 billion, our third consecutive year with sales over $1.6 billion. The earnings power of our business model was evident as we delivered a non-GAAP EPS of $1.51, up 19% over 2015. We generated $360 million in free cash flow and returned $195 million to shareholders in buybacks and dividends.
We ended the year with $1.6 billion in cash and $460 million in long-term debt. Our fourth quarter order surge follows a similar pattern to 2015's Q4. Once again, customers booked major tooling orders for mobile products early to ensure a smooth, manageable ramp over the spring and summer. Additionally, strength in microcontroller, image sensor, and analog test demand contributed to stronger bookings when compared to Q4 of 2015. Overall, we estimate that the SoC test market in 2016 was about $2.4 billion, up 16% from the $2.1 billion in 2015, driven primarily by strong mobility spending. This mobility demand falls into the complex SoC portion of the market served by our UltraFLEX platform, which set a new annual shipment record. Although our microcontroller and image sensor shipments were down for the full year, bookings for both were up strongly in Q4, a positive sign for 2017.
For 2016, our SoC sales totaled about $1.2 billion, up 15% from 2015. I'll step back to put these results in context when I talk about the industry trends shortly. Turning to memory test, we pegged the 2016 market at about $425 million, down from about $500 million in the prior year. However, our concentration is in flash package test, which remained strong in 2016, resulting in memory test sales of $147 million, up 5% from 2015. Of particular note is the success of our Magnum tester in the growing high-speed interface flash market. The use of UFS and similar high-speed interfaces will continue to expand as devices demand more bandwidth from their growing onboard flash storage. The Magnum is well positioned to ride the flash wave after successful deployments in five of the six major suppliers in 2016.
Combining SoC and memory, the total semi test market was about $2.8 billion, up 11% from $2.5 billion, and we estimate our share grew a point to 48%. Turning now to 2017, we expect the SoC market size to be in the range of $2.2 billion-$2.5 billion, down about $50 million from 2016 at the midpoint. Significantly, this midpoint is up 13% from the last off-cycle odd year of 2015 and up 25% from 2013. The reduced impact of parallel test on the semi test market is set to continue in 2017 as we remain above our trend line forecast of 1% annual SoC market growth. For 2017, we expect mobility, which includes all the devices you'll find in a smartphone, to remain the biggest end market driver. Our Q4 bookings also show an improving outlook for automotive end markets.
In memory test, we expect the market to be in the $450 million-$550 million range, with most of the growth coming from improved DRAM test demand. Universal Robots continues to shine. Q4 sales were $34 million, up over 50% from the year-ago quarter. For the year, sales were up over 60% to $99 million on a calendar year basis. In 2016, we expanded our customer-facing sales and support staff, reinforced the management team, grew the distribution channel, and launched a range of initiatives to extend our ease-of-use advantage and build industry awareness of the power of UR's collaborative robots. For UR, Q1 is typically sequentially down from a strong fourth quarter, and we expect the same this quarter. For the year, we continue to see growth of 50% or greater in this exciting segment.
System Test 2016 sales were $190 million, down $22 million from 2015 on lower spending in storage test. The defense and aerospace and production test units both saw single-digit growth for the year, and we expect similar results in 2017. Due to the lumpy nature of storage test, we expect the entire STG group's 2017 sales to be similar to 2016, ±10%. LitePoint sales for 2016 were $96 million, down nearly 50% in a market that was down by a similar %. As we noted in our last call, we resized our cost to align with lower market size and operated profitably in the second half of 2016. Our R&D efforts are going full steam ahead as we position ourselves for the new wireless standards expected in 2018 and beyond.
For 2017, we expect the market to be similar in size to last year, about $200 million. Overall, 2017 is shaping up to be another strong year for Semi Test and UR, while System Test and LitePoint should be about flat year-over-year. Stepping back, the macro trends that drove the business last year, and which we expect will drive it going forward, are very familiar. Increasing electronics complexity and increases in human scale automation. Growing electronics complexity helps across all segments of our test businesses, but most dramatically in Semiconductor Test. Here, complexity has been increasing from the moment the first transistor was created nearly 70 years ago. What's different now, compared to the 10 years prior to 2014, is that complexity, which drives increasing test time and correlates with tester throughput, is now growing faster than the productivity gains of the test equipment.
The industry is now routinely shipping in high-volume devices with more than 2 billion transistors, often with multiple processing cores and a disparate collection of IP blocks. This complexity is amplified as devices are fabbed with advanced process nodes using finer pitches and using advanced packaging technologies. The consumer performance and cost benefits that come from these technologies require small offsetting test intensity increases to optimize the overall balance of cost and quality. As we've discussed before, tester productivity advances through the use of increased parallel test offset much of this complexity trend up until about 2013. At that point, the diminishing economic benefit of the next doubling of parallelism, combined with the nonlinear time and cost escalation of producing reliable, fine-pitch tester interface boards and probe cards, changed the equation.
Increasingly, the lowest overall cost approach and fastest time to ramp comes from spending a little more in test capital to manage this growth, thus simplifying interface board and probe card design for lower overall cost and better yield. Additionally, test program complexity is reduced, both shortening development time and reducing debug. This slowdown in parallel test is having a positive impact on the size of the SoC market. We noted this trend several years ago, and the market data since continued to support the outlook for modest SoC trend line growth going forward. 2016 was a larger market than 2014, and we expect 2017 will be larger than 2015 and 2013. This growth in device complexity and associated longer test times, especially at the high end of the market for devices used in premium smartphones, was a major driver of our record Q4 Semiconductor Test orders.
Complexity is also important for our wireless test business at LitePoint. Last year and in 2017, we're on the other side of the coin, as wireless standards have not grown significantly more complex since the introduction of 802.11ac Wi-Fi and 4G cellular. We expect that will change with the adoption of new Wi-Fi standards in 2018 and 5G cellular later in the decade. For human scale automation, we are at a point where the fast setup time, safe operation, and low cost of UR cobots are enabling their use across an ever-broadening set of geographies and sectors of the global economy. The combination of lowering production costs and rising product quality continues to be the key attraction for this new class of automation.
For example, we recently highlighted a customer in India that uses UR cobots in the production of artificial lenses for cataract patients in the developing world. Operating in a low-cost region, the customer saw a 15% increase in product output after the addition of cobots to their process, making these lenses both more affordable and higher quality. The growth rate of UR cobot shipments is a testament to this expanding wave of adoption and our ambition to grow it even faster. Increasingly, customer awareness of UR cobots' benefits and capabilities, as well as increasing the channel resources necessary to execute the customer-specific requirements for a given application, has been our focus. At the same time, we continue to innovate to further reduce the time to deploy our cobot. With these investments, we see UR continuing to grow at 50% or more for years to come.
Let me turn to our capital allocation plans for 2017. As you know, we began a balanced approach in 2014 with the initiation of our quarterly dividend. We further added annual share buybacks while pursuing disciplined M&As such as Universal Robots. With much of the cash on our balance sheet being offshore, we issued convertible debt in the fourth quarter to raise additional U.S. cash to continue this balanced approach. In 2017, we plan to buy back a minimum of $200 million of shares and increase our quarterly dividend from $0.06 to $0.07. With that, I'll turn things over to Greg for additional comments and the financial details. Greg?
Thanks, Mark. Good morning, everyone. I'll start with the highlights of 2016, offer some comments on 2017, including our capital allocation plans, provide perspective on the strategic position and market trends at the operating segment level, along with fourth quarter results and first quarter outlook. On the financial highlights front, our $1.75 billion of sales in 2016, up 7% from 2015, and $1.51 of non-GAAP EPS are solid steps towards the $2 EPS plan we outlined in recent calls. In 2016, we accomplished our key objectives. We gained share in SemiTest for the fifth consecutive year with strong memory test performance. We grew Universal Robots' top line over 60% from calendar 2015, we allocated capital in 2016, where we had the highest return through buybacks and dividends, bringing our three-year capital returns to $583 million.
This includes buying back $446 million of our shares at an average price of $19.87 and returning $137 million in dividend payments since the start of 2014. We raised $460 million of seven-year convertible debt with 1.25% coupon rate. Factoring in the call spread overlay, shareholder dilution will not begin to occur until our stock is above $39.95. As we've described in prior calls, an increasing portion of our annual cash generation is offshore, which in 2016 reached 80% of the $380 million of total cash generated. As you can see from our cash resource slide, we have $1.6 billion of total cash resources with the convertible debt proceeds. Raising our U.S. cash balance allows us to continue our capital return programs rather than slowing down to wait for a potential repatriation holiday.
Our offshore cash and marketable securities of $768 million provide additional dry powder for possible foreign acquisitions or to be added to our U.S. balances if a repatriation holiday occurs. We have a new $500 million share repurchase authorization from January 1st, and as noted, we expect to repurchase a minimum of $200 million worth of stock in 2017 and have increased our quarterly dividend 17% to $0.07 a quarter, subject, of course, to ongoing board approval. Our strengthening growth prospects and tight operating model comfortably support this move. We're also maintaining dry powder in the U.S. for attractive M&A adjacencies, principally in industrial automation, where the intersection of enabling technologies and higher labor costs provide new automation opportunities. We'll continue to compare the risk-adjusted returns from possible M&A against buying back even more of our stock.
As I've noted in past calls, I don't need to fill any holes in our portfolio. We will remain very selective. Now I'll do a quick rundown of the segment level highlights and trends. Starting with SemiTest, we had another year of ATE market share gains, bringing us to an estimated 48% share for 2016. Memory test gains provided the lift this year. As described in the past, about half or more of our share gains come from focusing on the healthier segments that have long-term tailwinds versus trying to cover every possible bet. Targeting the new high-speed flash protocols with our Magnum product line and its frequency headroom and outstanding economics drove our memory share to over 30%. We also scored some meaningful DRAM wafer sort business with Magnum.
Across the SemiTest market, the steadily rising complexity trends that we've described are expected to continue driving up test intensity, along with the addition of more stringent quality requirements being placed on our customers from the major brands, further driving up test demand. As Mark noted, another key point to help explain the overall improvement in our core SoC market is that for complex mobility devices, parallel test efficiencies can no longer mask the underlying complexity increases. Also, parallel test productivity naturally diminishes as the biggest productivity gain arises from going from one to two devices rather than, say, four to eight. As in the latter, the system base cost has already been cut in half twice. We're well down the parallel test path, so the impacts of this compressive force on the test market should continue to decline.
The net of all this is, we tried to capture in our slide deck, is the SoC test market has shifted to a more favorable trend line trajectory with the more current even or odd years, such as 2016 and 2015, exceeding their predecessor even or odd years of 2014 and 2013, more than the 1% annual growth rate that we've used for modeling purposes. As you can see from the SoC range Mark noted, we expect 2017 to be a larger market than 2015. Also noteworthy, as of the prior multi-year cycles have been replaced with annual or seasonal cycles, driven by the love of new smartphone functionality and holiday electronics buying. Now shifting to Universal Robots. We grew calendar year sales 62% from $61 million in 2015 to $99 million in 2016.
We, along with third-party forecasters, expect the cobot market to grow from about $100 million in 2015 to $1 billion or more by 2020. The ROI payback with our easy-to-deploy cobots is often about six months and sometimes even faster. UR cobots are designed for easy deployment, high repeatability, and safety, with the target user being the shop floor operator on the manufacturing line. This ease of use enables UR cobots to literally be moved from one task in the morning to another in the afternoon. This, along with a lower initial cost to implement, makes UR cobots a compelling automation solution. As Mark noted, the constraint to even faster cobot growth, apart from general awareness, has been the need for more sales and trained field resources to follow up on the diverse opportunities across the many verticals and geographies.
While we still have work to do, we've made good progress investing in our sales and distribution channel to lessen that constraint and increase the velocity of business through each channel partner. Shifting to System Test. We operate at model profitability in 2016. Mil-Aero Avionics Testing is the bright spot, with smart weapon systems driving fighter upgrades, while in production board tests, we're getting a string of design wins at automotive suppliers with our high throughput tester designed for high panel count applications. In storage test, the business was down about 30% from 2015 levels as customers digested the large capacity additions put in place in the prior year. Turning to wireless tests at LitePoint, we repositioned the business back in the black for the second half of the year.
We expect between $90 million and $100 million of sales in 2017, about flat with 2016, and thereafter expect the next Wi-Fi retooling waves to offer some market lift in 2018. Longer term, 5G cellular is expected to drive a large retooling. Now to the more granular fourth quarter numbers. SemiTest orders were $524 million, a record for the fourth quarter, driven by mobility and automotive test demand. SoC test orders were $501 million, and memory test orders were $23 million. SemiTest service orders were $73 million of the total. SemiTest sales were $271 million in the fourth quarter, with SoC making up $214 million and memory test the balance at $57 million. SemiTest service revenue totaled $59 million. Universal Robots had orders and sales of $34 million.
Geographically, fourth quarter UR sales were 44% in Europe, 27% in North America, 22% in Asia, and 7% in the rest of the world. Our largest channel partner was only 3% of our total 2016 sales. Again, we have very wide diversity. System test orders were $47 million in the quarter, and sales were $50 million. Wireless test orders were $23 million, and sales $26 million in the fourth quarter. At the company level, our fourth quarter sales were $380 million, the non-GAAP operating profit rate was 19%, and non-GAAP EPS was $0.32. We had no 10% customer in the fourth quarter and two 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 $2 million to $148 million compared to the third quarter due to lower spending in SemiTest.
Non-GAAP gross margins were 55% in 2016, an improvement from the levels seen in the last two even years when mobility buying was also especially strong. Non-GAAP operating expenses were up $20 million versus 2015, driven principally by distribution investments in Universal Robots, along with their inclusion for a full year in 2016 versus only a partial year in 2015. In our test businesses, our total 2016 OpEx spending was under 2015, principally with reductions in wireless test. Cash and market securities total $1.6 billion or $8 per share, and 52% is onshore, reflecting the proceeds of the December convert, and 48% is offshore. We bought back $61 million of stock in the fourth quarter at an average price of $24.30 a share. Our convertible debt is recorded on the balance sheet at $353 million, which is net of imputed interest and offering fees.
At its face amount of $460 million, our debt is about one times our 2016 pro forma EBITDA. As planned, our inventory increased $21 million in the fourth quarter as we added materials to maintain attractive lead times. While most of the Semi-Test product orders from Q4 will ship in the first half of 2017, they are weighted towards the second quarter. Shipments or acceptances can be concentrated late in a quarter or early the next quarter and swing our quarterly reported results, but have no impact on our internal operations. Our sales for the first quarter are expected to be between $420 million and $450 million, our highest Q1 guidance range in over 15 years. Non-GAAP EPS range is $0.33 to $0.40 on 202 million diluted shares. 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 57%, about flat with the fourth quarter due to favorable product mix. For the full year, we'd expect gross margins to operate at about 55%, right in the middle of our historical range of 54%-56%. The first quarter OpEx, running at 36%-39% of first quarter sales, is up about $15 million from the fourth quarter due to further investments at Universal Robots, higher variable compensation accruals on higher anticipated profits, and SemiTest engineering NREs concentrated in the first half of the year. Overall, OpEx in our test businesses in 2017 should remain about flat with 2016 levels, while we will continue to make incremental investments in Universal Robots to further strengthen our distribution and ecosystem lead. We'd also expect inventory to step up significantly in the first quarter to support the second quarter shipments.
The non-GAAP operating profit rate at the midpoint of our first quarter guidance is about 20%. Our 2017 tax rate is expected to be about 15%, down from prior guidance due to a higher mix of offshore profits and the favorable impacts from the convertible debt and call spread overlay. Our non-GAAP interest expense, excluding the non-cash imputed interest from the convert, is expected to be de minimis, as the 1.25% annual coupon on the convertible debt should be offset by interest income on our cash balances. In 2017, we've earmarked $80 million-$100 million for CapEx, up slightly at the midpoint from 2016's $85 million. We start 2017 with strong momentum driven by SoC test and Universal Robots. With device complexity and quality standards increasing, we see a healthy long-term demand picture for SemiTest even before considering ongoing and targeted share gains.
Adding Universal Robots' 50%-plus annual growth, healthy performance in system tests, a leaner and more focused wireless test group, we see a very bright future. With that, I'll turn the call back to Andy.
Thanks, Greg. Ginger would now like to take some questions, and as a reminder, please limit yourself to one question and a follow-up.
At this time, if you would like to ask a question, please press star one on your telephone keypad. Again, that is star one. Please limit to one question and one follow-up. Your first question comes from Toshiya Hari from Goldman Sachs.
Hi, guys. Thank you for taking my question. Congrats on the strong results. My first question is on SemiTest, and kind of regarding the full year of 2017. When we take the midpoint of your guide for both the SoC test market and the memory test market, I think you're forecasting overall SemiTest, your TAM to be kind of flattish, if not up a little bit. Given that you've been consistently gaining share in the SemiTest market, is it reasonable to assume you could actually grow SemiTest revenue in 2017, which is something that we haven't seen in odd years in the past?
Yeah, Toshiya. It's certainly possible. A couple of things have happened since our last call. The market has strengthened a bit, which is why we've moved the midpoint of our 2017 guide up a bit. The market share gains pretty much have been consistent year-over-year for many, many years. If you go back over 10 years, we've picked up 20 points of test share. Although we talk about a point, any given year can be flat to up 2%, and we think we can continue to average that one point net gain per year. I wouldn't be surprised if it was up during the year-over-year revenue. The second half, of course, is still something we don't have a lot of visibility into, but the visibility we have to the first half is quite strong.
Great. Thank you. As my follow-up, I had a question on the competitive landscape in industrial automation. Clearly, the market is growing, and you guys are doing an amazing job growing Universal Robots. What are you seeing in terms of competition? What happened to market share in 2016 relative to 2015? What are your thoughts on pricing and margins as we head into 2017? Thank you.
Competition, we expect as this market develops, will continue to grow. I think the good news is that the market terrain is so vast and wide that competition's a little different here than most industrial businesses. The head-to-head competition is less evident, and it's more the competition of mind share and ROI awareness and integrator capability to deliver that to the customer are the bottlenecks, not competition. On the competitive front, we've seen several years ago, the large industrial companies begin to try to develop lower price point robots to move down into this human scale automation space. None of them have yet come close to the capabilities we have with the UR product. On the other end, you have emerging companies, startups Around the globe, China, Taiwan, U.S., other places, that are coming at it from the other angle.
Those companies, copying the hardware may be the most easy thing to do, copying the software, ease of use, the ecosystem, the distribution system is really hard. We continue to see that. We are confident this year on the competitive position. Margins should hang in. Things are still looking pretty good.
Great. Thanks so much.
Your next question is from Mehdi Hosseini from SIG.
Yes, thanks for taking my question. Two follow-ups. Just going back to cobot and given your 2016 performance, it seems like you have around a 50% market share. Should we assume that you should be able, at the minimum, maintain that market share given your market forecast for 2020, market reaching $1 billion of TAM? I have a follow-up.
Mehdi, this is Greg. Yes, that is a fair assumption. It's hard to know exactly what our market share is, but it could be 50, it might be closer to 60%, but it's in that neighborhood. As Mark outlined, we have a big first-mover advantage in our cobot in repeatability, safety, and particularly ease of programming, which is a huge advantage the flexibility provides. In the last year, we've invested considerably in the distribution and ecosystem, so we have many third parties developing their solutions on our platform, and it integrates with an API into our cobot.
Got it
We've developed the distribution channel extensively. When others come out with a capable cobot, and they will, it does take time to get that channel in place and get the channel productive. We feel confident that we can maintain this market share, looking at the next year and the year after.
Got it. Then on system tests, two follow-ups there. How should we think about overall end market trend in 2017? How does the mix change within the system test going to impact? I'm assuming that at some point you're actually going to do more of a SSD test. Is that true or not? If it's true, how would that change the mix of the system test, and would that have a margin impact?
In the short term for this current year we're in now, we think system tests will look similar to what it looked like last year. The storage test business, which is the wild card, because we have a new platform there that we've taken to different places, as you might recall, two and a half, three and a half to SSD. That's a versatile platform that has high automation and asynchronous slots, but it's going to take time to see if we can take that platform to other places. I think it's more of a late this year, maybe even 2018, we'll get a better read, can we expand the market for that platform? In the short term, storage test should be hitting model profits similar to last year.
Got it. Thank you.
Your next question comes from Tom Diffely from D.A. Davidson.
Yes, good morning. Getting back to your mobility strength, I'm trying to figure out how you view the world as far as end market mobility demand versus the change of manufacturing or the change of the foundry customer that's actually building the products. What is the bigger driver right now?
Is the question what's the bigger driver for the test market?
For the tester market, how much of your growth has been driven by that change in the manufacturing location, change of manufacturing customer, versus the actual end market demand driving the tester demand?
I'd say very little has been driven by the change of location. For the past several years, location's been pretty stable, and the growth has come all from increases in complexity and therefore test time.
Okay, good. Then on the DRAM side, you said DRAM was going to drive a lot of the growth in 2017. I'm curious how well you are leveraged to those portions of the DRAM market that you think are growing.
Yes. Our market share is lower in DRAM. To the extent DRAM grows, we won't grow proportional to the total market. We're more concentrated in flash. We do participate in DRAM, but let's say our total share in memory is 30, mid-30s right now. Flash might be in the 40-50 range, and DRAM might be in the 15-20 range.
Okay, thank you.
Your next question comes from Edwin Mok from Needham.
Great, thanks for taking my question. First question I have is on, go back to SoC tests. I think on the prepared remarks, you guys talk about some growth in MCU and image sensor booking for the quarter. Just curious if that's just some big customer come in to place order, or is it more broad-based? I guess, just tied to that also, how are you guys positioned in China right now? We obviously heard a lot of investment on the front end on China that might eventually drive their back-end tests market. Just curious how you're positioned in China, and how do you think that could potentially progress as you go to 2018, 2019, kind of longer-term timeframe.
Okay. On automotive m icrocontroller and image sensor, that is a broad-based strength that we saw in orders in the fourth quarter and continuing into the first quarter. That's good news. The whole ecosystem around automotive is getting increasingly healthy. Automotive tends to be stable overall, and what we're seeing more is a little pickup in the growth rate of automotive electronics. In image sensors, there's a little more diversity now in the, I would call it, the image sensor technology that's going to emerge over the next few years, and that's also driving a spreading around of the supply base a little more and more tests for that. Your other question, I'm sorry?
Regarding China and where you guys position and how do you think the market progress?
Yeah. China is characterized by the indigenous semiconductor companies in China. There are some large, powerful companies there. We've picked up market share last year in China, and we see that as a growing faster than average market for quite some time. It will be highly concentrated around a few customers and a few technologies. I'd say it's a strong position, growing position, and memory is a wild card at the moment. We're in position there with some of the investments going on, but when that might turn into volume production is kind of anybody's guess.
Great. My quick follow on UR. You mentioned that you expect to continue growing investment on UR. Is there a way to kind of think about it? Are you kind of increasing OpEx? The similar growth rate of the top line, less than top line? Any kind of metric or any way we can kind of think about how fast you're growing investment in UR?
Right. We recently increased UR's OpEx from about $7 million to $12 million between 2015 and 2016. Looking at 2017, we could take that $12 million up to about $17 million. We see another step up to support the much higher sales growth rate. We've talked about, in the past, that we were taking Universal Robots' profit rate down to 10% to get further ahead in distribution and ecosystem partners in anticipation of eventually running into more meaningful competition. We're still doing that, and we expect in 2017, we're going to move from about a 10% profit rate towards 15%, but we're more focused on capturing the growth and getting further ahead, which we've been doing more recently.
That sounds great. Thank you.
Your next question is from Farhan Ahmad from Credit Suisse.
Hi. Thanks for taking my question. I had a question in terms of your uptick in SoC market expectations for next year. How much of incremental strength are you seeing for AP versus other markets, and how should we think about the gross margins within SoC tests for next year?
Well, the SoC test margins, by and large, have been consistent for us for many years. We have also, I think, mentioned that in some of the even years when we have the large customer directing more buys because there's a bigger jump in their phone, that can and has pulled our margins down a bit. More recently, we've also described that complexity is helping even in the odd years. This even, odd year phenomena is moderating. We're seeing good demand because of more transistors, more test time in the odd years. There's far less moving around of our gross margin in SemiTest. Of course, in any particular segment, they could be a few points higher or lower, but overall, we are finding ourselves staying within this tight band.
Got it. One question on the long-term trend for the test complexity. If I think back, like long time ago, like in late 1990s, there was a very clear trend that the testing complexity was rising so much faster that the test market was outpacing the semi revenues by a significant amount. Is there any kind of trend that we may even see an acceleration in the complexity going up much faster, or the test times just decelerating a lot more that actually the test market could inflect a lot higher going forward?
So, I don't expect the craziness we saw at the end of the '90s to happen. That, I don't think was a systemic thing around technology as much as driven by a unit growth set of expectations. Right now, complexity growth looks to be on a beat rate that's relatively, I would say, consistent for the foreseeable future, the next, let's say, two years at least, where we have some visibility. I don't think there's a lot of time in the ecosystem of design to deal with that complexity in some of the ways that complexity's been dealt with in, let's say, memory test back 15 years ago, or microprocessor test 20 years ago, which was optimized test structures in the designs to try to reduce the time on the tester. That is an investment in time and money that has marginal returns.
As I said in my prepared remarks, people are really looking at the overall cost to ramp, and a little bit more test intensity simplifies dramatically the tooling and software and debug needed to ramp a part and improves the yield. So I think that's for the foreseeable future, the track people are going to be on at the high complexity end of the market. Got it. Thank you. That's all I have.
Our next question comes from Timothy Arcuri from Cowen and Company.
Thank you. I had two. One is a follow-on to what just got asked. Last year, the big orders in the fourth quarter were driven by InFO, and this year it sounds like it was a little more of a pull-in versus kind of a new technology that's going to get implemented. Is that the right way to characterize it? I guess the reason I'm asking here is because it looks like something structural is happening beyond just the parallelism effect. Because if you look at the buy rate, at least at SSD, it sort of has bounced around between 105-115 bps between 2013, 2014, 2015, 2016. Is there something bigger happening here, such that in your modeling, maybe you're assuming that the buy rate goes to 120 bps, 130 bps, 140 bps? Why can't that not happen? Thanks.
I'll start with that one. The buy rate has stabilized and improved a bit from several years back, so there is good news there. In terms of the bookings coming in earlier, when there's a concentration of sizable demand, it's not uncommon for a customer to put bookings in earlier with us to schedule deliveries over a longer time period so they can be assured that that stream of products is going to come when they need it. Our lead times tend to be very short for ordering one, two, three testers. If it's a sizable amount of testers, likewise, you need to put orders in at a different pace than a small number of testers. It's not much more complicated than that. Just to comment a bit, there's something else more fundamental going on.
We modeled a 1% growth rate in the market after a decade of a minus 3% CAGR. That's been very difficult to model, and we're actually running for the four-year period here ahead of that rate. It's certainly possible that we could get back to that 125, 130 range. I'd say it's a little too early to call. I'd like to see a little more evidence in some of the device segments that are less digitally intensive before we'd be that bold.
Got it. Okay. Thank you. Then I guess my next question is, if you run a screen on some of the proposed changes coming out of D.C., you guys screen pretty well on most of those. One particular side would be if there's going to be a lot of jobs brought back to the U.S., my guess would be that that's not really going to be very people-intensive. It's going to be pretty cobot-intensive, possibly. I'm wondering if you've seen any, sort of if your salespeople and your channels have seen any increase in activity or in inquiry activity because there's the potential for all these jobs to get brought back to the U.S. Thanks.
Tim, there's a lot of activity and interest with using cobots everywhere in the world, whether it's North America or Asia. Most manufacturers know if they don't use this new tool, they will be uncompetitive. In terms of the U.S., we've had some very good demand more recently, and we've built out the sales team and the technical support team, so our coverage is much better like we have done in other parts of the world. We've gotten some sizable orders from some large U.S. companies. We historically tend to get orders of two or three cobots, and then they add to it and add to it. More recently, there's been some sizable, bigger orders. I don't know if that's the beginning of a trend or it's just one situation.
The good news is we're prepared in the U.S. for these changes that may be upon us.
Awesome. Thank you.
Our next question is from C.J. Muse from Evercore.
Yeah, good morning. Thank you for taking my question. I guess first question, as you look at the 57% gross margin for Q1, curious, what is driving that within the mix? Is that higher Eagle Test content and/or should we be thinking about perhaps the new SSD tester coming in at a higher rate or perhaps earlier seasonal uplift from UR? Would love your thoughts around that.
C.J., it's even simpler than that. We've talked about our large customer driving demand through their partners. We don't have much from the large customers scheduled in the first quarter. That's largely after the first quarter. Just like the fourth quarter was, there wasn't a lot from the large customer in terms of the shipments in the fourth quarter or the first quarter, therefore, our margins are higher. It's similar to the phenomena we described years ago. In even odd years, the margins improve when sales are down because the large customer is buying less.
Okay, there isn't any sort of change in terms of better mix from your higher margin businesses helping as well?
Correct. No. The margins in all of our business have generally stayed consistent. LitePoint has very strong margins, but their volume is down, so that's hurt us a bit. It's not significant in the grand scheme of things. I think in the long run, we'll stay in this 54%-56% range, and occasionally we'll deviate out at 57%.
Got you. Okay. Very helpful. I guess as a follow-up, as you look at perhaps this tick-tock going away as the AP becomes more complex, and I'm assuming here we're talking greater capabilities around graphics to support AR/VR. Curious if that creates a greater opportunity for you across other AP makers, particularly as you look perhaps to Korea, and others. Does that create a revenue opportunity for you and/or share gains within the SoC bucket?
I think, first of all, it creates an opportunity for the market to grow and for us to grow proportional with it. Whether it's AR/VR or let's say, processing 4K video from multiple cameras in the phone, other kinds of sensors we're going to see in the next couple of years coming into phones, all of that, let's say, data-intensive processing is driving up AP transistor counts, core counts, and complexity, which dovetails right into the trend we've been talking about here. I think it's all positive.
Excellent. Very helpful. Thank you.
Our next question is from Krish Sankar from Bank of America Merrill Lynch.
Yeah, hi. Thanks for taking my question. I have two of them. The first one is, if you look at your quarterly profile for Semi-Test revenues, typically Q2 is the strongest from a revenue standpoint. Given the fact that the SoC test market is stronger, you're gaining traction in memory, do you think the profile is going to be different this year, or is it going to be the similar profile as in prior years with Q2 being the strongest for revenues for Semi-Test? I had a follow-up.
I think recently, in the first half of the year, we ship 50%-55% of the company's annual revenue. I don't think this year is going to be significantly different from that as best we can tell now. As I mentioned though before, the back end of the year is opaque at the moment for us in general. There are some things in memory that I think will most likely kick into high production in 2018, but there's a possibility that begins to ramp at the back end of this year. You're right, perhaps memory's a little bit of a wild card and upside in the back half. I think in terms of SoC, we should expect to see the same pattern we've seen in past years.
Got it. That's very helpful. As a follow-up, the SoC test market that you described last year and this year, can you tell us how much of that do you think was mobility within that SoC bucket last year, and how much do you think mobility is going to be for this year?
Yeah. I think year-over-year, mobility's flat. Maybe mobility, if in the broadest definition, you roll image sensors into it and other things, is running between $900 to $1 billion of test demand a year.
Got it. Thanks, guys.
Your next question is from Patrick Ho from Stifel Nicolaus.
Thank you very much. Maybe just following up on the question on memory test, Mark. You talked about potential upside as the year progresses. Given your strength on the flash side of things and with a lot of new fabs out there, is that where you're targeting some of that potential upside? Or as you mentioned on your prepared remarks, that the upside will come from the DRAM side?
I think we certainly see DRAM increasing this year compared to last because it was so weak last year. On the flash side, the key thing for us is the rate of deployment of these high-speed interfaces, such as UFS and a few others, into mobile and SSD devices. As that grows, we will grow disproportionately faster than our native share, so we'll pick up share. The deployment of those in high volume is what we're trying to gauge right now. My comments were more specifically to when will UFS and others really kick into high growth.
Great. That's helpful. As a follow-up on the UR side of things, you talked about the expense increase over the last couple of years to establish the foothold with your distribution partners and the channel. Longer term, do you see tactically using the gross margin or your cost of goods line to also protect that share as the marketplace grows? What I mean by that is, as more competition enters, will you be able to use the gross margin line as kind of a tactical maneuver to protect that share?
We don't think that is necessary. We are in the low 50s now, we have material cost-down actions underway leveraging the Teradyne supply line group. There's opportunities to improve that gross margin. Yes, there's a chance that those improvements get used in pricing should we find competition. The good news is we're going to have a buffer. I don't anticipate us deviating from low 50s, because we have this buffer and we have a very large lead. Usually, if you're going to do automation, most companies want to go with somebody who's done that type of automation before, has a recipe for it, has integrators that are trained for it, you find there's a better price because it's been done before. There's a lot of advantages that are accumulating on our side of the ledger.
I don't think we're going to need to use price.
Great. Thank you.
Your next question is from Stephen Chen from UBS.
Good morning. Thanks for taking my question. This is Neil on for Steven. Sorry if this is bad, could you let us know what the non-GAAP operating margin is for Universal Robots in the quarter?
Universal Robots non-GAAP operating margins. For the year, I'll tell you the year. For 2016, they averaged about 10% operating profit margin.
Okay. The way to think about that would be relative to the $7 million increase in OpEx that you were talking about for next year?
Yes.
Okay.
If we increase it seven again and get our growth, the operating profit margins, it depends what growth we get to, and we can meter some of the spending. The seven is if we see high growth, which we believe we will. If we think some of that growth will take a little bit longer, we might not go the full seven. Regardless, we expect Universal Robots to operate within a 10% operating profit rate range, plus working its way towards 15% over time.
One follow-up. You mentioned the strong demand from automotive customers. Can you give us an idea of what % of your SSD business is from automotive, if that's possible? We're seeing some strength in the industrial segment, some recovery there. Can you talk about whether that factors into your outlook for 2017?
Yeah. For Teradyne's business, automotive-related electronics represent roughly, let's say, 15% or so of our total revenue. Sorry, what was your second question?
I was just wondering if you're seeing any strength in industrial segment, if you have exposure there, given some of the strength that you're seeing it.
Yeah, I would say that is relatively stable, flat, not growing at the moment. The real growth has been, mobility looks like in 2017, being an odd year, will be flat to slightly up over last year, automotive and microcontroller will be up. Those are the kind of three pieces Analog, by the way, a little bit. Those are the three pieces that we see growing.
Great. Thanks.
Operator, we have time to sneak in just one more, please.
Okay, your final question will be from David Duley from Steelhead.
Thanks for taking my question. First question is, when you look at the incoming orders in the mobility space, were those concentrated with one customer, or was this broad-based demand?
Well, let's see. Our business in mobility every year for the past three to four years has had a certain degree of high concentration around, let's say, a specific driver. The overall mobility segment is the demand is broad-based. We do see pickups in demand from all the components you might find in a smartphone coming through fourth quarter into first.
Okay. I guess more specifically then on the application processor side, was this just one or two customers, or was this more of a broad-based demand for the more complex testers?
For application processors specifically, that is dominated by one customer. The other customers for applications processors, however, also had a pickup. Just proportionally speaking, a much smaller proportion of our bookings.
Okay, great. You've made an argument about essentially the SoC test market starting to grow again because the parallel test efficiencies are going to decline. Do you see that potentially happening in the memory market? I guess specifically my question is, it seems like there's pretty rapid bit growth in flash and 3D NAND, I'm kind of curious why, or when the market will grow at a higher rate, to match the bit growth in the market.
The bit rates have been growing rapidly in memory for decades, the thing that's been different about memory is the complexity there is not really complexity. A lot of what has happened is lithography shrinks and 3D vertical structures to add more capacity to the part, but the complexity of those devices isn't necessarily growing. They've been able to, because of the regular architecture of memories, do things with the design to optimize test and reduce test intensity. The thing there that could change this in terms of complexity change in memory are these high-speed interfaces. The high-speed interfaces are quite sophisticated and do require more extensive testing and obsolete existing testers.
If both on the, primarily on the flash side, I would say, as flash devices, whether it's vertical structures or bit density is less important to us than a step function change in the IO bandwidth, which is why I keep speaking about UFS and other types of interfaces. That complexity will drive retooling and test time increases. It could bump the memory test intensity over the next four to five years.
Okay, great. Just as a housekeeping, can you give me an estimate as what you think the size of the image sensor test market is?
That market varies year to year quite a bit, but let's say an average number to use might be $100 million.
Okay. Thank you so much. Nice quarter.
Okay, folks, thanks so much for joining us. We look forward to talking with you in the days ahead. Those remaining in the queue, I'll be back to you within the next 90 minutes or so. Thank you.
Thank you.
Thanks.
This does conclude today's conference call. Thank you for participating. At this time, you may now disconnect.