Illumina Inc (ILMN) 2016 Q3 法說會逐字稿

Welcome to the quarter three 2016 Illumina Inc. earnings conference call. My name is Katie, and I will be your operator for today's call.

(Operator Instructions)

Please note that this conference is being recorded. I will now turn the call over to Rebecca Chambers. Please go ahead.

Thanks, Katie. Good afternoon everyone, and welcome to our earnings call for the third quarter of FY16. During the call today, we will review the financial results released after the close of the market, and offer commentary on our commercial activity, after which we will host a question-and-answer session. If you have not had a chance to review the earnings release and earnings presentation, both can be found in the investor relations section of our website at Illumina.com.

Participating for Illumina today will be Francis deSouza, President and Chief Executive Officer, and Marc Stapley, Executive Vice President, Chief Administrative Officer, and Chief Financial Officer. Francis will provide a brief update on the state of our business, and Marc will review our financial results. This call is being recorded, and the audio portion will be archived in the investors section of our website.

It is our intent that all forward-looking statements regarding our expected financial results and commercial activity made during today's call will be protected under the Private Securities Litigation Reform Act of 1995. Forward-looking statements are subject to risks and uncertainties. Actual events or results may differ materially from those projected or discussed.

All forward-looking statements are based upon current available information, and Illumina assumes no obligation to update these statements. To better understand the risks and uncertainties that could cause actual results to differ, we refer you to the documents that Illumina files with the Securities and Exchange Commission, including Illumina's most recent Forms 10-Q and 10-K. With that I will now turn the call over to Francis.

Thanks, Rebecca and good afternoon everyone. As we shared a few weeks ago, our third-quarter revenue was $607 million, a 10% increase over Q3 of last year, and we expect Q4 revenue to be flat to slightly up sequentially. This is lower than the outlook we provided in July.

Since the close of the quarter, we have analyzed market and funding dynamics, capacity utilization, competition, and outsourcing trends. Based on these reviews, we do not believe that excess market capacity, outsourcing or competition materially affected our results or outlook. Utilization of our high throughput installed base continues to trend upward, normalizing for the large annual stocking orders received in Q3, both this year and last. Additionally, purchasing patterns at customers who have only HiSeq 2500s have been stable, supporting the view that these labs are not increasing their outsourcing to service providers. Lastly, from a competition standpoint, while we do compete for the allocation of funds, our win rates remain steady, and we do not believe this has been a significant factor.

We have identified the factors we believe drove our Q3 miss, and lower fourth-quarter revenue guidance. First, in the second half, we have been tracking opportunities for more than 90 HiSeq X instruments between funnel and backlog, and previously forecasted that more than 60 would ship. Due to a shift in certain US Government institutions' funding practices, and two customer-specific situations, approximately 35 of these instruments have moved out of the second half.

For the academic and government opportunities, projects of these institutions were previously awarded in three to five-year commitments, enabling customers to purchase instruments upon award. This year, smaller portions of these projects were awarded, with shorter or uncertain time frames, affecting our customers' ability to make capital commitments. Ultimately, we do not believe that this change will affect the size of these programs, which plan to sequence more than 100,000 genomes in the coming years.

Based on these programs, as well as other personalized medicine projects, including France's plans to establish population scale sequencing operations and the Cancer Moon Shot Initiative, we believe that sequencing sample growth remains robust. Furthermore, as we shared in the pre-announcement, one HiSeq X system that was expected to close in Q3 did not materialize, due to capital constraints specific to that customer.

Additionally, we removed a HiSeq X10 associated with a nation-scale project that was originally anticipated in Q4, and has since pushed to 2017, due to lab readiness. The other factor that contributed to the second-half shortfall was lower than anticipated HiSeq 2500 and 4000 orders, which we believe was driven by legacy HiSeq customers favoring the HiSeq X and NextSeq platforms.

The introduction of HiSeq X in January 2014 enabled whole genome sequencing to be performed much more economically, and as a result, samples have shifted to whole genome sequencing at the expense of other applications. Whole genome sequencing on HiSeq X now represents approximately 15% of all high throughput runs compared to 2% just two years ago.

Additionally, the release of NextSeq's v2 reagents in 2015 brought the quality on par with HiSeq. As a result, some high throughput customers have been adopting NextSeq, given its flexible work flow, which enables batching fewer samples and attractive operating costs. As a result, we will not see the second half uptick in high throughput instrument place placements we had previously expected.

To better identify trends like this earlier, we have initiated a global forecast improvement project, which I have asked Marc to lead, that will enhance both our visibility and forecast accuracy. Over time we are relatively indifferent to these shifts, as in both of these cases, the lower operating costs of HiSeq X and NextSeq enables our customers to sequence more samples. However, the timing of the elasticity taking effect, or the customer completing their transition varies across customers, and can create a temporary dip in spending.

Given this, HiSeq consumable utilization was slightly below our guidance range of $300,000 to $350,000, driven primarily by a few customers shifting to other platforms in the Americas, including the HiSeq 4000, and not a decline in utilization across the broader installed base. These new platform implementations take time, sometimes several quarters, and are expected to impact utilization in Q4.

Moving now to rest of our portfolio, benchtop instruments performed well in the quarter. NextSeq benefited from commercial customers scaling capacity, including NIPT customers in the US and China, and saw continued strength in consumable utilization, with pull-through at the top end of the guidance range of $100,000 to $150,000. Additionally, both instrument and reagent sales of MiSeq and MiniSeq performed in line with our expectations.

Microarrays grew more than 35% year over year, driven by another quarter of strong direct to consumer demand, as well as robust interest from agriculture customers, which increased 20%. Array orders grew more than 90% versus the third quarter of last year, as we received orders for another 2 million samples for our global screening array, or GSA, during the quarter, bringing the total to over 5 million samples.

Additionally, Infinium XD gained traction in Q3, and as this product continues to contribute to order growth. Both of these new products are now shipping, and have backlogs that extend over the next few years, which has led to an improved multi-year outlook for arrays.

I will now move to an update on a few of our key markets. Oncology testing shipments grew 24% versus the prior year, as our broad enablement strategy continues to drive the penetration of the oncology markets. Importantly, the research use only launch of our you TruSight Tumor 170 panel remains on our expected schedule for later this quarter. Additionally, we have submitted our companion diagnostic test for Vectibix, developed through our partnership with Amgen to the FDA for pre-market approval.

In NIPT, we continue to see customers shifting to testing in-house in our sequencers, and as a result, we performed about 20,000 samples through our testing facility. Over the last four quarters we have transitioned two large customers, and going forward, expect more stable TSO volumes, as well as continued growth in sequencing instruments and consumable sales. Our multi-year outlook for these markets remains optimistic, driven by oncology reimbursement and regulation playing out over time, and our ability to penetrate the average risk and global NIPT opportunities.

I would like to take a moment to discuss the announcement we filed today that Christian Henry will be leaving the Company at the end of January. As many of you know, Christian has been an integral part of our team since joining Illumina as Chief Financial Officer in 2005. He has held many different leadership roles, including general manager of the Genomics Solutions and Life Sciences Business, as well as most recently our Chief Commercial Officer. I would like to thank Christian for his many contributions over the years and wish him and his family the best, as they embark on their next chapter.

Effective immediately, I have asked Mark Van Oene, currently the General Manager of our Americas business, to step in as interim Chief Commercial Officer, while we conduct a search. Mark has been a member of the Illumina team since 2006, and has successfully led our Americas commercial team to remarkable growth. I am confident that with Mark's help, supported by the strong commercial team that Christian has built, we are in good hands during this transition phase.

In closing, the current product life cycle dynamics may take several quarters to resolve, but they are temporal in nature. We remain in the beginning stages of the large and untapped genomics market, and remain committed to introducing innovative technology over the coming years. I will now turn the call over to Marc, who will provide a detailed overview of our third-quarter results.

Thanks, Francis. As Francis mentioned, total revenue grew 10% year over year to $607 million, consistent with our preliminary estimate on October 10. Sequencing consumables and strong demand for microarrays drove the growth, offset by a decline in sequencing instruments.

Geographically, revenue in the Americas grew 2% versus the prior year, due to lower than expected HiSeq 2500 and 4000 instruments, and HiSeq 2500 reagent shipments, as well as a challenging HiSeq X comparison. Asia-Pacific revenue increased 35% versus Q3 2015, as China grew more than 85%. European revenue grew 16%, a slight beat to forecast, and we believe that the region's positive progress on execution will continue over the coming quarters.

Sequencing instrument revenue declined 26% year over year to $106 million. While the year-over-year drop was predicted due to the challenging HiSeq X comparison, this result was lower than our projection, given the miss in high throughput instruments, as Francis discussed.

Consumable revenue represented 65% of total revenue to equal $396 million, an increase of 23%, compared to the third quarter of 2015. Sequencing consumable revenue grew 24% year over year to $333 million, driven by our growing installed base of instruments, and particular strength in HiSeq X and NextSeq consumables. Please note that we have removed approximately 64 HiSeq instruments from our ending Q3 installed base, to reflect the units taken offline, due to the adoption of newer machines.

Services and other revenue grew approximately 18% versus Q3 2015 to $93 million, led by strength in genotyping services and sequencing instrument maintenance contracts associated with the larger installed base, partially offset by an expected decline in NIPT service revenue, given customer migrations to in-house testing.

Turning now to gross margin and operating expenses, I will highlight our adjusted non-GAAP results which exclude non-cash stock compensation expense and other items. I encourage you to review the GAAP reconciliation of non-GAAP measures, as well as our GAAP EPS guidance, which can be found in today's earnings release and presentation. Please note that all subsequent references to net income and earnings per share refer to the results attributable to Illumina stockholders.

Our adjusted gross margin for the third quarter was 72.5%, a sequential decrease of 30 basis points. Year over year, gross margin decreased 70 basis points given investment in sequencing consumable and array manufacturing capacity, as well as our clinical capabilities, which was partially offset by favorable consumable mix.

Adjusted research and development expenses for the quarter were $114 million, or 18.8% of revenue, including $14 million or 2.4% attributable to GRAIL and Helix. This compares to $114 million, or 19% of revenue in the second quarter. Adjusted SG&A expenses for the quarter were $117 million or 19.3% of revenue, including $9 million or 1.5% attributable to GRAIL and Helix. This compares to $127 million or 21.2% of revenue in Q2.

Given our updated revenue and earnings forecast for the year, we booked a $16 million benefit in Q3, primarily to operating expense, resulting from adjustments to our annual variable compensation plan accrual. Additionally, we removed $6 million of compensation expense from our Q4 forecast. Neither of these adjustments were anticipated in our previous forecast or guidance.

For 2017 planning purposes, investors should contemplate an incremental $0.22 EPS headwind, as the accrual returns to normal levels for the full year. Adjusted operating margins were 34.4%, compared to 32.6% in the second quarter, higher primarily due to the variable compensation accrual benefit, as well as lower legal expenses. Operating margin was lower compared to the 36% reported in the third quarter of last year, due to increased investments in headcount, GRAIL and Helix. Excluding GRAIL and Helix, operating margin was 38.2%.

Stock based compensation expense equaled $35 million, up sequentially from $32 million. We reported third quarter GAAP net income of $129 million, and EPS of $0.87 per diluted share, compared to net income of $118 million or $0.79 per diluted share in the prior-year period. Our non-GAAP net income was $144 million, which led to $0.97 in earnings per diluted share, which includes an $0.08 benefit from the compensation adjustment mentioned previously.

Non-GAAP EPS this quarter included approximately $0.05 and $0.02 of dilution from GRAIL and Helix respectively. This quarterly result compares to non-GAAP net income and EPS of $120 million and $0.80 in the third quarter of 2015. Cash flow from operations equaled $150 million, reduced by 100% of the GRAIL and Helix cash burn of $20 million this quarter.

Q3 DSO totaled 57 days, up slightly compared to 56 days last quarter, but still healthy given the unfavorable linearity we experienced. Capital expenditures in Q3 were $57 million, and we reported an additional $84 million increase in property and equipment recorded under build-to-suit lease accounting, where such expenses were paid for by the landlord. Consequently, Q3 free cash flow was $93 million.

We ended the quarter with approximately $1.5 billion in cash and short-term investments, including the consolidated cash balances of GRAIL and Helix. During the quarter, our Board of Directors authorized a new $250 million stock repurchase program, under which we repurchased $13 million of common stock, and have $237 million of authorization remaining.

Turning now to expectations for the remainder of 2016. We expect fourth-quarter revenue to be flat to slightly up sequentially. To be more precise, slightly up should be considered an increase of up to $5 million. This guidance no longer reflects our previously-expected uptick in high throughput sequencing instruments.

For FY16, we are now projecting GAAP earnings per diluted share attributable to Illumina stockholders of $2.92 to $2.97, and non-GAAP earnings per diluted share of $3.27 to $3.32. Please refer to our investor deck for details as to the drivers of our updated guidance.

We have updated our Grail dilution for the year to be approximately $0.30, primarily as a result of the weight of the common shares issued by GRAIL in the first half, compared to the reduced share count in the second half, driving an inflated full-year share calculation. Our share of the GRAIL losses on a quarterly basis including in Q3, continues to be at approximately 50%, as communicated on our last earnings call. Helix dilution is expected to be $0.10.

In closing, I would like to spend a minute on the forecast process improvement project that Francis mentioned. We have already started the initial phase, which is expected to run until mid-December, and we'll identify key opportunities for improvement, including any immediate changes that we can make to enhance our visibility.

We will fold the good work already underway in Europe into this process, and identify areas for improving global consistency, taking the best of the best practices from each region, and adding best-in-class practices where needed. This will inevitably lead to a second stage of the project next year, which we anticipate will incorporate longer term tool and process implementations. I look forward to updating you on our progress periodically.

Thank you for your time. We will now move to the Q&A session.

(Caller Instructions)

Operator, we will now open the lines.

(Operator Instructions)

And our first question comes from Tycho Peterson from JPMorgan. Please go ahead.

I'm going to try to ask two in one here. The first part is, with product cycles lengthening have your assumptions around where utilization needs to go to drive another incremental purchase changed at all? Could you elaborate a little more on the HiSeq consumable miss? That's one of the things we've got on a lot of questions on, post your pre-announcement.

Sure. So let me start by saying, look, it has been a couple years since we released the last big instrument launch, but in general, I wouldn't say the plan is for instrument life cycles to get longer. They just are ready when they're ready.

Now, that is independent from expectations around utilization, which is driven much more around the capacity in the market, as well as the right fit of sequencers for a particular application. So the two are fairly independent from our perspective. And what was the second part of the question?

More color on HiSeq consumable miss. You ruled out a number of things. There's certainly some lingering questions as to whether this is a blip, or going to be more persistent.

We spent, as I said in the script, a bunch of time analyzing what was a driver of this miss. One of the questions we had internally was, is there an overcapacity in the market, and we looked at the utilization of those instruments. One of the analyses we did was look at customers who had just HiSeqs, and the 2500s, and look at what their utilization patterns were doing, and those patterns continue to be stable. So, we're not seeing a change in the utilization of the HiSeqs in the market, at all. So what we are seeing is, as people are making choices around which high throughput instrument to go for, historically where they would have gone for a HiSeq, we're seeing some customers now go to an X for their whole genome work, and in some cases to a NextSeq.

Tycho, I would say we talked about some customer transitions as well, and for certain customers, those transitions from one platform to the other that Francis mentioned does take time. It's not dissimilar from us implementing a new system, our customers implementing a new platform. That can cause a temporal dip. We saw that in Q3, with certain customers. I believe that could continue into Q4.

Go ahead.

I was just going to add but across -- that is a select handful of customers. It isn't across the broad HiSeq installed base. To Francis' point, the broad HiSeq installed base is actually quite stable. If you adjusted for that, that would be very much noticeable in the number.

If I could just ask one more, you came out after the first quarter and talked about the fact that you still thought the market growth was in the mid-teens, and obviously you were growing a little below that. Is that still your view, and do you see a path back to mid-teens growth longer term?

I think Tycho that's something that we'll probably update at some point in the future, our long-term revenue expectations. We're not prepared to give any update on that at this time.

All right. Figured I'd give it a shot. Thanks.

And our next question comes from Doug Schenkel from Cowen. Please go ahead.

So through what I think is fairly fair to describe as what's been a pretty poor period of execution relative to Company-provided financial targets, in a period which has lasted now seven quarters, sequencing consumable revenue growth has ranged between 24% year over year and 45% year-over-year. This, in a period where total revenue growth has dropped as low as 6% year over year.

So, a two-part question. One, are you seeing anything that leads you to question the sustainability of sequencing consumable revenue growth at robust double-digit levels? Based on your prepared remarks, it seems like while consumable revenue growth may be lumpy from line to line in the model, that you have s a lot of conviction in the outlook for sustained consumable revenue momentum.

And then the second part is, related to that, recognizing that in spite of the fact that consumables revenue growth has been robust, and that you are the world's premier sequencing company, you keep missing numbers. So with that in mind, are there business model changes that need to be made, to make you less susceptible to capital cycles? Thank you.

I think those are all fair observations, Doug. I'd start by saying that the consumable growth, as you pointed out, has been reliable and robust over the last many quarters. And that, for us, is a good indicator of the level of the sequencing activity happening in our customer base.

The volatility that we have seen has been primarily driven on the instrument side over the last few quarters. We talked about some of the dynamics playing out in the Americas in Q3, where in some cases there were changes in funding practices, that caused some customers to not spring for the capital purchase, but still look to do sequencing. In other cases, we talked about dynamics where customers were purchasing either Xs or going to NextSeqs, instead of 4000.

So a lot of the volatility has been around the instrument dynamics, and we've seen fairly strong growth in consumables throughout that whole period. We expect that to continue. There's no reason for us to believe that will change in the future.

The only caveat I would add to that is, of course, the large stocking orders that we talked about in Q3 this year, and we had last year. Those do have a potential effect on the comparisons going forward.

Obviously it's a function of the installed base growth of the prior period.

And then Doug, on the -- you mentioned about us missing on the execution. I think the most important thing right now that we're focused on, one of the most important things is that visibility improvement that we need to get. Clearly, we've been surprised by some of the -- we expected for example a big uptick in Q4 in instruments, and as Francis just mentioned, that's no longer the case. So we need to work pretty hard to understand how to -- what's going there and how to improve that going forward and it's going to take a little while. I think there's some immediate things that we can do and have already started to do, and then we'll need some systems and tools and process improvements that we'll implement next year.

I think look, the last few quarters have definitely highlighted for us areas where we need to strengthen our execution capability. So we identified areas in Europe earlier in the year, where we thought we could do a lot better from an execution perspective. We have made a number of changes there, and we're pleased with how those changes are playing out in Europe.

We've identified other areas, as Marc talked about, where we can do a better job in terms of getting ahead of some of the trends that are playing out in our customer base, and improve our own visibility and reporting. And so, the project that Marc is driving for us will be key to that.

And guys, I don't know if you can still hear me but if you can, really the core of the second part of the question was just getting at -- I think what's been a frustration for everybody involved, investors as well as you, you're doing well on the consumables side. Sequencing demand is still robust, yet the stock moves around a lot based on what happens, or what doesn't happen, with capital equipment placements. You just described what you can do to improve visibility and execution, but is there anything more that you would contemplate, that could make this story less about how many instruments you're placing every quarter, and more about the sustained and robust consumables growth that we've seen for a while, and expect to see moving forward, based on what you just described?

Sure. We are looking at a number of things we can do, some of which we already have in place, that we will be talking more actively to our customers about. Things like reagent rental opportunities to get access to our instruments, also talking actively about our leasing programs. And so certainly as we look forward, we expect those to be more important part of the conversations we have. And to your point, the more we get customers into those models, the more predictable the revenue flow will be.

Great. Thank you.

And our next question comes from Jon Groberg from UBS. Please go ahead.

Thanks. Maybe just to piggyback off of that, maybe I got some of these numbers wrong, but I think you said you had a funnel of 90 HiSeqs Xs going into this year, 35 moved out of the second half into 2017. Can you maybe talk a little bit about -- one of the big concerns that we hear from investors is HiSeq utilization sample growth continues at a nice pace.

As you move into 2017, let's say that those instrument purchases may be picked back up, but you have a really difficult compare on the HiSeq X side. So can you maybe talk a little about the funnel that you're seeing for Xs still? And just to follow up on thinking about 2017, Marc, you mentioned you need to remember $0.22 of incremental potential dilution as you move into 2017. Anything else we should be particularly aware of, for 2017? Thanks.

So let's talk about the demand for the Xs. So as we look forward, we continue to expect to see orders in the range of 20 to 30 instruments a quarter. As we looked at the second half of the year, when we were sitting at the midpoint of the year, we were tracking opportunities that added up to over 90 X units.

With that, we gave the guidance we did. Now, driven by some of the things I talked about in terms of the change in the funding practices, in certain US Government agencies, those weren't realized, and we had a couple of X opportunities move out of Q3.

And then one population sequencing opportunity move out of Q4 into 2017. And so, there have been specific dynamics that have caused movements or specific opportunities, but the overall demand for X continues to stay stable in that 20 to 30 units a quarter range, and we expect that to continue going into next year.

Jon, just to clarify, the 90 was our funnel of opportunities we were tracking for the second half of this year. So that doesn't necessarily include the parts of the funnel that are already existing for 2017 opportunities.

And the changing government funding practices doesn't necessarily mean those do occur in 2017, for that component of the 35.

That's right. Jon, on your question about headwinds and tailwinds for 2017, we're still in the middle of building our 2017 budget right now. Just from a directional standpoint, you've got the $0.22. That's right.

I'd also couple other factors that will clearly affect 2017 is GRAIL and Helix, and where those end up from a budgetary standpoint. Plus both of them will have a full year versus just the partial year ramp that they've had this year. Those are still being worked out at this moment. So those are probably the ones I would call out at this time. Everything else will be largely business as usual.

Thanks.

Thank you. And our next question comes from Derik de Bruin from Bank of America-Merrill Lynch. Please go ahead.

Can we talk a little about your oncology? You said sales were up, your shipments were up in oncology 24%. Could you look at this in terms of boxes and instruments and consumables, and parse that out?

The question I'm trying to get to here is like, we've heard from a few vendors that the reimbursement environment has been a little tough, and so they've pulled back a little bit on their panel sequencing. I'm just wondering if could you just clarify what you're seeing in the oncology market, is there any signs of weakness there? Is reimbursement creating a headwind in that segment?

No, we're not seeing any of those play out in our oncology business. We said it grew 24% year on year, which is -- it moves up and down from quarter to quarter, but given the relatively small revenue base as a percentage of our total. But we continue to see growth there.

We continue to see growth in panel use, across our customer base. There are some encouraging signs, in terms of reimbursement that are starting to show up, and all that is driving the growth that we are seeing. So we're not seeing the headwinds that you talked about.

And could you just -- any color on like the boxes versus consumables split?

We don't have that level of granularity to share at this point in time.

And then just one other question, one follow-up. So on the -- you talked about the HiSeq 2500 reagent usage being stable. I assume that a lot of that, that gets done on the HiSeq 2500 is mostly RNA-Seq applications, or a large chunk of that.

I'm just curious, could you talk about that utilization on the 2500, could you talk about what the split is on RNA-Seq, versus exome versus genome? The question I'm wondering is RNA -- are your RNA customers accelerating, and this might be potentially hiding the fact that you might be seeing some exomes and whole genome moving elsewhere? I'm just curious about the mix in terms of 2500.

RNA is an important application for us. It's done on the HiSeq, but really it is mostly a NextSeq -- it's probably most important on the NextSeq. HiSeq tends to be a pretty workhorse instrument for us. We do see a lot of exomes, and we see some genomes on that instrument, so it's not driven by RNA.

And Derik, just to clarify, so the HiSeq customers who only run HiSeq as their high throughput platform choice, they have been, as we mentioned in the script, stable here. Where we have seen a decline in HiSeq consumables, it's due to those transitionary points we made earlier, customers transitioning to other platforms, who in many cases, you see a reduction in their HiSeq consumables and an increase in either NextSeq or HiSeq X. And in some cases, when I said earlier, that transition takes time, you do see a temporal dip in the total spending by those consumers on consumables. So I don't think anything is being masked here by any other trends, I think it's as simple as those dynamics driving it.

Derik, we are seeing more genomes being sequenced. Whereas a couple of years ago there was only 2% of X runs that accounted for the capacity, today, that has increased to mid-teens. So, we are seeing more whole genomes be sequenced as a proportional percentage of samples that are being run. If that's what you're referring to, we would agree with that, but in general, it's not --

I'm more curious about the 2500 utilization. I know people are not doing a lot of whole genome, other than on the rapid run mode on the 2500. It's more about just -- it's more about a cannibalization question on that. I'll take it offline. I'll follow it up with you later.

Okay.

And our next question comes from Dan Arias from Citigroup. Please go ahead.

Francis, can you just comment on the HiSeq upgrade cycle in general? What percentage of the 2500s have been turned over at this point, and what at this point is a reasonable expectation, just in terms of the portion of the installed base that might eventually convert?

Sure. We actually don't provide those specific numbers around the percentage of the installed base that's upgraded. So I can't break it down there. I can say that the upgrades, or what was driving a lot of the demand for the 3000 and the 4000, but the 2500s have been remarkably sticky, for people who have liked the rapid run mode there, people have had validated work flows that they've been running on the 2500, which has caused them to continue to be good workhorse instruments in production. So we don't really break down any more specifics than that.

Okay. If I could maybe ask a follow-up, on your comment on the favorable economics of the X system. What is your sense at this point for how much of the next wave of sequencing work, or demand, might be tied to the next step-down in the costs of sequencing? And then along those lines, just given that it does look like cost is a factor, I'm curious why you think customers might be seeking out the lower price point that the X offers, but maybe not looking for the whole genome cost advantage of the 4000 versus NextSeq, when it comes down to buy a box.

We continue to hear from our research customers, especially, that there are lots of projects that they would like to take on, and they will do that as the price of sequencing comes down. And so whether it's much deeper sequencing they want to do in oncology research, or more single cell research, there's a whole set of projects that we hear about from customers, that they don't feel they can take on at this price point. They are looking forward to the price points continuing to go down.

So I continue to deeply believe in the elasticity of the demand in the research customers. The reality is, we understand so little of the functionality of the genome in almost any area you can think of, that I believe that there's a lot more sequencing to be done, and a lot of it will be enabled through lower price points. So that's one point.

I think that what we are seeing is customers are making the decision to do whole genome, have the potential to look at both the 4000 and the X, and the economics associated with the X are still significantly better than the 4000. And so that's why -- that's what driving the decision I talked about in my remarks, where if they have any volume around whole genome, or they have access to a whole genome sequencing service, those economics can be very compelling for some customers.

Okay. Thanks.

And our next question comes from Ross Muken from Evercore ISI. Please go ahead.

So as we think about some of the markets over time that you have highlighted, or applications, and the pacing at which penetration has occurred in each of them, whether it be in parts of clinical or applied, et cetera. Where do you think maybe something like NIPT, where we've maybe gotten to a grander scale than we had envisioned 12, 24, 36 months ago, where do you think, what other markets have we seen better penetration than one would have expected? And then, where have you seen things generally move slowly, and how does it make you think about the pacing of adoption in terms of time, versus the ultimate size of the opportunity, which I'm guessing in many of these cases like in oncology, et cetera, probably hasn't changed.

It's a great question. So I'll start with where you started, with NIPT. And I'll agree with you that clearly the pace of that market has not lived up to our expectations. And that if you look at where we are, as we come to the end of this year, we are around 30% of covered lives in the US have access to NIPT, testing and we expected that will be further along than it is.

That's still coming, and so we expect that will play out in 2017 and beyond, but we did expect to be further along than we are today. There are areas that did perform a lot better. So for example, the growth that we're seeing in China has been ahead of where we expect it to be.

Some of it is driven by the Chinese PMI and customers buying as part of the Chinese PMI. Some of it has been customers buying in anticipation of being a service provider to the Chinese PMI projects. But a lot of it is driven by adoption, clinical adoption in the Chinese market, and that has been further ahead than where we'd expected.

China is already now our number-two market, country market in the world. And so that's been an area that has moved faster than we expected.

The whole genome sequencing market has moved further and faster than we'd expected. We've touched on in the remarks that today whole genome sequencing represents 15% of the sequencing done in high throughput instruments. That's up from 2% just two years ago, and that has vastly exceeded our expectations.

We're starting to hear from a lot of customers on the research side certainly, that they do want to move to whole genome sequencing, and that they see that as a very evolutionary, natural next step for them in their research, to get a better handle on the biology involved. That's been another area that I'd say has gone better than we'd expected.

Great. And I apologize, Marc, before I ask the question, because I'm going to make you repeat yourself again. I have a lot of questions in the inbox on the Q4 and into 2017 cadence of the benefit we saw this quarter, the reversal next quarter, and then I think through that accrual dilution. And then tie that in with the GRAIL dilution in second half, just so, again, maybe through repetition, some of us can actually walk through this a little bit easier.

Yes, so just to clarify, there's no problem going through it again. The benefit that we saw this quarter, I already mentioned, in terms of the dollar effect of that. And then, the same for next quarter.

So let's recap those. About a $13 million benefit this quarter, and a $6 million benefit next quarter.

$13 million to OpEx.

To OpEx. I'm sorry. So it's a total of $19 million. I think when you look at that relative to the second half guidance, it's about an $0.08 -- sorry, an $0.11 benefit for the second half. The $0.22 headwind for next year contemplates 100%, going back to 100%.

So bear in mind, what you're seeing as the benefit in the second half of this year, takes us from where we were coming into Q3, to where we expect to be coming out of Q4. So it's only the second-half effect, whereas the $0.22 is a full year impact. Does that make sense?

Yes. And then the GRAIL, the incremental GRAIL dilution from the shares being higher?

Yes, so the way that works is on a quarterly basis, when you look at quarterly EPS, you take into account our ownership of GRAIL. And in the first two quarters of the year, it was around 90%, and in the second two quarters, it's around 50%. And that's what we communicated we had accomplished last quarter.

But when you do either a year-to-date or a full-year basis, you take the weighted average of the shares over the number of quarters included. So on a full year, it's over four quarters. On a year to date, it's over three quarters. And of course in the first -- the way we did that particular transaction, in the first two quarters, there were more shares in that calculation than there are in the second two quarters.

When do you that weighted average effect over four quarters, for a full year, it equates to more like 80%, or even above 80% of the losses that we would record, instead of the 50% that we're now down to for the quarter. What that means, if you take it to its natural conclusion, is the four quarters don't sum to the full-year EPS. There's a chart in our investor deck that clearly shows that happens. That's just the way the accounting works.

Got it. That helps. Thank you.

And our next question comes from Amanda Murphy from William Blair. Please go ahead.

Just had a couple questions here, on some commentary you've made. First on the NextSeq, you mentioned that there's work flow benefits to that platform. Could you just give us a little more detail around what types of applications you're seeing shifting to the NextSeq? Is this something that's been going on for some period of time, and is there anything that you can leverage from that platform into other platforms, to drive increased performance?

Yes. So one application that is starting to -- we're getting a lot of traction on NextSeq on, is NIPT. We see a lot of our customers that are running LVTs move to the NextSeq. What they like is the flexible work flows. They like the fact they don't have to batch as many samples as they would have to, with the HiSeq.

And that's a phenomenon that's been building. And so it's been -- it's didn't show up suddenly this quarter. It's been building over time.

I think it was important, the v2 chemistry was an important step for us with the NextSeq, where there was an initial period, where people were looking to see the quality of the runs on NextSeq compared to the other instruments, the four channel instruments. With the v2, customers started to get comfortable that we had achieved parity in the quality of the output that they could get, and they started trialing those instruments and validating them. And so this was a trend that's been building.

There are definitely things we are taking away from that, on a number of areas. One is, we are very pleased with the traction that we're seeing, for example with two-channel chemistry and the benefit that it gives to our customers. Internally, that's something we're thinking through, how we leverage that even more broadly.

There are other lessons we're taking around -- NIPT is one of those markets where we are focused specifically around the end-to-end work flow, and making our solution easy for our customers to consume in the NIPT market, and you see that with the VeriSeq launch that we're doing, as well. And for us, that represents a good way for us to get good traction in the applied markets. And so that's another lesson we're taking from the success we're seeing in NextSeq.

Is it fair to say that's probably mostly in the US then, just given that O-US, it seems people were starting more with NextSeq and NIPT?

No, we're seeing that in countries outside the US too. For example, we're seeing that certainly in China, is a good example of a NextSeq market.

That will continue to be the case next year, when we launch VeriSeq NIPT in Europe, with CE Mark.

Got it. I had a quick follow-up to Dan's question on the X. So historically, you've given us some context around what the run rate looks like on SBS chemistry just more broadly. Is there a way you can help us think about that same concept on the X? So just thinking about the platform that's out there now, and assuming that you don't have to make major instrumentation changes? How much more run rate do you have there, in terms of thinking about cost per genome, or whatever metric? Just on what's out there today.

Obviously I can only answer that question at a very high level. I'll say that, look, we are really optimistic about the headroom that all of our core technologies have. So the SBS chemistry, the flow cells, and so on.

And so from a technology perspective, we certainly believe that we have headroom to push the performance of the X on a number of dimensions. For us, the question's always going to be also around what the market needs to unlock the elasticity, and how much of that will be driven by price point, how much will be driven by informatics, and so on. But from a capability perspective, I think we have good line of sight into continued improvements on the X, but also on our other instruments.

Got it. Thanks very much.

Thank you. And our next question comes from Bill Quirk from Piper Jaffray. Please go ahead.

Two quick interrelated questions. First, Marc, would you be so kind to elaborate on the new forecasting project, and I'm thinking maybe you could give us an example of a process that will be changed as a result of this? And then again somewhat related, just help us think a little bit about -- what would potentially stopped additional customers from transitioning to other platforms and obviously disrupting your consumables flow, like we saw in this quarter? Thanks.

I would say on the forecasting project, I think of it in terms of both our visibility and predictability of the pipeline. So there are a number of different things that we can do, including connecting the market insights that we have, as well as improving the market insights that we have, and connecting those to the forecasted pipeline. And driving more discipline around our CRM database, around our pipeline definitions and so on.

It's much broader than just forecasting, which is why I said this could continue into next year. It's between the sales process and the forecasting process, itself. We might need to make changes in both areas.

I think we do it really well in certain spots, in a number of regions. And as we've always said, we felt like Americas had a very disciplined process. But I think we don't do it as consistently globally as we probably -- as I know we can.

So driving global consistency and the way that we do it, the items we look at, the dashboards we use, even the taxonomy. What do you mean by defining opportunity versus upside, and those kinds of things. Like I say, in some places, we do that really well but I'd like to drive more global consistency around that.

And then your second question on transitions, I think generally, it's okay for us, if our customers are transitioning to other products and platforms. There's normally good reason for that, and we still believe in elasticity, and over time, that revenue comes back. The fact that it causes a dip, as those customers transition, I think is part, goes with the territory.

And so key there, again, back to the forecast process, is our ability to predict that, versus being surprised by that. But the actual dynamic itself, I don't think is one we need to worry about, as long as our customers don't take multiple years to transition, which would be more damaging to their business than it is to ours.

Of course. Understood. Thank you. One real quick follow-up which is, the X order that slipped out of the quarter, was that a 5 or a 10? Thanks.

We didn't say. It was a large X order.

Got it. Thank you, Francis.

And our next question comes from Isaac Ro from Goldman Sachs. Please go ahead.

Want to spend a minute on expenses. If I look at your guidance for the balance of 2016, was hoping you could maybe give us a sense of the SG&A aspect of the spend, and to the extent that informs a run rate, it was a little bit lighter than we expected this quarter. I'm sure that will rebound a little bit into year-end, but want to hopefully get a sense of a normalized run rate going forward.

Yes, Isaac. One of the reasons it was -- the SG&A was lighter this quarter was back to that compensation. By the way, I want to make sure I correct that because I misspoke earlier when I was answering Ross' question. In Q3, we saw a $16 million benefit in OpEx, not $13 million, as I said.

In Q4 it was $6 million. But that was one of the key drivers of that. We're in the middle, as I said earlier, of our 2017 budgeting right now. And of course, as we do that we're going to be very cognizant of our spending rates, and so on.

We are going to continue to invest in sales team, and people who carry quota. That's clear. But the normal balance we go through every year is trading off between how much do we spend on R&D, and as you can tell, it's trending up quite high, and much higher than our peer group, because we believe in the technology and the headroom Francis mentioned earlier.

The trade-off might be sales and marketing or even G&A. I try to drive G&A down, and provide leverage to drive it down as a percentage of revenue, and that will always be the case. But we'll make trade-offs between R&D and sales and marketing.

Remember, Isaac, that Marc's commentary is primarily associated with core Illumina and that GRAIL and Helix are going to continue to ramp throughout the year, and those obviously will increase operating expenses for our consolidated financials as a result.

That makes sense. Thank you. A quick follow-up on GRAIL, hoping you can provide an update on, I think last we heard, assay design hopefully on track to be finalized early next year. Wondering if that's still the case, and if you could comment on their progress? Thank you.

The GRAIL team is making really good progress. We will at some point schedule a broader update on where we are with GRAIL. What you have seen is the team has started work on the circulating cell free tumor atlas, and that data we expect to come out in the back half of next year.

And so while the team is making good progress on defining the assay, and is roughly on track, we expect the data that comes out of that will lead to some modifications of the assay. The team continues to recruit an incredibly strong team, and so, we are pleased with the progress they're making. We're very pleased with the talent level of the team that we're building there, too.

Thanks.

And our next question comes from Steve Beuchaw from Morgan Stanley. Please go ahead.

Hi, good afternoon. Just two quick follow-ups from me. First is on arrays. I thought the commentary on arrays and the growth sustainability there sounded a bit more positive on the sustainability than we've heard in the past. In the past, we've heard about a mix of volume and offsetting price. Can you speak a little bit to the mix between volume and price that you're seeing now in arrays, and how those evolve to get you to your view that there's a pretty sustainable healthy growth profile?

I think it's a fair read of our take on the market, that we are seeing strength in that market, driven by a number of factors that I'll talk about. And internally, the team that runs it calls it an array-a-thon. What's driving that momentum, there are a number of things.

One is the direct to consumer business we're seeing real strength from our customers in that segment, exceeding frankly our expectations, substantially. And so that's driving some of the growth you're seeing in the array market. We talked about the strength of the ag business, and those are some fairly large orders that we're taking in that market. And so those create a really healthy backlog for us, and give us cause to be optimistic about the revenue flows for the array business.

That's a business that I think really benefits from elasticity, and so I think you'll see both play out. You'll see prices continue to go down, as we move to routine testing in the ag business. And you'll see the lower prices drive more demand, as we're seeing in the direct to consumer business, as well. So it's a great example of real strong elasticity in a market.

And then, I would love to hear updated perspective from you on the ramp of the commercial opportunity at Helix. I appreciate that this is still early but just watching the flow of press releases on partnership announcements from Helix over the last couple months, I thought it would be a helpful topic to cover here, as we think about planning for the model, and where consumer could go? Thanks.

It's a great question. You're right, there have been a number of partners that have been signed up to Helix, and a number of announcements about those partners. So I think there are partners like obviously Mayo and LabCorp and Duke and Good Start Genetics and National Geographic and Invitae.

But there are also interesting partners like the team from ExploraGen, and what they're doing is they're creating an application that harnesses DNA and the science of taste to create personalized culinary experiences, and their first application is the Vinome. And what they want to do is create a direct to consumer application that will match wine to consumers.

That's an example of an application that we couldn't have predicted, and frankly we were hoping to see by launching a platform like Helix, to encourage all sorts of experimentation and interesting applications show up. And we're seeing that, even beyond what we've announced to the public.

Helix is driving a really interesting pipeline of opportunities in all areas, of whether it be health or wellness or genealogy or whatever. It's really interesting what they're doing. The key next step really, I think for Helix, is the product launch, and that's what the team's focused on right now.

Thanks so much.

And our next question comes from Tim Evans from Wells Fargo. Please go ahead.

Marc, just still trying to think about maybe the run rate for the GRAIL dilution, going into 2017. I mean, it looks like it's going to be like $0.15 in Q4, but I know that's not the run rate. Is there any way you can help us think about how we should be modeling that into next year?

I'm not sure where you got the $0.15, unless you're including this $0.08 of incremental. And I wouldn't do that. The $0.08 of incremental is a 2016 full-year issue.

Now, it could happen again. That way, if there are any changes in the common stock of GRAIL, something we need to account for, going forward. But as I mentioned just now, the EPS impact of GRAIL in this quarter was $0.05.

So that's, after all the non-controlling interest calculation, and everything else. That doesn't include the $0.08. $0.05 is the current run rate.

Q4, I'm expecting, as you saw, I took the dilution up by a couple cents. I'm expecting expenses to ramp a little bit higher in Q4, and then that will determine the run rate going forward.

And then the next question is to what extent does GRAIL continue to grow their spending in 2017? And that's part of the budget decision that we're all in the middle of right now. Does that help?

Yes, it does. Thank you.

Thank you. This concludes the question-and-answer session. I'll now turn the call back over to Rebecca Chambers for closing remarks.

Thanks, Katie, and thank you everyone for joining us this afternoon. As a reminder, a replay of this call will be available as a webcast in the investor section of our website, as well as through the dial-in instructions contained in today's earnings release. This concludes our call and we look forward to our next update following the close of the fourth fiscal quarter.

Thank you, ladies and gentlemen. This concludes today's conference. Thank you for participating, and you may now disconnect.