Ladies and gentlemen, thank you for standing by. And, welcome to the GSI Technology's Second Quarter Fiscal 2019 Results Conference Call. [Operator Instructions]. Before we begin today's call, the company has requested that I read the following Safe Harbor statement. The matters discussed in this conference call may include forward-looking statements regarding future events and future performance of GSI Technology that involves risk and uncertainties that could cause actual results to differ materially from those anticipated. These risks and uncertainties are described in the company's Form 10-K and filed with the Securities and Exchange Commission. Additionally, I have also been asked to advise you that this conference call is being recorded today, October 25, 2018, at the request of GSI Technology. Hosting the call today is Lee-Lean Shu, the company's Chairman, President and Chief Executive Officer; with him are Douglas Schirle, Chief Financial Officer; and Didier Lasserre, Vice President of Sales. I would now like to turn the conference over to Mr. Shu. Please go ahead, sir.

Welcome, everyone, and thank you for joining today's call to review our second quarter 2019 results. The revenue for the quarter was $12.8 million, above the high end of our guidance provided on the first quarter's call of $11.5 million to $12.3 million. Gross margin in the second quarter were 63% above our guidance of 57% to 59% provided earlier last quarter. Doug Schirle, our CFO, will provide additional insight into the factors that saw some revenue growth and the margins in his commentary. Our APU team made significant progress during the quarter implementing the development of our patented in-placed associated computing technology, the APU. We recently received the first wafer from the fab and we are now focusing on assembling package chip for testing. It is an exciting time at GSI as we approach this crucial step of determining the viability of our first AI product. In that testing process, our Israeli team were [indiscernible] liable in algorithm to identify [indiscernible] and we also plan to engage with our customers, both enterprise and institution. They are willing to test a very early stage chip in their specific application. The team has worked very hard to get to this critical point, but we still have a lot of work ahead. We are eager to embark on the next phase of this journey with the APU. We recently announced a collaboration with the Weizmann Institute of Science, one of the world's leading scientific research institution. This collaboration is the result of a pending marked in scope of our APU team as they look to build relationship with target of our customers and institutions interest in evaluating our unique solution for complex data and analytic applications. A leader in the field of bioinformatics, which combines scientific and technology footprint, to analyze and…

Thank you, Lee-Lean. We reported a net loss of $351,000 or $0.02 per diluted share and net revenues of $12.8 million for the second quarter of fiscal 2019 compared to $1.7 million loss or a loss of $0.08 per diluted share on net revenues of $9.6 million for the second quarter of fiscal 2018 and a net loss of $1.6 million or $0.08 per diluted share on net revenues of $11.3 million for the first quarter of fiscal 2019, ended June 30, 2018. Gross margin was 62.6% compared to 50.4% for the prior-year period and 51.4% in the preceding first quarter. Strong revenue growth in the second quarter was driven by strong demand from our largest customer and a partial shipment against a one-time order for university supercomputer build-out in Europe. In addition, these factors drove an increase in gross margins, which also benefited from product mix and a decrease in our inventory reserves compared to the preceding quarter. Total operating expenses in the second quarter of fiscal 2019 were $8.4 million compared to $6.7 million in the second quarter of fiscal 2018 and $7.4 million in the preceding first quarter. Research and development expenses were $5.8 million compared to $4.2 million for the prior-year period and $4.9 million in the preceding quarter. Research and development expenses in the quarter ended September 30, 2018, included an expense in the amount of approximately $1 million related to a non-production mask set for our initial APU product. Selling, general, and administrative expenses were $2.7 million compared to $2.5 million in the quarter ended September 30, 2017, and $2.6 million in the preceding quarter. Second quarter fiscal 2019 operating loss was $394,000 compared to $1.7 million in the prior quarter and $1.8 million in the comparable period a year ago. Second quarter fiscal…

[Operator Instructions]. We'll take our first question from Mike Crawford with B. Riley FBR.

I was hoping you could update us a little bit further on the prospects of the APU. When you acquired MikaMonu nearly 3 years ago, I think they were -- you -- I think you were talking about hiring a team of about 16 people to bring that to market and I think that number is nearly double that now and you also had talked about certain enterprise customers seeing this as maybe even 1,000 times better than existing solutions. And you know what -- how does that compare with what you see now that you have some wafers back from the fab?

Okay, run of the team, Israeli software team is about close to 30 people and also in order for us to field the APU product, we over [indiscernible] the processing logic debunking, so that's about, you know, plus 10 people. So totally I would say it's about 40 people for the new, the new group. And what was the second question?

Was asking about where we, after we've seen the wafers, where we see how the part stacks up versus the competition?

Okay, the performance. We, okay, so APU is a very, very good in the similarity search. Okay, so at this moment in the market, competing solution, we don't see anything close to our solution. So I think that a thousand times improvement is still bad in terms of similarity search. So that is going to be our focus point.

But to be clear, you said once we've seen the wafers, the wafers are out but we -- they're being assembled so we haven't actually tested them yet. So everything that we're discussing is based off of software simulation.

Okay. So then that's good to hear that, you know, you still see the same relative several order of magnitude benefit over competing technology. So, as you assemble these wafers and you get them out for sampling and get some feedback and then you come and do it again, is this -- you know how might revenues layer in or in what periods might revenues start to layer in relative to the effort?

So that's a difficult answer or a difficult question to answer. The reason I say that is again we have -- the wafers are out. We don't know what level of functionality we have with first silicon at this point. So, we'll know more in the next couple weeks when we get assembled or packaged units in our hands. As Lee-Lean mentioned in the script, we will be sending boards, eval boards to Israel so they can run their algorithms and libraries through the device. And then also we have a few friendly customers that are willing to look at the boards as well. So, you know certainly it depends on the, you know what kind of debugging we need to do and the timeframe for that before we can determine when the revenues will kick in.

Maybe if I could take one final stab at it, that you know if assuming that you -- there are some problems detected and then you come back and you get another set of wafers that then perhaps could pass muster under a scenario like that, which maybe that's typical. Is this something where you know by this time next year we would be seeing meaningful revenue or it's just also too difficult to answer?

Yes, I think it's really hard to predict how the debugging process will go. But, typical you will see half year to one year, you can see the real work in silicon. I think that's typical.

We'll go next to Kurt Caramanidis with Carl M. Hennig, Inc.

Good quarter. Question for you, how did this kind of develop with these friendly customers? I think if I remember right, you were going to debug kind of on your own and what made some of these customers want to help in that process? Are they using some of their IT people? Kind of how will that go and how did that come about?

So certainly we've always anticipated that we were going to look at the part in 3 different ways. Certainly we're going to look at it as a component level on memory testers and then we're also going to send the eval boards to Israel to do our internal library and algorithm check. And then it's always important to get real data and real software flowing through the parts. And that's where we reached out to some of the folks who we had presented our solution to early on and saw that it was a very compelling solution for them. And so they were willing and in some cases wanted to sign up to have early access to the silicon. And so that's a process that's been happening for over a year at this point.

So do you see -- when do you see there's some depends on here but the board getting into these first friendly's, is that -- it sounds like it's after you first get it put together, run the simulation and then -- would that be like January, February type of thing or--?

Yes, so again it depends on with first silicon looks like. Obviously if there's some functionality to it then that timeframe is -- it'll be sometime we're looking hopefully in the end of first quarter, calendar quarter of next year. But again it completely depends on the functionality of the first silicon.

Okay, great and finally, any AI conferences. It sounds like George has been at a number of them. Any of them coming up between now and the end of the year that are worthy of putting on your website?

So we're actually going to be hosting a panel at ODSC West and I believe it's November 2. So it's coming up next week. And so we'll be hosting the panel and we'll have folks from you'll see Wayfair, I believe eBay, I believe Google, Wal-Mart labs and Clarify, I believe are the 5 folks that will be on our panel. And it'll be specifically to discuss similarity search.

Will you be able to run that live on the web?

I don't know what the process is going to be. I know it will certainly be recorded. I don't know if it'll be live though.

Okay but otherwise you'll get it up -- that's great. Okay, super.

We'll take our next question from Jeff Bernstein with Cowen.

Just a couple questions. Have you guys been able to just do -- just probe these wafers and get some electrical characterization and how does that look?

No, no, this wrong. We pass wafer probe. We just go straight to assembly.

Got you.

So we will see when we receive the package unit.

Got you. And then just on the mask set expense, so because this was a prototype basically it gets expensed in this quarter now that it's delivered or how do these get treated sort of going forward if there's another mask set?

Yes, typically when we've run a product on a technology that we've never run before, we'll expense that to R&D, which is what we did this time. This is the first time--

Got you.

We run product [indiscernible]. Future mask sets will all be charged to cost of goods sold over a one-year period.

Got you, terrific. And then lastly I just would mention, I don't know if you guys saw there's a high EEE spectrum article out, a Q&A with Micron Technologies memory mastermind, essentially blessing the APU approach as kind of one of the futures of where processors need to go?

Yes, we haven't seen that specific article but we've seen a few in the past that certainly bypassing the von Neumann model and processing and memory is key. We and we certainly have seen several articles on that subject.

[Operator Instructions]. We'll go next to Jim Kennedy with Marathon Capital Management.

Would you just remind us and I guess in layman's terms why Weizmann is pursuing looking at this for bioinformatics? What the APU does that perhaps another chip architecture cannot do at this point? I mean what can you, can you give me some of the just fundamental benefits that cannot be achieved by other alternatives?

Sure. It's not that they can't be achieved with other alternatives, is that they're done much slower with other alternatives. So they're trying to accelerate the time that they can go through processes like drug discovery and virtual screening. So, they're -- they are using solutions now. They're mostly CPU based and they're very, very slow. And so when we've presented our device to them, they're very excited. And then we ran some simulation for them specifically on the software platform that they were using and the results came back very, very encouraging. And so that's why they're certainly going to be one of the first friendly's to look at the eval board.

Is it proper to call this machine learning or artificial intelligence or is there another terminology we should be using?

It's not machine learning. So this is, this falls under kind of what Lee-Lean alluded to earlier, which is similarity search. So we're essentially doing similarity search based off of molecules. And so this is not machine learning. It falls under similarity search.

At this time we have no questions in the queue, we do want to give everyone a final opportunity. [Operator Instructions]. We'll take a follow-up from Jeff Bernstein with Cowen.

Yes, just on the aerospace/defense side, so the characterization of the rad-hard parts at the -- sort of to the higher standard, you know does that have any impact in terms of where you can go with this part with other customers? And just any color you can give on any other thoughts about additional customers in the pipeline?

So, long term, we were always going to go to that higher standard. We knew that was a requirement for this particular program for the first customer. What was news to us was that the prototype to build out before the production quantities came in was expected to be at a lower standard. And they have come back and said they would like even the prototype devices to be at the highest standard that they're going to be buying it production. So it won't affect the long term because that was anticipated. It just delayed a little bit the prototyping because that was a change from our customer.

Got you and any feel for interest level from other customers at this point?

Yes, certainly l mean there are other programs we're discussing. This just happens to be the first one in the pipe, but we're discussing with other customers not only our rad-hard but also the rad-tolerant programs. And those will be kicking in later in calendar 2019 and in -- and then into 2020.

At this time, we have no further questions in the queue.

Thank you all for joining us today. We look forward to speaking with you again when we report also for the fiscal 2019 results. Thank you.

Ladies and gentlemen, this concludes today's conference. We appreciate your participation.