Ladies and gentlemen, thank you for standing by. Welcome to GSI Technology's First Quarter Fiscal 2020 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 the future performance of GSI technology that involve risks and uncertainties that could cause actual results to differ materially from those anticipated.

The risk and uncertainties are described in the company's Form 10-K filed with the Securities and Exchange Commission. .

Additionally, I have been asked to advise you that this conference call is being recorded today, July 25, 2019, at the request of the 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. .

Good afternoon, everyone, and thank you for joining us to review our first quarter 2020 financial results. On the heels of a strong performance in fiscal year 2019, the new year is off to a solid start. In the first quarter, we delivered year-over-year revenue growth of 15% and the sequential growth of 3%.

We continue to see robust sales of our SRAM line, especially, SigmaQuad-IV category, our highest ASP and the gross margin product. This has led to further gross margin improvement with Q1 increasing almost 400 basis points over the year ago period and 210 basis points over the prior quarter.

As a result, we saw a reduction in our operating loss in the quarter compared to the fourth quarter of fiscal 2019..

Cash generated from the strength of our core business is supporting our deal product road map and the funding of our ongoing R&D budget..

Our R&D spending and our engineering teams are focused on Gemini, our in-place associative computing technology. I'm pleased that we are tracking to our norms. On our prior conference call I stated that we anticipated delivery of our second silicon device from TSMC, which we received in May.

Our AI team is pressing the device on evaluation board and continue to develop software library and algorithm..

In addition to the cheminformatics I mentioned on the May conference call, we successfully launched proof-of-the-concept of Gemini's capabilities in database analytics as well as facial recognition for target customer.

In both cases, Gemini significantly exceeds the performance of latest current system and validate the capability with effective one-off solution..

We anticipate government payments later this year with fully functional Gemini chip in the multiple CAST systems on these scale database applications. We will also perform a [indiscernible] and there has been to [indiscernible] in preparation for release of Gemini in fiscal 2020..

Today, our team of over 70 engineers based in the Israel and in the U.S. are focused solely on our ALU map. Our next milestone for Gemini is having the evaluation board for customer demo, which we anticipate beginning in late summer or early fall.

Once we have production-ready board, we will conduct performance benchmarking test to quantify Gemini's performance relative to competitive solutions..

As we try to reach milestones, we will get further clarity on the timing of the launch of our AI product. This is an exciting time at GSI as we achieve the milestone for Gemini and prepared for first shipment of our radiation-hardened SRAM device. Didier will provide more details on the state of our RadHard further in a minute.

The next stage of our company's evolution with new products, solid financial results and the exciting new market opportunities point to a promising future. I would like to thank my team for all their hard work and the commitment to our vision. .

Now I'll hand the call to Didier who will discuss our business performance in further detail. Please go ahead, Didier. .

Thank you, Lee-Lean. Switching gears to our other new product category. We have made significant progress on our radiation-hardened SRAM. We successfully completed the series of critical testing. The RadHard continues to track completion of our internal qualifications end of this year.

The next step will be customer qualifications putting us on track for initial shipments by calendar year-end of 2019..

Looking at our sales breakdown, in the first quarter of fiscal 2020, sales to Nokia were $6.0 million or 45% -- 45.7% of revenue compared to $5.2 million or 46.5% of revenue a year ago and $5.2 million or 41.1% of revenue in the prior quarter. We expect sales to Nokia to continue being strong for the remainder of the calendar year 2019. .

Military and defense sales were 21.0% of first quarter shipments as compared to 19.7% a year ago and 22.1% in the prior quarter..

SigmaQuad continues to be our best-performing product category with sales of 67.9% of first quarter shipments compared to 59.7% last year and 57.4% in the preceding quarter. .

I'd like now to hand the call over to Doug. Go ahead, Doug. .

Thank you, Didier.

For the first quarter of fiscal 2020, we reported a net loss of $125,000 or $0.01 per diluted share and net revenues of $13 million for the first quarter of fiscal 2020 compared to a net loss of $1.6 million or $0.08 per diluted share and net revenues of $11.3 million for the first quarter of fiscal 2019 and a net loss of $102,000 and [indiscernible] per diluted share and net revenues of $12.7 million in the fourth quarter of fiscal 2019..

Gross margin was 63.3% compared to 51.4% in the prior year period and 61.3% in the preceding fourth quarter..

Total operating expenses in the first quarter of fiscal 2020 were $8.5 million compared to $7.4 million from the first quarter of fiscal 2019 and $8.1 million in the preceding fourth quarter..

Research and development expenses were $5.6 million compared to $4.9 million in the year -- in the prior year period and $5.6 million in the preceding fourth quarter..

Selling, general and administrative expenses were $2.9 million in the quarter ended June 30, 2019, compared to $2.6 million in the prior year quarter ended June 30, 2018, and up sequentially from $2.6 million in the preceding fourth quarter..

First quarter fiscal 2020 operating loss was $229,000 compared to an operating loss of $1.7 million a year ago and an operating loss of $365,000 in the prior quarter..

The first quarter fiscal 2020 net loss included interest and other income of $147,000 and a tax provision of $43,000 compared to $23,000 in interest and other income and a tax provision of $10,000 for the same period a year ago. In the preceding fourth quarter, net loss included interest and other income of $186,000 and a tax benefit of $77,000..

Total first quarter 2019 pre -- total first quarter fiscal 2020 pretax stock-based compensation expense was $651,000 compared to $542,000 in the comparable quarter a year ago and $580,000 in the prior quarter..

At June 30, we had $64.7 million in cash, cash equivalents and short-term investments and $7.3 million in long-term investments compared to $61.8 million in cash, cash equivalents and short-term investments and $9 million in long-term investments at the fiscal year ended March 31, 2019..

For the upcoming second quarter of fiscal 2020, our current expectations are net revenues in the range of $11.6 million to $12.6 million, with gross margin of approximately 54% to 56%..

Operator, at this point, we'll open the call to Q&A. .

[Operator Instructions] We do have our first question from Kurt Caramanidis. .

I missed that you did some other testing on facial recognition and something else.

And how did those results look?.

Yes. Kurt, it's Didier. Yes, correct. So as you know in the past, we've done the cheminformatics demonstration, this time around looking at 2 different applications. So the first was facial recognition, and the second was database analytics.

Specifically, in this case, it's the database where it's patent searches, so they're looking for words or sentences..

The demonstrations went well and they were based -- certainly the first one, the database analytics, was based off of our first silicon Gemini 1, which we talked about, we know had some major bugs in it. So it -- the demonstration was -- certainly exceeded all of our expectations.

With that said, they're waiting for the same test or the same demo with the second silicon of Gemini, which certainly looks a lot better. We have the chip out and we're just putting them on boards now. So we're, within probably a month or 2, doing that demonstration for them again. .

Okay. Is that government-related on the facial recognition? Or... .

No. It's -- I mean the answer is the customer we're working with is not a government entity. It's -- do they have some varying customers as government entities? Yes. But this is not the government directly. .

Okay.

So for this Jefferies conference, do you think you'll have anything demo-able? Or is that going to be -- it depends on how things go?.

Are you talking about ODSC West?.

Well, the Jefferies conference you're going to be at in about a month. .

So the Jefferies conference. No. I mean we will not have any hardware with us if that's what you're asking. No.

I mean as far as the time line goes, again, there are -- they're putting the devices on the boards and they're running all the GNL, which is --I'm sorry, which is our libraries and algorithms So they still have some work to do before we start doing the demos for the customer.

So as I mentioned, we're probably a month or 2 away from doing the demos with the second silicon device. .

Okay.

And am I understanding right then, you're talking about launching -- or the release of Gemini in fiscal 2020 which would tell me based on the extra spend that you got to do, that would be calendar Q1 of 2020?.

Yes. Correct. .

Okay. And does that mean that you've got people... .

I believe, in Q3 that we... .

Go ahead. .

Yes. Go ahead, yes. .

Well, so you will be talking with prospective customers between now and year-end and then you would expect you would be able to ship in calendar Q1 if all goes according to plan?.

Yes. I think, yes, for the -- I think for the prototype kind of quantity, yes, I think we will do that. .

Okay. What is this extra IP that you're buying? Is that part of the original purchase of MikaMonu? Or is this something new? And what about that second generation? Is that coming along? I know that has -- you've talked about -- it has a much wider use. How is that... .

Maybe I can talk to where we are in the development of the second generation of the product. But the IP that we referred to in the earnings release this afternoon has to do with some -- basically it's -- IP, it's a part of the product development that we'll be doing rather than generating all of that ourselves.

[Audio Gap].

design of the product. We're actually purchasing some IP that can be dropped into the chip to get us out quicker. It's similar to what we did in first chip. We had some IP that we put in. That was about $600,000 or $700,000. This chip is much more complicated.

There's a lot -- there are a lot more features in the chip, a lot more complication in terms of development -- developing it. And what we're purchasing this time is significantly more extensive and expensive than what we did in the first chip. And then if you want to talk about where we are in the... .

Yes, I think that what Doug mentioned is the IP, I think, you referred to in our -- that's in our earnings report, that's a -- those are the IP we plan to put into our second Gemini product. And the second Gemini product, we have a lot more integration than the Gemini 1. Basically, we will put all the system-controlling function into the chip.

So there's a lot more interfere -- IP and all system-controlling function we need to put in the chip. So that's a hard demo in itself. .

And this is not related to the MikaMonu acquisition at all. .

Okay. That's what I thought. Yes, I just wanted clarification.

Is like that a year behind or so, Gemini 1, let's call it?.

Probably more like 2 year behind. .

Okay. .

You're talking about the introduction of Gemini 2? Look, Kurt, just to be clear, you're talking about the introduction Gemini 2?.

Yes. .

Yes. It's more like 2 years. .

2-year time frame from now. .

We have our next question from Jeff Bernstein. .

Just a clarification and questions. Just so Gemini 2, I think you were calling that Socrates before.

Is that -- that's not something different, right?.

Okay. Yes, yes. We spend a certain time now -- some time now to do the feasibility study, and Socrates is the target to -- for the training -- the AI training market, and also we -- the feasibility study and all the customer synergy, we feel is much better we doing the product development along the search -- demanding search strategy..

So basically, we are changing a little bit to come back to focus on the search market again. So that's why it's the Gemini 2. .

Got you. Okay. And then you talked about the facial recognition and the patent database search.

It sounds like you have software compilers out and people can use off-the-shelf kind of software now and aim it at your chip? So is that stepdown?.

Yes. So we're writing the -- yes, so we call them GNLs, which is the GSI libraries. And so we'll have generic libraries that will be written for certain functions. There will also be folks that want to write their own, so we'll be having classes here in California to teach them how to write their own GNL.

We also writing algorithms for certain applications as well to accelerate even further the hardware..

So the answer is, it's multilevel, there's libraries, there's algorithms. But we'll be writing those ourselves, but then we'll also be teaching other folks to be able to write them themselves if they like. .

Got it. Got you. Okay. And then the demos so far, any comments you would make about sort of power use and heat dissipation? Obviously, you talked about the speed kind of being there.

But what about those elements?.

It's hard to measure that yet. We anticipate, certainly, it will be much lower power. But you have to remember the -- all the demonstrations we've done to date have been based off of first silicon with the identified flaws and bugs. So it'll be much more interesting to see what the results are going to be with second silicon.

So second silicon -- and we certainly have a couple of bugs left, but they're minor bugs, nothing major. And so in that respect, we should start seeing more substantial, I guess that we can call them, benchmarking, what the chip really looks like..

So I think we'll have a better feel for really what the performance is and then we can take into consideration some of the powers as well. .

Got you.

And then in Gemini 1, how much memory is there actually in that chip? And is that something that people care about? And does Gemini 2 have a higher memory? Or can you just talk about that?.

Yes. So there's different layers of memory within the chip. So we have L1 within the chip, we have L2 within the chip, and we have MMB, which is really our secret sauce, the Main Memory Block. And so then we all have -- but the real difference between Gemini 1 and Gemini 2 will be the L1..

If you look at the L1 and Gemini 1, it's roughly 96 megabit of memory. If you look at the L1 in Gemini 2, it's going to be 8 times that. So it's going to be over 750 megabit, and that's what's huge. And we're also making the actual device smaller, physically smaller, so that we're going to be able to load 32 cards into a 1U chassis.

And so, therefore, we're going to be able to essentially take data directly from the host and load it into -- onto our chip without having to worry about an L4 DDR DRAM loading some of the database. So that's really the significant difference between Gemini 1 and Gemini 2 is the -- I mean, obviously, it's much faster as well.

It's 1.6 gig versus 800 to 1 gig on Gemini 1. So the performance will be faster, but certainly, that L1 increase is really the big driver. .

Got you. So you can distribute the work across many of these chips and so the database size is not limited by the individual chip. It's how many cards in the chassis, et cetera. .

It's scalable, correct. .

We now have our next question from Jim Roumell. .

I just wanted to get an update on the Weizmann Institute collaboration and how that's progressing.

Is it segueing into something larger? Can you just kind of give a little color to that?.

So they actually have on their premise our hardware with the first silicon device. And that's what we demo-ed with them. It worked so well with them. They actually took it in-house, and they're using it to find new molecules as we speak. With that said, they are looking very forward to seeing the second silicon.

So when we have a built-up system, we're going to swap out the board they have now with the second silicon board. So it can continue their tests. And really, as I mentioned, we did some bypassing in the first silicon demo just to get around the major flaws. We won't have to do that in this second silicon device.

And so therefore, they're going to -- certainly looking much more forward to that. But as of today, they're still -- they're using our first demo board to do their market research as we speak. .

If you -- do you have any sense, the process they went through before selecting the APU chip and kind of feedback you're getting from how they're using it today?.

I'm not sure if I fully understand the question. So they were using CPU boards in the past, and that's sort of what is -- has dominated this market.

And certainly, it's -- I'm not -- when you go through the process, we told them based off of what they're doing with the similarity search and using your device, what we could do for them and that's when they interested.

I mean obviously, initially it was a PowerPoint presentations, but once we got the first silicon into their hands, they realized certainly a huge performance event for them. Even with the buggy first silicon, they went from minutes for a search to seconds. And then once we give them the complete parts, it'll be milliseconds per search.

So certainly, their initial design or system was using CPU. .

We have our next question from Ari Shusterman. .

Ari taking the question for Raji Gill. So I wanted to start off by asking about the market opportunity for your Gemini chips.

Have you seen any changes in -- and I guess, what is the difference and have you seen any changes in the -- how do you expect it to grow in the next 5 years?.

So the -- certainly, this is an ever-evolving market. Every time you turn around, there's another application for some type of similarity search. As we mentioned, 1 of the 2 new opportunities we had in the database analytics, specifically for patent searches, I know that's something that is new to us.

Nine months ago or a little bit before that, the cheminformatics was new to us. So it seems that there's ever-evolving need for similarity search. So as far as the TAM goes, we see the TAM growing as these new opportunities arise. .

All right.

And have you had any impact -- how has the macro environment impacted your business? Or not much of an impact?.

I'm sorry, can you repeat the question?.

Yes.

So when it comes to macro uncertainty, has that impacted, I guess, the Nokia's behavior or other customers' behavior, any impact to your business?.

The fact that we're going to be offering a Gemini chip?.

No. I was talking about just the macro uncertainty. .

Not sure what you're asking. You're talking about trump and tariffs and all that or... .

Correct. .

That has had no impact on us, at least in terms of our ability to ship product. Didier can talk about the customers, but we've thoroughly gone through the process with our attorneys looking at shipping in Asia. And because of our structure, there's no U.S.-based content and anything that's manufactured offshore and shipped offshore.

So right now, as it stands, the tariff situation is not an issue for us. .

[Operator Instructions] We'll now take our next question from Jeff Bernstein. .

On the RadHard business and if you could give us any kind of update on customer interest there.

And also, what's going on with the rad-tolerant parts?.

Sure. The -- right now, we certainly are seeing -- the -- for both RadHard and rad-tolerant, we're certainly seeing plenty of interest. Mostly of the interest is longer-term, kicking in some time 2020 and beyond. We certainly have the one customer we've talked about for some time, that's more near-term. But as far as interest, the level is very high.

In fact, some of the -- a lot of the customers we're talking about the same test for offering larger-density devices in this market space just because no one has, and they really -- there's certainly a need for higher-density SRAM market. So as far as that goes, there won't be a shortage of customers in this market. .

And we have our next question from Kurt Caramanidis. .

Just one follow-up.

Are you using the success with the Weizmann? Has that opened other doors to other drug companies?.

The answer is yes. Certainly, we need to go through the demo-ing. Well, we're trying to, right now, obviously, get engaged with a couple of customers in each of the market spaces, but we're certainly resource-limited right now because there's -- as we've talked about the first silicon device, there was some workarounds.

But with -- we're anticipating, with the second silicon device, with less bugs that we should be able to open up demo-ing. It's our -- it's -- and I kind of mentioned it earlier, it's our anticipation some time in probably the October time frame.

We're going to start doing some sessions here within headquarters, where we're going to have a few of the boards setup in our data center here at headquarters, and we'll be able to do both shallow-dive and deep-dive trainings for customers that are interested in.

And there's already been folks on both the commercial side and the government side that have already said that they want to be part of some of the initial teachings that we have. So that -- yes, certainly there's that interest. .

[Operator Instructions] It does not appear we have any further questions at this time. I would like to turn the conference back over to the speakers for any additional or closing remarks. Thank you. .

Yes. Thank you all for joining us. We look forward to speaking with you, again, when we report our quarter fiscal 2020 results. Thank you. .

Ladies and gentlemen, this concludes today's call. Thank you for your participation. You may now disconnect..