Lee-Lean Shu - Co-Founder, Chairman, CEO and President Douglas Schirle - Chief Financial Officer Didier Lasserre - Vice President of Sales.

Robert Mertens - Needham & Company Jeff Bernstein - Cowen Prime Advisors Kurt Caramanidis - Carl Hennig Incorporated.

Ladies and gentlemen, thank you for standing by. Welcome to the GSI Technology's Preliminary Fourth Quarter and Fiscal 2017 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.

These risks and uncertainties are described in the company's Form 10-K filed with the Securities and Exchange Commission. Additionally, I've also been asked to advise you that this conference call is being recorded today, May 4, 2017, 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..

Good afternoon, everyone, and thank you for joining us. Today, we reported fourth quarter net revenue of $10.4 million, which was within the range of guidance that we provided earlier in the fourth quarter.

The revenue decline was consistent with our expectation, though lower than anticipated and resulted, in part, due to an inventory correction at our largest customer that began in the third quarter and continued to affect fourth quarter sales.

We also saw continued slowness in our primary telecommunication and the networking market and continued weak sales in Asia.

We were pleased to see our gross margin came in above our guidance for the fourth quarter at 56.4%, which was driven by the continuing very favorable mix of higher-margin products, including in our military and the defense business.

We remain focused on expanding our market position in the high-speed SRAM and low-latency DRAM segments, including our extremely high-performance SigmaQuad radiation-hardened SRAM products targeted at aerospace and the defense applications, which we expect to introduce in second half of calendar 2017.

In the meantime, we have stepped up our marketing efforts to highlight these products to increase attendance of industry conference and our workshop.

Meanwhile, we are continuing to develop - to devote majority development resources toward the creation of a new category of products based on our patented in-place associated computing technology and intellectual property.

These products will utilize massive parallel data processing capability to greatly improve computation, search and the response time for use in a variety of big data applications, including machine learning and the convolutional neural networks, computer vision and cyber security.

We are very optimistic about our prospects for fiscal 2018 and beyond..

For the year, we reported a net loss of $115,000, or $0.01 per diluted share, on net revenues of $48.2 million, compared to a net loss of $2.2 million, or $0.10 per diluted share, on net revenues of $52.7 million in the fiscal year ended March 31, 2016. Gross margin was 54.8% compared to 50.7% in the prior year.

Total operating expenses in fiscal 2017 were $26.9 million compared to $29.8 million in fiscal 2016. Research and development expenses were $15.8 million in fiscal 2017 compared to $12.1 million in the prior fiscal year.

And selling, general and administrative expenses, which included litigation-related expenses, were $11.1 million compared to $17.7 million in fiscal 2016. Litigation-related expenses in fiscal 2017 were minimal at $142,000, down substantially from $6.7 million in fiscal 2016.

The litigation-related expenses were incurred primarily in connection with the commercial and trade secret lawsuit against United Memories, Inc. and Integrated Silicon Solutions, Inc. As previously reported, the trial of the UMI/ISSI litigation concluded on November 25, 2016 with mixed results, and appeals of the case remain pending.

Expenses related to patent infringement and antitrust litigation involving Cypress Semiconductor, which was settled in May 2015 also contributed to fiscal 2016 expenses.

We reported a net loss of $1.3 million, or $0.07 per diluted share, on net revenues of $10.4 million for the fourth quarter of fiscal 2017, compared to a net loss of $87,000 or $0.00 per diluted share, on net revenues of $12.2 million in the fourth quarter of fiscal 2016 and net income of $348,000, or $0.02 per diluted share, on net revenues of $11.5 million in the third quarter of fiscal 2017, ended December 31, 2016.

Gross margin was 56.4% compared to 50.3% in the prior-year period and 56.6% in the preceding third quarter. Fourth quarter of fiscal 2017 operating loss was $1.5 million compared to an operating income of $234,000 in the prior quarter and an operating loss of $156,000 a year ago.

Total operating expenses in the fourth quarter of fiscal 2017 were $7.4 million compared to $6.3 million in the fourth quarter of fiscal 2016 and $6.3 million in the preceding third quarter. Research and development expenses were $4.2 million compared to $3.4 million in the prior-year period and $3.8 million in the preceding third quarter.

The increase in R&D expenses is related to the development of a new category of products based on our patented in-place associated computing technology and intellectual property.

Selling, general and administrative expenses, which included litigation-related expenses, were $3.2 million compared to $2.9 million in the quarter ended March 31, 2016, and up sequentially from $2.4 million in the preceding quarter.

Total fourth quarter pre-tax stock-based compensation expense was $518,000 compared to $429,000 in the prior quarter and $463,000 in the comparable quarter a year ago. Depreciation expense for the fourth quarter was $343,000.

Sales to Nokia were $4.2 million, or 40% of net revenues, during the fourth quarter, compared to $4.5 million, or 39.5% of net revenues, in the prior quarter and $4.3 million, or 35.1% of net revenues in the same period a year ago.

Fourth quarter direct and indirect sales to Cisco Systems were $745,000, or 7.2% of net revenues, compared to $1.1 million or 9.8% of net revenues from the prior quarter and $1 million or 8% of net revenues in the same period a year ago.

Military/defense sales were 24.3% of fourth quarter shipments compared to 15.9% in the prior quarter and 19.5% in the comparable period a year ago. SigmaQuad sales were 53% of fourth quarter shipments compared to 54.2% in the prior quarter and 53.5% in the fourth quarter of fiscal 2016.

In March 31, 2017, we had $49.9 million in cash, cash equivalents and short-term investments and $12.9 million in long-term investments; $57.8 million in working capital; no debt; and stockholder's equity of $86.4 million.

Looking forward to the first quarter of fiscal 2018, we currently expect net revenues to be in the range of $11 million to $12 million. We expect gross margin of approximately 55% to 57% in the fourth quarter. Operator, at this point, we will open the call to Q&A..

[Operator Instructions] And your first question will come from Rajvindra Gill, Needham & Company..

Hi, this is Robert Mertens on behalf of Rajiv. My first question would be towards your in-place associated computing. Can you sort of walk me through your products roadmap and what sort of applications that this might be targeting first.

Is it specifically search and data centers and then some other applications or a wide array of applications once the product rolls out?.

As you mentioned that where we're going to be focusing on initially is the search function, also areas like data mining and recommender systems. The way that our part is architectured, that's the area that we have the largest advantage. We've spoken in the past that our part has a 500x power performance advantage in those sectors.

We certainly feel in the other areas of AI we have an advantage but it's not as dramatic as this. So this is the area that we're certainly going to focus our efforts initially..

Okay, thank you. And then I think I had - last time we spoke that you were sampling a chip in the first half of 2018 and then maybe we'll see revenue in 2019.

Is that correct? Is that for the fiscal year or calendar year?.

The calendar year..

Okay. And then also I was thinking on the last call you had 2017. So the first quarter you're guiding towards somewhere around $13 million in revenue, and now it's between $11 million and $12 million. Can you sort of walk me through what brought down that estimate a little bit.

Are there design wins that are being pushed out from the first quarter?.

So it was actually twofold. We have seen a bit of a rebound from Alcatel. As Lee-Lean mentioned earlier, there has been an inventory correction and so we're seeing a small bounce-back starting in this quarter. We anticipated that bounce-back to actually be a little bit larger.

But the bigger factor is we had a few military customers that we're putting in last time buys for older DIREV devices that were put in place when we made that comment. Ends up one of the customers who was going to be placing or had placed the last time buy was able to qualify our new DIREV.

Initially they had indicated they would not do that and they decided to qualify the new DIREV. So one of the last time buy orders that we anticipated hitting this quarter isn't going to happen, so instead they'll continue to use our parts in the future with the newer DIREV ..

Okay, great. And one more for me before I step back in the queue.

Could you just talk us through a little bit about what caused the gross margin beat this quarter and how we should think about that going forward?.

Well, during the quarter, it's tough to predict exactly what's going to be happening to the quarter. We pull our information together and put together forecasts, and as you've seen, Alcatel has not been what we were necessarily expecting. And additionally, the military business was up quite a bit this quarter.

And if you take like, I think, we're about plus 25% in the military business for this quarter and that business has extremely high margins, and that's probably the biggest reason for beating the forecast. And then in terms of going forward, we've recently revised our gross margin target with the military business coming on and the APU.

Looking out 3 to 5 years, we're targeting somewhere between 59% to 63% gross margin. So that is up from where we are today. But it should be better than what we've been seeing this past year which has been in the - closer to 55% range.

But the rad-hardened military business is an extremely high gross margin business and the APU processor in-place associated processor will be better than the legacy products that we've been shipping recently..

From Cowen Prime Advisors, Jeff Bernstein..

I was a little surprised to see you talking about the neural networking and machine vision. I thought that maybe the architecture of the chip did not allow for as complex instructions.

Does that mean that you will be doing that more in software and it would be better suited towards things like automotives that aren't in a big giant data center? Or can you just talk about how you get to the neural networking and machine vision on the architecture?.

Yes, all architecture is really - when you look at our architecture, it's really a massive parallel processing. We call it in-place, okay, which means that we are through the market parallel processing inside the memory array. So today, the GPU many for the - machine learning and many GPU end layer.

We have a similar characteristic, the market parallel processing, but we do have the advantage. If we can be inside the memory, so we integrate the CPU and the memory interface. Okay, that's our advantage. Of course, I mean, today's algorithms is developed based on the GPU.

So we do have some [indiscernible] things, okay? But we're clearly in angle of search that's where because of the venues that we have. But how cheap can the search without chip can rolls up through the market parallel processing just like GPU.

So I think, once we have our product in the marketplace and people start familiar with the concept of our APU product, then I think we definitely can be certain market penetration in the machine learning and the convolutional neural network..

And Jeff, let me add to that. You had mentioned software, so let's be clear that this is a hardware and a software solution that we're offering. So we do have a custom chip that we're designing and hardware, but it will have libraries that go along with the chip and the libraries will be suited for different functions and different market spaces.

So there will be libraries that are more suited for, like you said, the auto industry. There'll be some that are more for the data centers, more there for cyber security. So it will also have a software component to it..

So in pure speed, you should be better than the PSPs GPUs, et cetera.

Are you going to have more complex software running on it and therefore have a little bit less of a power performance advantage on some of the more complex kinds of algorithms?.

We will have a speed and a power advantage. And as Lee-Lean mentioned a second ago, the GPU solutions today have to go off-chip to retrieve data. I mean the data stored in GDDR DRAM is basically off-chip.

So when you're going back and forth from GPU to memory, that is not only losing cycles time-wise, but it's also all the I/O cycles going back and forth uses a tremendous amount of power. As Lee-Lean mentioned, we do ours in place. So we have the data stored on-chip and we do the calculations and the search on-chip.

So we are not going off-chip to receive or fetch data. And so that - you're seeing a lot of time and a tremendous amount of power, so that's both of an - we have both those advantages..

[Operator Instructions] Next, we'll go to Kurt Caramanidis with Carl Hennig Incorporated..

Anymore winds on the rad-hard business of late? I think you have 2 to date..

We'd say 2. We don't have any orders yet because we don't have our products yet. So our first ES device - in fact, we pushed out a couple of months, the package design took a little longer than anticipated because this requires a custom package, specifically for a dye.

So we're looking to having ES parts sometime in September, October, which the ES parts won't actually be for sale. They're not given away. And then we'll have our production parts, which originally we're hoping for December, now looking like February of 2018. With that said, we have 2 folks that have immediate needs for the parts.

They are anticipating it very quickly. But then we have 2 others as well that will be placing orders that looks like for the initial batch of ES parts in the September, October. With that said, we have identified other customers.

It's just a matter of when are they going to get funding for whatever program they have and what's their window for qualification. But 2 are certainly pressing. 2 others will most likely be in the queue for the early samples. That like I said, we'll be shipping sometime, best case, September, and more likely, October..

Okay. So we'll see that in either Q3 or Q4 revenues..

That will be right..

Calendar. Calendar Q3 or Q4. Is this ....

Right..

Right. Yes, I don't think it can hit calendar Q3. I think it's going to be more like calendar Q4. Yes, because October falls into 4 and I think September is going to be a real push for us that that will happen..

And then just kind of reflecting back, Lee-Lean and Didier maybe. 18 months ago you just buy the associated computing business.

Where are you at today with your confidence compared to where you were when you first bought that business?.

Yes, I think we really have a unique architecture. I think that's an - that's actually better than my original expectations. So today, we have another conversation with our customers - target customers, the people in the big data area and the people who - in the care computing, I mean, so far, I mean, all we've got is it's a fantastic solution.

It's really up to us to deliver the product..

Didier, the same thing?.

Yes, absolutely true. The customers that we had initially spoken to that showed interest are still very excited about the part.

Certainly, they would like to have it immediately, but there is no other solution they see that does what we can do anywhere in the horizon, so they're still working with us closely to make sure that we're aligned for when we finally do have the silicon available..

And if we get that at the end of the year which I guess is the target, where are we, let's say, a year from today that's being sampled or what should we look at say a year from now, where do you think we would be? I realize that's kind of difficult, but I'm trying to get my arms around modeling and where about we would be, let's say, a year from now, you might be 4 months into sampling?.

No. So the design will be done at the end of the year. So that doesn't mean we will have silicon at the end of the year. At that point, we have to spread the message and get silicon and put it in the package and evaluated.

So we're looking at having some kind of product that we can evaluate sometime I would say in the second calendar quarter of next year. And that's why we're talking midyear is what we're hoping for we've got something available to show to customers.

But, yes, at the end of this year [indiscernible] the product available, that's when the design is complete..

[Operator Instructions] We'll go to Rajvindra Gill with Needham & Company..

This is Robert again. Just sort of a question on your in-place associated computing, if it has power requirements much lower than existing solutions than the GPUs.

Do you see this as a potential for, say, computer vision or natural language processing for IoT devices and namely maybe running off of a battery or something is that maybe farther down the line than you guys are looking into right now?.

No, no, we are not looking for the mobile application. I think our - right now, at least for now, our primary target is for data centers. Somewhere down the line, maybe we will develop something for mobile but not at this moment..

[Operator Instructions] I'm seeing no other questions at this time. I would like to turn the conference back over to management for any additional or concluding remarks..

Thank you, all, for joining us. We look forward to speaking with you again when we report our first quarter 2018 results. Also for those attending, we will be presenting at the B. Riley investor conference later this month in Los Angeles. Thank you..

Ladies and gentlemen, that does conclude today's presentation. We do thank everyone for your participation..