Published on August 2nd, 2020 📆 | 2320 Views ⚑0
GSI Technology, Inc.’s (GSIT) CEO Lee-Lean Shu on Q1 2021 Results – Earnings Call Transcript
GSI Technology, Inc. (NASDAQ:GSIT) Q1 2021 Earnings Conference Call July 30, 2020 4:30 PM ET
Lee-Lean Shu – President and Chief Executive Officer
Didier Lasserre – Vice President, Sales
Douglas Schirle – Chief Financial Officer
Conference Call Participants
Ari Shusterman – Needham & Company
Jeffrey Bernstein – Cowen, Inc.
Kurt Caramanidis – Carl M. Hennig, Inc.
Please stand by, we’re about to begin. Ladies and gentlemen, thank you for standing by. Welcome to GSI Technology’s First Quarter Fiscal 2021 Results Conference Call. At this time, all participants are in a listen-only mode. Later, we will conduct a question-and-answer session. At that time, we will provide instructions for those interested in entering queue for the Q&A. [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 results and uncertainties are described in the company’s Form 10-K filed with the Securities and Exchange Commission.
Additionally, I have also been asked to advise you that this conference call is being recorded today, July 30, 2020, 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 to review our fiscal quarter, our first quarter 2021 financial results.
GSI has developed a truly disruptive technology with the Gemini APU processor. We are now working on getting the message out to industry and customers on the APU advantage in several search application, including drug discovery, facial recognition, and military and defense application.
To do that, we published a whitepaper in April, documenting the performance benchmark on a one [bidding] [ph] item database. Recently, the Linley Group, a leading publisher of semiconductor technology analysis, supported our funding in their recent whitepaper on the APU. This paper describes in detail Gemini’s revolutionary architecture and give an overview of the competitive advantage with performance examples. Both whitepapers are available on our website.
Recognition is growing for in-memory computing as the ultimate solution for analyzing market dataset, the overall [room] [ph] from human processes. Very few designs with the computation function closer to the data, software from human model indication related to having separate compute and memory functions.
Beyond performance, Gemini’s hardware and software solutions also enables tremendous capability. With the Gemini-I chip installed on the PCIe board that plugs into standard servers, which we call the Leda-G board, we have shown that Gemini has a linear scalability to manage 40 billion records and beyond by combining multiple systems with multiple boards.
The Linley Group’s whitepaper highlights Gemini’s advantage for competitive solutions, in particular, CPUs and GPUs designed to have a high compute capability working with smaller datasets, has limitation when working with large datasets. But the dataset can now exit the [bidding] [ph] license. CPU and GPU solutions require more installed hardware, PC and servers, as well as increased power to keep pace with the expansion of data and this, of search application. This drives out the total cost of ownership, making this solution very expensive.
In contrast, Gemini’s architecture is the optimal solution for search applications for massive datasets. Each of Gemini’s millions of processors can load dataset upgrade from own chip memory, allowing for much greater data growth by removing the human processor/memory bottleneck.
Gemini outperformed standard processor by 100 times or more, reducing power consumption by 70% in search applications. Gemini systems use less hardware for smaller footprint and use significantly less power. This reduced the overall total cost of ownership while delivering better outcome.
As GSI enter new markets and engage with customers, unfamiliar with our company and our technology, we face business challenge magnified by the COVID. Resources such as industry conferences and on-site training workshops, which could become more business process to build a sales pipeline, are unavailable under COVID restrictions.
We are experiencing these indications everywhere, in California and North America, Europe and Asia and with government agencies. The quarter 4 is more challenging to connect and build brand awareness with [indiscernible] for Gemini-I.
As a result, the goals that we have set for Gemini-I in calendar year 2020 allow more challenges to achieve. Our legacy business is also impacted by COVID restriction this year. Overall, our business visibility is reduced in the environment. As said, we anticipate improvement on Nokia in the second half of our fiscal year.
GSI remains on solid financial footing with a strong balance sheet, with $59 million in total cash balance with no debt.
Despite these challenges, we are making progress with customers in the field application and have shipped boards to early adopters. We expect to see DRAM win this year in the few search applications, drug discovery, facial recognition and the various military and defense application. So as I stated earlier, we have limited visibility at this time by tech design wins and modest sales of the Leda-G board by yearend 2020.
In the meantime, our AI team is focused on hardware and software improvements on several front. First, we have continuously improved Gemini-I to increase speed and lower power consumption. By [lispo] [ph], we anticipate providing further improvement to Gemini-I’s performance.
Second, we are developing the software to provide customer with the tool to develop readily adaptable solutions. So we will launch a compiler that will make it easier for customers and the researchers to write their own software. The compiler is for GSI developed Titan-based programming language with supporting APU-specific instructions.
We are also developing a software deducting tool calling NanoSim, that allows a programmer to automatically generate APU machine code and see simulator from their own high-level language program. This software will also allow programmers to visualize and manipulate APU processing array and the instruction interactively. NanoSim is in early stage of design and is undergoing internal prototyping.
We have completely developed a few product-specific libraries, including GSL, short for GSI Search Library. GSL is used to support search functions inside a database. For example, GSL supports a function like framing K-Nearest Neighbor, or KNN, used on billion-scale technology research, as well as [indiscernible] K-NN to be used on molecular research.
We increased awareness of the APU as per knowledge to improve our solutions and develop an ecosystem for our technology. GSI’s FRACTALS Project is collaborating with SHREC, National Science Foundation Center, for Space, High-performance, and Resilient Computing. The partnership will give the GSI team access to a wide variety of higher technology experts.
Lastly, as the CEO and largest shareholder of GSI, I’m heavily invested in developing — delivering on our goals to monetize the investment we are making APU. My team is fully committed to this goal well. We are at a crucial point in our evolution from a legacy company to an innovator in the semiconductor. We understand that it is confidence we have to our shareholders to execute our long-term strategy successfully. I sincerely thank you for your support as parity of our shareholders.
Now I will hand the call over to Didier, who will discuss our business performance in further detail. Please go ahead, Didier.
Thank you, Lee-Lean. I want to expand on some of the points covered by Lee-Lean. GSI’s FRACTALS project partnership with SHREC brings resources to GSI that will help us increase our solutions performance and expand our sphere of influence in this community. SHREC is made up of universities, government agencies and companies from the technology, military and defense and space sectors.
FRACTALS project is focused on Gemini solutions. And in working with members of SHREC, we will be able to leverage many resources from this community. Their focus in this partnership is to advance our radiation hardened, radiation tolerant product lines and the radiation tolerant APU. This new collaboration has already benefited GSI, creating new relationships working with us to improve the Leda-G board functionality.
In the Linley Group white paper, it references the Weizmann Institute work. The paper discusses the APU’s performance using k-Nearest Neighbor, also known as k-NN, for biomedical and pharmaceutical research identifying molecules based on their characteristics. k-NN is a search algorithm that is flexible and easy to implement for classifying data based on similarity, making it highly suitable for analyzing large data sets like molecular databases. However, the high computational costs of performing this search with CPUs and the need for lots of memory in performing a k-NN search limits its use in this application.
GSI technology is excited to be helping researchers at the Weizmann Institute search for COVID-19 treatment. The Weizmann Institute performs k-NN searches using a Leda-G board loaded with 38 million record database of molecular fingerprint. Given the APUs flexible design, they can load the database in different ways, such as 512 bits per fingerprint and 2.3 gigabyte or 1024 bits per fingerprint and 4.6 gigabyte. The Leda-G board can even handle larger fingerprints, such as 2048 or 4096 bits that would take too long to run on CPUs.
The APU’s performance truly stands out as the database has become larger and more complex. The 680 million entry database can be formatted into a data set of 43.5 megabyte of 512 bits fingerprints and loaded on to 4 Leda-G boards, taking only 4.5 seconds to be searched. By comparison, the data set would require more than 1 hour when processed on a CPU-based system.
Another time-saving feature of the APU is the ability to process multiple queries in parallel on the data. This ability improves batch search’s performance also. This is an important differentiator for the APU, which can do 4 parallel searches as compared to a GPU, which cannot do parallel searches, but instead, do them in batches sequentially. For example, on the data set of 38 million molecular fingerprints discussed above and loaded on 1 Leda-G board, a single query will take 1.1 seconds. The card can process 100 different queries at the same time for molecules in 1.41 seconds or 71 queries per second, a function that the CPU cannot perform and the GPUs are not as efficient.
The Weizmann Institute uses Leda-G board to accelerate the discovery of active compounds that may prove useful in COVID-19 treatments. The Gemini-I APU is being used by researches – or researchers to identify compounds structurally similar to Remdesivir, an antiviral drug recently shown to be effective in treating severe COVID-19 patients. Weizmann researchers with Remdesivir as the query molecule are using the Gemini APU on the Leda-G board to perform very rapid similarity structure searches in their in-house database in their search for compounds to develop new treatments for COVID-19.
As Lean stated earlier, we are making Gemini-I faster and more power efficient. We target having an enhanced version available later this year. The AI team is also moving ahead on the next generation of the APU. We are in the early stages of building awareness of our new chip’s capability and functionality.
Shifting to the sales breakout for the first quarter. Sales to Nokia were $1.8 million or 24.9% of net revenues compared to $6 million or 45.7% of revenues in the same period a year ago and $2.4 million or 28.3% of net revenues in the prior quarter. Military/defense sales were 30.9% of first quarter shipments compared to 21% of the shipments in the comparable period a year ago and 30.9% of shipments in the prior quarter. SigmaQuad sales were 46.3% of the first quarter shipments compared to 69.7% in the first quarter of fiscal 2020 and 44.7% in the prior quarter.
I’d now like to hand the call over to Doug. Go ahead, please, Doug.
Thank you, Didier. We reported a net loss of $6.1 million or $0.26 per diluted share, on net revenues of $6.6 million for the first quarter of fiscal 2021, compared to a net loss of $125,000 or $0.01 per diluted share on net revenues of $13 million for the first quarter of fiscal 2020 and a net loss of $3.8 million or $0.16 per diluted share on net revenues of $8.5 million in the fourth quarter of fiscal 2020.
Gross margin was 46.1% compared to 63.3% in the prior year period and 52.5% in the preceding fourth quarter. The changes in gross margin were primarily due to changes in product mix sold in the 3 periods.
Total operating expenses in the first quarter of fiscal 2021 were $8.7 million compared to $8.5 million in the first quarter of fiscal 2020 and $8.4 million in the prior quarter. Research and development expenses were $5.8 million compared to $5.6 million in the prior year period and $5.6 million in the prior quarter. Selling, general and administrative expenses were $2.9 million in the quarter ended June 30, 2020, compared to $2.9 million in the prior year quarter and up from $2.8 million in the previous quarter.
First quarter fiscal 2021 operating loss was $5.7 million compared to $229,000 in the prior year period and $3.9 million in the prior quarter. First quarter fiscal 2021 net loss included interest and other income of $106,000 and a tax provision of $487,000, primarily resulting from the settlement of a tax audit in Israel for fiscal years 2016 through 2019 compared to $147,000 in interest and other income and a tax provision of $43,000 for the same period a year ago. In the preceding fourth quarter, net loss included interest and other income of $148,000 and a tax provision of $65,000.
Total first quarter pre-tax stock-based compensation expense was $755,000 compared to $651,000 in the comparable quarter a year ago and $644,000 in the prior quarter. At June 30, 2020, the company had $64.6 million in cash, cash equivalents and short-term investments and $4.9 million in long-term investments compared to $66.6 million in cash, cash equivalents and short-term investments and $4.1 million in long-term investments at March 31, 2020.
Working capital was $67.3 million as of June 30, 2020, versus $70.9 million at March 31, 2020, with no debt. Stockholders’ equity as of June 30, 2020 was $86.7 million compared to $89.6 million as of the fiscal year ended March 31, 2020.
For the upcoming second quarter of fiscal 2021, our current expectations are net revenues in the range of $6 million to $7.2 million with gross margin of approximately 41% to 43%.
Operator, at this point, we’ll open the call to Q&A.
Thank you. [Operator Instructions] We’ll take our first question from Ari Shusterman with Needham & Company. Please go ahead.
Hey, guys. This is Ari taking the question for [indiscernible]. Thank you for taking my questions. So I first want to ask about like how we can think about Nokia trending for the rest of the year versus defense customers. Thank you.
Yes, so the forecast we’ve seen from them, they should be slightly up this quarter and then trending up again the quarter after. So we had anticipated the 2 quarter down, the March and the June quarter, which happened. And then, the second half was forecasted to trend up, and the forecasts are still holding to that.
Got you. And can you talk about – I know you said the guidance this quarter was weak due to COVID, some of the delays. Can you maybe give some more color on that, like how exactly – I mean, it was especially big, right, lower-than-expected revenue?
So if you look at – we track revenues, obviously, continuously. And when we track revenue this past quarter through May and June, it was very steady from the quarter before. And then – I’m sorry, April and May were very steady. June was down substantially versus March of the prior quarter. And that’s part of the – the backlog was already loaded before all the shelter in place is for April and May.
And so, you could see that it really affected the June quarter. So it’s really the booking. Again, if you look at the shelter in place, a lot of these folks have never worked from home, and I think that’s still a challenge for a lot of folks. And so, we’re seeing some of the slowness as a result of that.
Got you. And just like one more question regarding the Rad-Tolerant product. Can you tell us about how they’ve been trending? If – yeah, if there is any – like anything coming up in the near term? Thank you.
So the Rad-Tolerant is progressing well. We have some opportunities. We’re trying to close. The Rad-Hard has been a little bit more challenging. If you look at the Rad-Hard, they tend to go into more national asset type of applications, so higher security-type applications. And so, what we’re finding with part of the shelter-in-place is that we’re not able to have some of the communications with the customers that we need.
And the reason for that is, since it’s a national asset, those communications have to be secured communications. And there are certain rooms and facilities where you can do that. Well, right now, most of our customers and ourselves aren’t allowed to go into these buildings, and so the conversations we need to have cannot be done over e-mail or over a cell-phone, which is insecure or unsecure I should say.
And so the Rad-Hard is a little bit more challenging in this environment. The Rad-Tolerant is progressing.
Got it. And for gross margin, why there is such a big decline that you’re expecting?
Why did it go down? Can you repeat the question? It didn’t come through clearly.
Okay. So the 41% to 42%…
Yeah, the gross margin – yeah, the gross margin is strictly a function of product mix and also the fact that revenues are down. You’ve got the impact of overhead on gross margin. It’s just lower, higher margin, higher ASP products being shipped in the quarter or forecasted to be shipped in the quarter.
It also impacted the June quarter and we expect it to continue to get back to September quarter. Our long-term outlook with Rad-Hard, Rad-Tolerant and then APU sales and if existing business picks up again, it’s certainly well above 50%. And with the new business, it should be 60% or more. I think it’s just a short-term thing, dependent on the business conditions we’re experiencing right now.
And we’ll go ahead and take our next question from Jeff Bernstein with Cowen. Please go ahead.
Yeah, hi, I just wanted to follow up on the SHREC partnership. Is this the Department of Energy high-performance computing research? I guess, there was $3.3 million in federal funding that was recently allocated to this. Are you getting some of that money or how does this work exactly?
Yeah, so it’s not a grant. What is it? Think of it as more of a group or consortium that we get involved with. And so, it’s actually a membership fee that you have to pay to get into it. And again, what it does is it gives you access to researchers, PhD candidates, professors at these universities. And also, it has folks in there, some government agencies are in it along with prime contractors and then other folks like us in it.
So why it’s important to us is not only are we getting exposed into the environment that we want with these government agencies and prime contractors, but also we have access to these professors and researchers that are doing additional work for us. As we mentioned in the script, they’re looking at our Leda-G board for improvements and enhancements, things like, if you look at the Leda-G board, we have an FPGA on there, so they’re looking at ways to be able to utilize that FPGA, and other applications and other functions, besides what we’re using it for. So it’s just a way to get exposure and also help build the ecosystem for the Gemini program.
[Operator Instructions] And we’ll take our next question from Kurt Caramanidis with Carl M. Hennig, Inc. Please go ahead.
Hi, guys. You said you will get it shipped to early adopters. I’m just wondering if that’s the prototype earlier this year, the Weizmann or if there are other people that are early adopters.
That actually purchased the board?
Yeah, either purchased or whether it’s through Weizmann. And then I think you had a prototype sold in January, February. And then the commentary mentioned, shipped to early adopters. So I’m trying to figure out what that means?
Right. So actually – correct. So they’re actually – the last 2 boards that we shipped out have actually gone to more government agencies. Well, one of them is a prime and another one is a government agency in the U.S. And so – and we’ll probably see that trend continuing, especially with the work we’re doing with SHREC. We should see a few more of these boards shipping. It may not ship this quarter, but it will certainly ship by the end of the year. We should have a handful more boards that have gone out to other entities.
[Operator Instructions] We’ll take a follow-up question from Jeff Bernstein with Cowen. Please go ahead.
Yeah. Just on the Rad-Tolerant, Rad-Hard opportunities. I think you guys have talked about needing to get Rad-Hard parts actually into space to satisfy some people’s desires to know if they’re going to work. Is that true with the Rad-Tolerant as well? And obviously, you can’t have a lot of meetings around Rad-Hard, and you also have to find somebody who’s want to put one of these things in space. How long do you think that ends up taking?
Right. So that – so putting them in space is called heritage. So that heritage is much more of a Rad-Hard issue. The Rad-Tolerant is really a different market space. And the fact that, historically, the folks that have used the Rad-Tolerant have actually kind of gotten themselves, but what I mean is there hasn’t been a Rad-Tolerant market. Folks either offered commercial parts or they offer Rad-Hard parts.
And so what was happening is some of the folks that were required to have a product and didn’t want to pay Rad-Hard prices would take commercial parts and then try to upstream them through testing. And so what we’re doing is we’re just taking that burden away from them and offering them a part instead. So we’re not going to fall under the same heritage, at least in most cases. We’re not going to fall under that same Rad-Hard heritage issue with the Rad-Tolerant.
Great. And then – so I guess, in the pricing differential sort of appeal to people, I hear a lot of the case that you talked about, just using commercial off the shelf. I mean, a lot of the lowest orbit type applications, people – these satellites aren’t up there for all that long, and they’re using commercial off-the-shelf and it works fine. So are you sort of up-selling them on the insurance policy that Rad-Tolerant is going to serve that function? Or are there other market segments that you’re going after?
So there’s other segments. So you’re talking about the LEO satellites. And like you said, some of the LEOs are up there for a while and some of them aren’t. You’re right. Some of these new launches are happening. Satellites are up there for 2 or 3 years, and they’re done. And so folks like that don’t really care. But there are imaging satellites that are LEO, where they’re going to be up there for more than a few years, and so that’s important.
And also, what we’re finding is again, the closer you get to Earth – I’m sorry, to the sun, the more issues you have with all these, call it crazy or what have you. And so what we’re seeing is some of these boards or projects we’re working on, things like probes that are going to Mars or going to some of these asteroids to do mining or what have you, that’s actually going away from the sun. And so they don’t need Rad-Hard, the Rad-Tolerant is sufficient. But these are things that are going to be out there for a while. And so just getting commercial off-the-shelf parts isn’t going to work. They do need to have some insurance that these things have some form of robustness. They cannot take the chance of sending something to Mars or what have you and have it fail within months or a year.
That’s great. Thank you.
I’m seeing that there are no further questions. This concludes today’s question-and-answer session. I would now like to turn the call back over to Mr. Shu for any additional or closing remarks.
Thank you all for joining us. We look forward to speaking with you again when we report our second quarter fiscal 2021 results. Thank you.
Once again, that does conclude today’s conference. Thank you very much for your participation. You may now disconnect your phone lines.