GSI Know-how, Inc. (GSIT) Q3 2024 Earnings Name Transcript

GSI Know-how, Inc. (NASDAQ:GSIT) Q3 2024 Earnings Convention Name January 25, 2024 4:30 PM ET

Firm Members

Lee-Lean Shu – Chairman, President and Chief Government Officer

Didier Lasserre – Vice President, Gross sales

Douglas Schirle – Chief Monetary Officer

Convention Name Members

Brett Reiss – Janney Montgomery Scott

Jeff Bernstein – Silverberg Bernstein Capital

Operator

Girls and gents, thanks for standing by. Welcome to GSI Know-how’s Third Quarter Fiscal 2024 Monetary Outcomes Convention Name. Right now, all individuals are in a listen-only mode. Later, we’ll conduct a question-and-answer session. At the moment, we’ll present directions for these excited about getting into the queue for the Q&A.

Earlier than we start in the present day’s name, the corporate has requested that I learn the next protected harbor assertion. The issues mentioned on this convention name might embody forward-looking statements concerning future occasions and the longer term efficiency of GSI Know-how that contain dangers and uncertainties that would trigger precise outcomes to vary materially from these anticipated. These dangers and uncertainties are described within the firm’s Type 10-Ok filed with the Securities and Trade Fee.

Moreover, I have been requested to advise you that this convention name is being recorded in the present day, January 25, 2024, on the request of GSI Know-how.

Internet hosting the decision in the present day is Lee-Lean Shu, the corporate’s Chairman, President and Chief Government Officer. With him are Douglas Schirle, Chief Monetary Officer; and Didier Lasserre, Vice President of Gross sales.

I might now like to show the convention over to Mr. Shu. Please go forward, sir.

Lee-Lean Shu

Good afternoon, and thanks for becoming a member of us.

I’m happy to share a number of key updates from an eventful third quarter. Beginning with the product growth, we achieved two main milestones that may maintain us on observe to advance the Gemini APU household.

First, in November, we efficiently accomplished the radiation-hardened testing on the Gemini-I APU for compute-in-space purposes. The check outcomes affirm that Gemini-I has met traits to be a radiation-tolerant processor. We’re actively engaged with a number of satellite tv for pc corporations that want radiation-tolerant APUs, and we’re inspired by the robust curiosity.

Transferring to our next-generation APU, we accomplished the desk of Gemini-II within the third quarter. In late February, we’ll consider the preliminary spin and anticipate to start sampling Gemini-II chip within the second half of calendar 2024. Gemini-II has 8 instances of inner reminiscence and 10 instances higher efficiency than Gemini-I and notably decrease value. This dramatically enhance in value efficiency permit us to focus on a wider vary of purposes.

The 96-megabyte of inner reminiscence in Gemini-II can match many AI fashions solely on the chip, enabling in-place information processing with out accessing exterior reminiscence, exterior [DRAM] (ph). This could give Gemini-II great benefit on compact edge purposes like drone and ADAS. We anticipate beginning preliminary alpha deployment with choose clients within the second half of calendar 2024.

Lastly, within the third quarter, we shipped radiation-hardened SRAMs to 2 clients for 2 new packages and we acquired a second SBIR Direct-to-Part II contract within the quantity of $1.1 million.

Trying forward, we’re within the early stage of growing the structure for our next-generation Gemini-III chip. We had a dialogue with a number of hyperscalers about APU design choices to greatest deal with the rising markets’ want within the information heart. Along with serving to hyperscalers decrease information heart energy consumption, the APU also can present vital profit to GenAI finish customers by decreasing inference prices. We initiated preliminary dialogue with two main hyperscalers who’ve expressed curiosity in our expertise and proceed to work with inner groups on varied early-stage ideas.

Turning to our monetary efficiency, the third quarter income of $5.3 million met our steering, with a gross margin of 56% was on the midpoint of our steering. The sequential enchancment in gross margin displays product combine this quarter.

Our working bills, which elevated each year-over-year and sequentially, included a one-time expense of $2.4 million for pre-production masks for Gemini-II. We view this as an funding in future development as Gemini-II will tremendously develop APU market attain.

In parallel, we proceed dialogue with potential strategic companions to assist Gemini-II launch and help with Gemini-III growth. As our most bold chip in the present day, we anticipate that Gemini-III would require vital capital funding and partnership to assist mitigate our funding wants. We stay targeted on meals and monetary administration to gas our product roadmap and growth into new markets.

Not too long ago, Cornell College revealed a analysis paper, spotlighting the distinctive effectivity of our Gemini-I APU to accelerating location filtering in DNA mapping. Didier will talk about the element of the paper and its implications for GSI in his feedback.

In closing, it was an eventful third quarter with great progress on our Gemini roadmap, prudent expense administration and motion to extend our monetary flexibility to assist our development. We stay laser-focus on bringing our progressive APUs to market and driving long-term worth creation.

Now, I’ll hand the decision over to Didier, who will talk about our enterprise efficiency additional. Please go forward, Didier.

Didier Lasserre

Thanks, Lee-Lean.

I need to present some extra context on why we’re so assured out there potential for our APU structure, particularly for inference workloads.

First, the unequalled flexibility of our variable bit processing is essential. With 2 million undefined bit processors that may be toggled from 1 bit to 2 million bits cycle by cycle, our APU can adapt to actual time — I am sorry, in actual time to maximise effectivity. This dynamic bit-wise configurability can course of lengthy strips 1 bit at a time and is good for inference since analysis exhibits that completely different bit precisions are extra environment friendly for various fashions.

Second, our APU structure breaks the Von Neumann mannequin by eradicating the information fetch perform. This progressive design delivers larger efficiency with decrease energy consumption. As Lee-Lean talked about, these capabilities immediately deal with the important wants of information facilities and rising purposes by reducing information heart energy consumption and decreasing inference prices for GenAI finish customers.

Importantly, I need to emphasize that our APU represents true in-memory — I am sorry, true compute-in-memory structure. In contrast to competing chips that declare compute-in-memory, they’re really close to reminiscence compute, and our APU has logic bodily built-in within the reminiscence. This basic distinction in structure will in the end allow our APU to realize the transformative velocity and effectivity features we anticipate as we scale. Our true compute-in-memory structure offers us a sustained aggressive benefit.

To speed up ecosystem growth, we’re specializing in getting APU within the arms of key companions within the navy, hyperscalers and academia. The true-world deployment and libraries will showcase the advantages, develop use circumstances, and assist our go-to-market capabilities.

One instance of the technique helps us promote and monetize Gemini-I is a not too long ago revealed analysis paper from Cornell College. We’re happy to announce that the Cornell paper demonstrates our APU one — I am sorry, Gemini-I APU’s distinctive efficiency advantages for genomic purposes. Leveraging the APU’s massively parallel in-memory structure, Cornell researchers confirmed as much as 6 instances quicker DNA sequencing filtering in comparison with a 16-core CPU. This showcases our expertise’s benefit for data-intensive workloads requiring fast, low-precision comparisons.

The research additionally revealed robust potential to speed up different purposes with related data-matching wants, together with medical information evaluation, search, safety, and extra. With easy scaling, our APU will be packed into cost-effective high-density servers to multiply this efficiency for real-world deployments that may decrease energy budgets for hyperscalers in comparison with GPU options.

These outcomes implement — I am sorry reinforce our vital market alternatives throughout sectors that depend on effectively discovering patterns and similarities inside huge information units. We stay targeted on delivering the game-changing in-memory compute efficiency to clients throughout a number of industries.

As Lee-Lean talked about, we anticipate receiving first silicon gadgets of Gemini-II in February. After preliminary analysis and debugging, we’ll goal a second spin this summer season and provoke benchmarking shortly after. Our $2.3 million in SBIR funding will assist this growth. As a reminder, this contains our not too long ago introduced second SBIR Direct-to-Part II $1.1 million contract to create specialised algorithms for the U.S. Air Drive Analysis Laboratory. The goal purposes embody in-craft purposes resembling search and rescue, object detection, transferring goal indication, change detection, and SSIM in GPS-absent conditions. GSI may even develop algorithms utilizing information from the U.S. Area Drive to showcase the efficiency advantages of its compute-in-memory APU2 built-in circuit.

In abstract, the flexibility of structure, hands-on buyer engagements, and ecosystem partnerships offers us confidence in our market alternative. Now we have a strong product roadmap to ship steady improvements that we are able to — that we imagine will capitalize APU adoption throughout a number of industries within the coming years.

Let me swap now to buyer and product breakdown for the third quarter. Within the third quarter of fiscal 2024, gross sales to Nokia have been $807,000 or 15.2% of internet revenues in comparison with $1.3 million or 20% of internet revenues in the identical interval a 12 months in the past, and $1.2 million or 20.3% of revenues within the prior quarter. Army/protection gross sales have been 28.2% of third quarter shipments in comparison with 26.2% of shipments within the comparable interval a 12 months in the past, and 34.8% of shipments within the prior quarter. SigmaQuad gross sales have been 46.9% of third quarter shipments in comparison with 45.2% within the third quarter of fiscal 2023, and 55.8% within the prior quarter.

On one final notice on product gross sales within the third quarter, we shipped over $600,000 of a prototype radiation-hardened SRAM to 2 completely different clients. These will probably be deployed in two separate satellite tv for pc packages.

I would like at hand the decision over to Doug. Doug, go forward, please.

Douglas Schirle

Thanks, Didier.

We reported a internet lack of $6.6 million, or $0.26 per diluted share, on internet revenues of $5.3 million for the third quarter of fiscal 2024, in comparison with internet losses of $4.8 million, or $0.20 per diluted share, on internet revenues of $6.4 million for the third quarter of fiscal 2023 and a internet lack of $4.1 million, or $0.16 per diluted share, on internet revenues of $5.7 million within the second quarter of fiscal 2024.

Gross margin was 55.9% in comparison with 57.5% within the prior-year interval and 54.7% within the previous second quarter. The modifications in gross margin have been primarily resulting from modifications in product combine and quantity offered within the three intervals.

Complete working bills within the third quarter of fiscal 2024 have been $9.7 million in comparison with $8.5 million within the third quarter of fiscal 2023 and $7.2 million within the prior quarter. Analysis and growth bills have been $7 million in comparison with $5.5 million within the prior-year interval and $4.7 million within the prior quarter. Promoting, basic and administrative bills have been $2.7 million within the quarter ended December 31, 2023 in comparison with $3 million within the prior-year quarter and $2.5 million within the earlier quarter.

Third quarter fiscal 2024 working loss was $6.7 million in comparison with $4.8 million within the prior-year interval and an working lack of $4.1 million within the prior quarter.

Third quarter fiscal 2024 internet loss included internet curiosity and different revenue of $155,000 and a tax provision of $71,000, in comparison with internet curiosity and different revenue of $61,000 and a tax provision of $84,000 for a similar interval a 12 months in the past. Within the previous second quarter, internet loss included internet curiosity and different revenue of $71,000 and a tax provision of $33,000.

Complete third quarter pre-tax stock-based compensation expense was $649,000 in comparison with $655,000 within the comparable quarter a 12 months in the past and $676,000 within the prior quarter.

At December 31, 2023, we had $21.6 million in money and money equivalents, in comparison with $30.6 million in money, money equivalents and short-term investments at March 31, 2023. Working capital was $23.1 million as of December 31, 2023 versus $34.7 million at March 31, 2023, with no debt. Stockholders’ fairness as of December 31 was $39.6 million in comparison with $51.4 million as of the fiscal 12 months ended March 31, 2023.

For the fourth quarter of fiscal 2024, we anticipate internet revenues in a variety of $4.8 million to $5.4 million, with gross margin of roughly 55% to 57%.

Operator, at this level, we would prefer to open the decision to Q&A.

Query-and-Reply Session

Operator

Thanks. We’ll now be conducting a question-and-answer session. [Operator Instructions] Our first query is from Brett Reiss with Janney Montgomery Scott. Please proceed together with your query.

Brett Reiss

Hello, gents. You’ll be able to speak to me like I am six-years-old. May you simply clarify what’s an rare $2.4 million cost for a pre-production masks? I am not an engineer. Simply in case you can provide me some extra readability on that?

Douglas Schirle

Yeah. So, each time now we have a product, now we have to have a masks set ready to run within the fab to fabricate the wafers. Now, usually, after we incur prices for masks set for manufacturing product, we’ll capitalize that into prepaids and amortize it over a 12-month interval. Nonetheless, the one exception to that’s when now we have masks set ready on a brand new course of expertise that we have by no means run earlier than, we’ll cost that expense to R&D expense. So, on this quarter — properly final quarter, we taped out Gemini-II on a course of expertise we have by no means used. It is a 16-nanometer course of at TSMC. Now, since we have by no means used that course of earlier than, we cost that $2.4 million to R&D expense.

Brett Reiss

All proper. So, that is one shot. It isn’t going to be reoccurring?

Douglas Schirle

No. It can recur sooner or later when now we have one other product that tapes out on a course of expertise that we have by no means used earlier than. So, it is rare. It does not occur yearly. It is at most each two or three years.

Brett Reiss

Okay. The money, which was my margin of security in my funding right here, retains dwindling down. I am a bit involved about that. A few of these initiatives with bringing in different three way partnership companions, what is the timetable on that? And the way lengthy do now we have earlier than we burn by means of the remaining money?

Douglas Schirle

Nicely, we’re at present taking a look at varied alternatives, and Didier talked about speaking to hyperscalers and others. To return out with our subsequent product, it is going to require vital investments, so we’re searching for companions or different sources of funding. Along with that, now we have a constructing that is — we personal that is price fairly a bit of cash, and we’ll be trying into probably promoting that constructing within the close to future.

Brett Reiss

Proper. What’s the constructing appraised at? And in a base case situation, what do you assume you possibly can promote it for?

Douglas Schirle

We predict we are able to most likely get someplace within the vary of $10 million to $13 million.

Brett Reiss

Proper.

Douglas Schirle

And there isn’t any debt tied to that. It is totally paid for.

Brett Reiss

Proper. May you simply give me a broad define of what the construction of a three way partnership accomplice’s capital funding in GSIT would take? Would they pay an upfront milestone cost with different funds to observe? Would it not be an fairness funding? What do you assume the construction of that may seem like?

Didier Lasserre

So, it might be both of these. What I imply is when Lee-Lean was speaking a couple of accomplice for funding the subsequent program, that is particularly for Gemini-III. So, Gemini-II was totally funded internally. However for Gemini-III, we’re searching for a accomplice for that, almost definitely a buyer funding accomplice. However except for that, that may be extra of an NRE kind of funding, so there can be milestones related to that. However except for that, we’re additionally open to fairness investments within the firm as properly.

Brett Reiss

Okay. I am going to drop again in queue. Thanks for taking my questions.

Douglas Schirle

Thanks.

Didier Lasserre

Thanks, Brett.

Operator

[Operator Instructions] Our subsequent query is from Jeff Bernstein with Silverberg Bernstein Capital. Please proceed together with your query.

Jeff Bernstein

Yeah, hello. Good afternoon. So, only a query on the — and congratulations on inserting these rad-hard SRAM components for analysis. If you happen to have been to win these packages, about how a lot income and over what time interval would possibly you be capable to acquire from these two satellite tv for pc packages?

Didier Lasserre

Certain. So one of many packages was — and once more, that is only a prototype amount for demonstration functions. Certainly one of them was simply over $500,000 and the opposite one was about $150,000, and it totaled 41 components that we shipped. You are able to do fast calculation with the ASPs are on these. And so, these are simply the, once more, prototype portions. So, clearly, you’ll be able to multiply by one thing. We do not have the portions but. These are packages that they are seeking to launch throughout the subsequent 12 months. And so, it might be someday second half of 2025 on the soonest earlier than they launch manufacturing. However definitely it might be north — if the prototype is $500,000, you’ll be able to think about what a manufacturing is likely to be.

Jeff Bernstein

And so, these can be like GEO satellites. So, these aren’t satellite tv for pc networks. These are going to be particular person bigger satellites?

Didier Lasserre

These ones are X, one is GEO, one is definitely LEO.

Jeff Bernstein

Okay. And can you get an computerized with these analysis components? Are you going to get a experience into house on considered one of these and truly get form of provenance from that, or are we nonetheless seeking to get that some other place?

Didier Lasserre

No. So, definitely, now we have — as we have spoken prior to now, now we have different prototype gadgets we have shipped out already within the final couple of years. And so, it might be any of the packages we have mentioned prior to now or these two. Certainly one of these seems like is pretty accelerated of their timetable. So, there’s an opportunity one of many ones we simply shipped this final quarter may stand up pretty shortly.

Jeff Bernstein

Okay. After which, simply on Nokia lowest income from them in perpetually, what is the story on Nokia and the outlook for that router that you just promote into?

Didier Lasserre

So, they’re contract producers, as a result of we ship these components to 2 separate contract producers. And so they each had a bit little bit of stock, so that they have been burning by means of some stock. So far as the — we get, I believe we have talked prior to now, a 12-month rolling forecast from Nokia, and people nonetheless are coming in across the run fee we have been seeing for the final couple quarters. So, just a bit bit north of what we did this previous quarter, someplace in $1 million, $1.1 million form of vary 1 / 4.

Jeff Bernstein

Okay. And at Needham, I believe you additionally talked about along with searching for a accomplice, probably a monetary accomplice, probably a growth accomplice, probably a hybrid of each of these for Gemini-III. I believe you additionally talked about one thing about probably licensing IP. It feels like for what would most likely be an edge form of case for semiconductor IP for doing in-memory processing. Are you able to simply discuss that? Would that be form of an upfront license after which royalty stream, or simply give some colour round that?

Lee-Lean Shu

Yeah. The IP — properly, we’re extra targeted on the Gemini-III. So, it might be IP or it might be the product growth for the client. So, principally, I believe now we have — as we talked about within the convention name, I imply we made fairly good progress over the quarter. So — yeah, so we’re nonetheless engaged on it.

Operator

Thanks. There aren’t any additional questions presently. I would like at hand the decision again to Lee-Lean Shu for any closing feedback.

Lee-Lean Shu

Thanks all for becoming a member of us. We look ahead to talking with you once more after we report our fourth quarter and the total 12 months fiscal 2024 outcomes.

Operator

This concludes in the present day’s convention. You might disconnect your traces presently. Thanks in your participation.