GSI Technology Inc (GSIT) 2025 Q1 法說會逐字稿

Ladies and gentlemen, thank you for standing by.

Welcome to the GSI Technology's first-quarter fiscal 2025 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 involves risks and uncertainties that would 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 have also been asked to advise you that this conference call is being recorded today, July 25, 2024, 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 call over to Mr. Shu.

Please go ahead, sir.

Good day, everyone, and welcome to our first-quarter FY 2025 earnings call.

Let's start with the Q1 results.

Our first-quarter revenue of $4.7 million was within our guidance range despite lower SigmaQuad sales and lower sales to our military prime contractors.

Additionally, the mix of products we saw affect our gross margin causing gross margin to come in at a low end of our guidance range.

Right now, our team is focused on generating new revenue from our Gemini APU to address the expected decline in our legacy SRAM business.

Our main goal this year is to launch a new business line with Gemini-I and secure more SBIR contracts with Gemini-II.

This will help us reverse the revenue decline, achieve financial stability, and ultimately return to growth when we bring Gemini-II online.

Let me explain how we're working toward this goal.

We're preparing to launch our new GXL platform for fast vector search, built on Gemini-I, which will be available as a cloud-based or on-premise solution.

Our groundbreaking technology significantly reduces the time needed to create and update large vector database indexes increasingly used in GenAI and e-commerce, providing a critical advantage for companies that rely on timely data updates.

Didier will provide more details on this.

Additionally, we're expanding our Department of Defense Small Business Innovation Research, or SBIR, proposals pipeline for the Gemini APU.

These contracts can range from $75,000 to over $2 million, typically paid out over 9 to 18 months.

So far, we've announced Phase I and Phase II SBIR contracts, where we're meeting our milestones and are on track to complete them in calendar 2025.

Didier will expand further on these SBIR opportunities in his comments.

We're also on track to begin customer sampling Gemini-II software in early calendar 2025.

The chip has been successfully mounted onto a board, allowing our software team to work with it.

Our goal is to finalize the firmware and complete the software library by December 2024.

We have uncovered minor bugs on the first silicon.

The good news is that there are workarounds for all of those to continue firmware and software development without mass change, thanks to our success in working with the chip at multiple levels.

We are actively seeking a strategic partner to secure resources, funding, and expertise for future development.

To increase our chances of successful engagement, the team is preparing to demonstrate more than 10 times improvement in 1 bit and 2 bit processing with Gemini-II.

We aim to validate this concept by the December quarter with a 1 bit and 2 bit compute-in-memory demo.

Successfully showcasing this breakthrough will enable us to have more productive discussions with potential strategic partners.

As I stated earlier, we are diligently working to launch and monetize the Gemini I and II platforms.

We are preparing to launch the GXL platform to establish a recurring revenue stream in the fast vector search market.

In addition, we're actively writing multiple SBIR proposals to build a pipeline of future revenue streams and offset the remaining development costs of Gemini-II.

With this unique compute-in-memory capabilities, our compute-in-memory APU promises to deliver essential power and time-saving solutions to the rapidly expanding AI market.

Now, I'll hand the call over to Didier, who will discuss our business development and sales activities.

Please go ahead, Didier.

Thank you, Lee-Lean.

I want to expand on two topics that Lee-Lean highlighted, the pending launch of GXL and the building of our SBIR pipeline.

Let’s start with the fast vector search and our new platform for this market called GXL.

First, let me explain the problem GXL solves and why it’s important.

The size of vector basis is growing fast, especially with GenAI and vector similarity search for ecommerce.

Many of these databases use the HNSW algorithm to build indexes because it provides quick and accurate search results.

However, building an HNSW index takes a long time, which is a problem for databases needing fast and accurate information.

One way to fix this is by speeding up the calculations needed to build or refresh the index by running them in parallel.

These calculations take most of the time, so making them faster is crucial.

Our in-compute memory APU technology is designed for this.

It enables massive parallel processing, solving the slow index-building issue.

The APU processes data directly in memory at the bit level, with millions of bit processors working simultaneously.

This means it can handle large amounts of data quickly, reducing the index build time by about 85% compared to traditional CPU-based solutions.

The three main benefits of our GXL solution are scalability, it provides fast access to information for GenAI and ecommerce recommendations, even as these applications grow to billions of vectors.

Second benefit is lower operational costs.

By using computational resources for less time, especially in cloud computing, where costs are billed hourly, it saves money.

And the last benefit is faster updates.

New products can be added or deleted into the index more quickly in ecommerce applications, leading to timely recommendations and increased revenue opportunities.

We recently published a white paper and a blog on medium about efficient HNSW indexing.

Be sure to check those out for more information.

The GXL platform is being prepared for launch, and we aim to begin connecting with the vector database companies that we have identified and then expand to large ecommerce players.

Switching to the build-out of our SBIR pipeline and our focus on landing these Phase I and Phase II contracts from the Department of Defense, or DoD, we view this strategy as a highly efficient and effective way of increasing awareness of GSI’s APU within the DoD.

These SBIR awards validate our statements regarding the APU’s performance and benefits with many large divisions within the DoD and can bring a significant multiplier effect for future opportunities as many of these applications will have dual use for defense and commercial.

This is why we have focused resources on submitting proposals and then prioritizing the execution of these contracts.

We are working on a Phase II SBIR contract with the Air Force worth $1.1 million, which we started in 2023 and are now over 25% complete.

Additionally, we have a Phase II SBIR contract with the Space Development Agency valued at $1.25 million, and we are over 50% complete on this project.

We recently secured a new Phase I SBIR with one of the largest divisions of the DoD, a division that's new to GSI APU's technology.

It has a value of up to $250,000.

We will give more details on this latest win once the contract is signed.

We are currently preparing several other SBIR proposals, adding up to a total pipeline of $6 million.

Our next submission will be for a Phase II SAR space-level on-board contract worth $1.25 million.

In addition to SBIRs, we have initiated the process of seeking additional government funding to enhance the resources we could dedicate to further developing G-II.

This funding would enable us to expand the Gemini-II product line by creating a version for advanced edge AI applications.

On a final note regarding APU opportunities, we sold a Gemini-I Leda card to a major aerospace and defense contractor, who will evaluate this technology for various underclosed applications.

We shipped the card in the June quarter, and the [PO] also provided GSI software training in the September quarter.

Although these sales generate minor revenue, this opportunity is valuable for prospecting engagement with this prominent contractor.

We are excited to see this customer using our technology.

Regarding the two SAR processes we've mentioned in previous calls, one is on hold as we finalize some documentation with that customer.

The other is looking to expand the scope of our services that we can provide in different areas of their operations where they could possibly use the APU.

Our team is actively investigating how to accelerate the delivery of these additional services.

Now, let me switch to the first quarter's customers and product breakdown.

In the first quarter of fiscal 2025, sales to Nokia were $998,000 or 21.4% of net revenues, compared to $1.9 million or 33.5% of net revenues in the same period a year ago, and $694,000 or 13.5% of net revenues in the prior quarter.

Military defense sales were 31.9% of first-quarter shipments compared to 33.8% of shipments in the comparable period a year ago and 35.5% of shipments in the prior quarter.

SigmaQuad sales were 36.3% of first-quarter shipments compared to 58.6% in the first quarter of fiscal 2024 and 42.4% in the prior quarter.

Regarding our SRAM business outlook, while our largest customer has announced some business challenges, our current expectations for orders from them remain stable for the remainder of the calendar year.

Several other SRAM customers are currently managing their inventory levels, and we anticipate they will resume placing orders before the end of the calendar year.

Additionally, we believe that the customers who were previously reduced their use of SRAMs have now stabilized.

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

Doug, go ahead, please.

Thank you, Didier.

The company reported net revenues of $4.7 million for the first quarter of fiscal 2025 compared to $5.6 million for the first quarter of fiscal 2024 and $5.2 million for the fourth quarter of fiscal 2024.

The decline in revenue was due to lower SigmaQuad sales and lower sales for military prime contractors.

Gross margin was 46.3% in the first quarter of fiscal 2025 compared to 54.9% in the year-ago first quarter, the fiscal 2024, and 51.6% in the preceding fourth quarter of fiscal 2024.

The decrease in gross margin in the first-quarter fiscal 2025 was primarily due to product mix and the effect of lower revenue on fixed costs and our cost of revenues.

Total operating expenses in the first quarter of fiscal 2025 were $6.8 million compared to $8.2 million in the first quarter of fiscal 2024 and $7.2 million in the prior quarter.

Research and development expenses were $4.2 million compared to $5.2 million in the prior period and $4.8 million in the prior quarter.

Selling, general and administrative expenses were $2.6 million in the quarter ended June 30, 2024, compared to $3 million in the prior year quarter and $2.4 million in the previous quarter.

First-quarter fiscal 2025 operating loss was $4.7 million compared to an operating loss of $5.1 million in the prior year period and an operating loss of $4.5 million in the prior quarter.

First-quarter fiscal 2025 net income included interest and other income of $55,000 and a tax provision of $57,000 compared to $80,000 in interest and other income and a tax provision of $51,000 for the same period a year ago.

In the preceding fourth quarter, net loss included other income of $108,000 and a tax benefit of $85,000.

Net income in the first quarter of fiscal 2025 was $1.1 million, or $0.04 per diluted share, reflecting a one-time gain of $5.7 million on the sale and leaseback transaction related to sale of the company's headquarters compared to a net loss of $5.1 million or $0.21 per diluted share for the first quarter of fiscal 2024 and a net loss of $4.3 million or $0.17 per diluted share for the fourth quarter of fiscal 2024.

Total first-quarter pre-tax stock-based compensation expense was $658,000 compared to $820,000 in the comparable quarter a year ago and $693,000 in the prior quarter.

At June 30, 2024, the company had $21.8 million of cash and cash equivalents compared to $14.4 million at March 31, 2024.

Working capital was $25.7 million as of June 30, 2024, compared to $19.1 million at March 31, 2024.

Stockholders' equity as of June 30, 2024, was $38 million compared to $36 million as of the fiscal year ended March 31, 2024.

During our fourth-quarter earnings call in May, we announced that the company had initiated a comprehensive strategic review and established a special committee of the Board to evaluate various strategic alternatives.

We have engaged Needham and Company as our strategic and financial advisor to assist in this process.

Currently, we are in discussions regarding potential strategic opportunities and are encouraged by the direction of these preliminary discussions.

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

(Operator Instructions) Luke Bone.

Hi.

I'm wondering if you all could do a brief -- kind of skim over the cost benefit and contemplations about potential leasing or sale or going this more diverse route, which it seems like you're laying out in this call, as well as a potential for a key partnership.

If you could give a succinct current refinement of all those prospects and your leading direction?

What exactly are you asking?

Are you asking about how we decided to sell the building?

Oh no, not the building.

I'm talking -- referring to the APU or Gemini APU project.

And since the last call -- prior calls here, we're building up to hopefully a decision on a route for funding and development to progress into Gemini-III.

I just wondered if you had kind of a way of like a succinct way to describing what your current leading prospects are for either the major funding or leasing of that technology?

Yes.

We have still continued discussion on Gemini-III with hyperscaler and HBM vendors.

But the key is really we need to demonstrate our GPU, APU capability on the 1 bit and 2 bit processing.

That's why we are focused doing it, and we want to prove it on the G-II and by the end of this year, the calendar year.

Yes.

So there's definitely interest, but they'd like to see the capabilities of the parts so where we can show them some benchmarks.

So we're working on that now.

Michael Cooper.

Hi, guys.

Can you tell me, with these benchmarks, have you validated these benchmarks in house and now you have to formalize the process for outside clients and then start showing them?

Do you know the answers to these questions that are coming down the pipe?

Yeah.

We have a calculation, but it's just a calculation line.

And right now, we are trying to get the G-II working, so we can run the actual device.

That's what we are doing basically right now.

Yeah.

We're actually working with the first silicon, and the first silicon has turned out to be a very solid for first silicon.

I think as we've mentioned before, there are no major bugs, nothing that we can't work around.

We're right now accessing the part.

They've been mounted on board and trying to gather as much data and information as we can, working with what we have.

Right.

And in order to be able to run those benchmarks, we just need all of the algorithms and the libraries on the software side complete.

On the hardware side, as Lee-Lean said in his section, there have been no bugs that are showstopper at this point.

And so it's just getting the libraries completed so we can run those benchmarks.

And in those libraries and the algorithms, there's less critical pinch points, if you will?

Well, it's just different, right?

With the hardware, you have a lot of moving parts and you have a lot of potential bugs; with the software and the libraries, it's just time.

It's just the folks that's being done by our Israeli facility, and it's just a matter of time to complete those libraries.

And if I could just extend one more question about, okay, let's say you achieve all this.

What does the market look like?

How large potentially within your channels?

And what's the gross margin?

And is it like the hyperscalers might buy hundreds or thousands of a $30,000 system?

Or what does that

-- [should look like]?

Sure.

So Gemini-II is a little different going to market than Gemini-I.

So Gemini-I, if you recall, we have some cloud-based usage models, and then we also sold some board/systems.

Gemini-II is a little different, I shouldn't say a little different.

We can expand to that, some of the models we're looking at will be cloud-based usage kind of revenue.

We will be selling some boards and some systems, but we also have, which is very key to some of the markets we're looking at, the opportunity to sell actual chips.

So with Gemini-I, a chip sale was not viable.

And we've talked about the dependence on an FPGA that was on the original Leda board with Gemini-I.

With Gemini-II, there's not that restriction so we can sell the chip as well.

And so depending on the customer, they may want to buy a chip, they may want to buy a board, they may want to buy a server that has multiple boards.

And then in some of our revenue streams, it will be, like I said, the cloud-based kind of revenue.

Yeah.

The Gemini-II is targeting at the Edge AI processing market.

We are not targeting it at a cloud basis.

But anyway, the Edge AI market, [the research report] -- the whole Edge AI computing is like $20 billion market right now in 2024 and growing at mid-20, to mid-30 on the depending on who is publishing the data.

I think it's a [fast-growing] market, and that's where the market we are targeting.

Greg Greenberg.

Yeah.

Can you give us an update on the radiation hardened programs or anything that's happened in the last quarter and what you see for one, two, three years from now?

Thank you.

Sure.

So we shipped a prototype order this past quarter to a customer actually out of Europe.

And so that's going to be further ongoing testing.

There's still some active programs that we have had some samples into over the last couple of years.

And so those are just waiting to go into production.

And then there are a few others that we're still targeting.

Hard to say.

We've always said that we kind of are looking for the kind of the $10 million a year kind of run rate.

And again, we talked about this as north of 90% gross margin on this product.

So we're still saying that in the years ahead.

It's just been -- as you know, if you've been on these calls in the past, it's just been a much slower process than we had ever anticipated.

And I think you mentioned, maybe a few quarters ago, that you didn't need it to go up in space.

Or has it been already up in space?

What's the status there?

So, yeah, you're referring to heritage.

Yeah, as far as I know, we don't have heritage yet.

And so we're waiting for some of these launches.

I know that we're also seeing what we can do about possibly getting some parts in the space just for the heritage only.

It would not be -- this is going to sound a little strange.

It's not associated with the program we're working on now.

It would just be a way to exercise the parts in space just to prove that they actually work in those environments.

All the testing we've done, which simulates that environment, is done on earth.

All the tests that are done tells you it will work, but these folks want actual proof.

So we're looking for a mechanism now to try and get the parts up so we can have that official heritage stamp on our parts.

Okay.

And then back to APU.

So you're still expecting a second spin.

When would you expect that back?

Yeah, so we're not actively spinning the part yet.

So as we discussed, first silicon looks very solid.

There are some minor bugs.

They've all had software workarounds at this point.

And so as we continue to finish up these libraries and algorithms we talked about, we'll be able to determine whether or not there are any other bugs.

And once that's done, then we'll do the spin.

So as of now, we're holding off on any spin until we can identify any other minor bugs.

So I can read into that strategic alternatives doesn't require a partner to be able to see a second spin to know what you guys have so far, correct?

Correct.

Yeah, the strategic partnership is more for the next generation part and also just to expand our business in general, but we won't need that just for the second spin.

That's correct.

Robert Christian.

Yeah, have you guys identified any possible companies that are interested in licensing your technology?

So we are actually working with right now a -- so this is on the customer side, not on -- I'm not sure if you're talking about potential competitors or customers.

So on the customer side, we do have a major networking telecom customer that is looking at our technology to be able to incorporate some of the IP into their ASIC.

And those are conversations we've been having for some time now, and it looks like it's getting further along.

We actually started the conversations before we had Gemini-II.

And obviously, Gemini-I wasn't what they're looking for; it had to be Gemini-II technology.

Now that it's out and once we get some of the benchmarks, then that will progress even faster.

Okay, good.

And is there any other companies out there, competitors that's interested in buying the company as it is right now?

Well, we have been having a lot of discussions with Needham and things are encouraging, but there's nothing to talk about yet.

It appears that there are no further questions at this time.

I would now like to turn the floor back over to Mr. Shu for closing comments.

Thank you all for joining us.

We look forward to speaking with you again when we report our second-quarter fiscal 2025 results.

Thank you.

This concludes today's teleconference.

You may disconnect your lines at this time.

Thank you for your participation. (Operator Instructions)