Sign in

You're signed outSign in or to get full access.

GSI - Earnings Call - Q3 2025

January 30, 2025

Executive Summary

  • Q3 FY2025 revenue was $5.414M (+2% y/y; +19% q/q) and gross margin rebounded to 54.0% from 38.6% in Q2, driven by higher revenue, mix, and absence of severance costs that impacted the prior quarter; operating loss improved to $(4.055)M and EPS was $(0.16).
  • Management initiated Q4 FY2025 guidance: revenue $5.4–$6.2M and gross margin 55–57%, implying further sequential margin expansion as SRAM demand normalizes and cost actions flow through.
  • Mix metrics show continued SRAM recovery: SigmaQuad at 39.1% of shipments (vs 38.6% in Q2) and Nokia at 4.4% of revenue (down from 17.8% in Q2), while military/defense shipments were 30.0%; management also cited a key SRAM customer tied to AI chip manufacturing as a 2025 growth driver.
  • APU roadmap advanced: Gemini-II first silicon is functional with software workarounds; second spin planned for mass production; PLATO (LLM-focused, low-power) launched as next-gen with 12–18 month timeline; SBIR activity expanded with a U.S. Army Phase 1 award up to $250k and ongoing milestones with AF/RL and SDA.
  • Street consensus via S&P Global could not be retrieved at time of analysis; beat/miss vs Wall Street estimates cannot be assessed (S&P Global data unavailable).

What Went Well and What Went Wrong

  • What Went Well

    • Gross margin recovery to 54.0% (from 38.6% in Q2) on higher revenue, improved mix, and absence of prior-quarter severance costs; operating loss narrowed to $(4.055)M and EPS improved to $(0.16).
    • Strengthening SRAM trajectory with normalized inventories and rising demand from a key customer integral to leading AI chip manufacturing: “We anticipate this customer to become our largest revenue contributor in fiscal 2025.” — CEO Lee-Lean Shu.
    • APU milestones: Gemini-II functional first silicon (software workarounds), on-track February tape-out and May availability; PLATO for edge LLMs launched with a faster, lower-cost path versus prior 3D plan.
  • What Went Wrong

    • Revenue remains low on an absolute basis ($5.414M) and EPS is still negative (GAAP), highlighting ongoing scale challenges and reliance on legacy SRAM while APU ramps.
    • Nokia revenue dropped to $239k (4.4% of net revenues) from $812k (17.8%) in Q2, and military/defense shipments retreated sequentially to 30.0% (press release), reflecting mix and customer-specific variability.
    • Transcript vs press release discrepancy: management remarks referenced military/defense shipments as 25% on the call vs 30.0% in the press release; we anchor to the press release for official KPI and note the variance for caution.

Transcript

Operator (participant)

Welcome to GSI Technology's Q3 Fiscal 2025 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 the queue for the Q&A. 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 have also been asked to advise you that this conference call is being recorded today, January 30, 2025, at the request of GSI Technology.

Lee-Lean Shu, the company's Chairman, President, and Chief Executive Officer, will be hosting the call today. 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.

Lee-Lean Shu (CEO)

Good afternoon, and thank you for joining us to review our Q3 Fiscal 2025 Financial Results. I am pleased to report that our Q3 demonstrated meaningful progress in our financial performance and the ongoing initiative with the APU. Let me start with our financial highlights before diving into our product development. In the Q3, we generated revenue of $5.4 million, representing a 2% increase year-over-year and a 19% sequential growth from the previous quarter. More importantly, our cost reduction initiative had begun to bear fruit, resulting in a reduction in operating and net losses compared to Q3 of fiscal 2024. Looking at our aerospace business, I am encouraged by the continued momentum in our core aerospace segment.

Revenue growth this quarter in SRAM came from existing customers resuming orders at their inventory levels low moments, and the growing SRAM demand from our customer for our 144Mb Rad-Hard SRAM system are integral to manufacturing leading AI chips. We anticipate this customer becoming our largest SRAM customer in fiscal year 2025. Let me share a few key developments on the APU front. So far, our APU technology is progressing steadily. We made good progress using the Gemini 2 first silicon, which had a minor bug. Fortunately, software workarounds made the chips sufficiently functional to put on the board for testing and software and library development. We will be able to deliver neural network models for our Gemini 2 SBIR program with the Air Force in the March quarter based on this first silicon. We are now preparing for a second spin of Gemini 2 for mass production.

Didier will provide additional details on Gemini 2 in his prepared comments. In parallel, we have begun work on a new chip, Plato. Plato will be designed to address the growing market for large language models in AI application as a solution with its low-power features. The chip architecture will be built on Gemini 2's architecture. We are pursuing multiple funding pathways to support this $50 million development program, including discussion with strategic and financial partners. Our government business continues to expand to multiple SBIR programs. We recently announced a phase I contract with the U.S. Army, securing $250,000. In this phase, we will explore and identify best-fit applications to develop specialized edge computing AI solutions with Gemini 2. This will strengthen our relationship with agencies within the Department of Defense. Didier will share more details on this new win and the status of our SAR projects.

In summary, while we are encouraged by our progress, we remain dedicated to disciplined execution across all areas. Our improved operating efficiency, growing Rad-Hard sales, and strategic initiative in commercial and government markets position us well for sustainable growth as we continue to evaluate strategic alternatives. We start handing over to Didier to provide more details on our APU business opportunities, then cover the Q3 sales breakdown.

Didier Lasserre (VP of Sales)

Thank you, Lee-Lean. As Lee-Lean mentioned, Gemini 2 is on track for a February tape-out and availability in May, and we'll have a new leaderboard in June. This aligns with the milestone for the Air Force's SBIR. Gemini 2 enhances AI capabilities with neural network models together with SAR imaging capability to target high-value applications in the defense and aerospace markets. Meeting these milestones should put us in good shape to complete algorithm development and library building for delivery by Q3 of calendar 2025. Leveraging Gemini 2's architecture, we can accelerate the development of our next-generation chip, Plato, with a cost-effective fast-to-market strategy. Plato's ultra-low-power design will target rapidly growing markets for the edge and large language model solutions.

Plato is a strategic pivot from our previous plan for developing a new next-generation version of Gemini with a more economical and faster development path than the previously proposed 3D Gemini 3. The distinguishing feature of Plato targeted for the LLM market is that it can deliver data center performance levels using very low power, making it suitable for use at the edge. As Lee-Lean Shu mentioned, we are in discussions with potential customers who could be initial users and may also be funding sources with an expected development timeline of 12 to 18 months. I am pleased to say that our ongoing SBIR projects are on track and meeting our milestones. As a reminder, we are currently working on phase II contracts with Space Development Agency and Air Force Research Laboratory and most recently announced a phase I contract with the U.S. Army.

We are preparing to deliver a YOLO algorithm for the Air Force Research Laboratory in the current quarter, including the benchmarks for a real-time object detection application. As a reminder, YOLO, which stands for you only look once, are algorithms that immediately determine the exact placement of objects in an image by drawing boxes around them and identifying the types of objects. Our second SBIR with the Space Development Agency is waiting for the delivery of a Gemini 2 board. As I mentioned earlier, we are on track to deliver this within the required timeframe. Today, we have received 45% of the $1.25 million contract and will receive the balance upon the board's delivery. We are now executing the phase I for our latest win, the U.S. Army contract, to determine which of their ongoing projects would benefit from Gemini 2's capabilities.

After phase I, we will have the opportunity to compete for phase II, which could be worth up to $2 million. The focus will be on edge computing AI solutions using Gemini 2. Our key objectives are integration, feasibility assessment, and AI algorithm validation for military applications. The Army is the largest of the five branches of the Department of Defense, and with this SBIR, we are gaining exposure to this branch. Lastly, regarding our SAR projects, there has been ongoing progress in fiscal Q3. The Asian Defense and R&D organization that we have mentioned in the past is evaluating Gemini 2 for low-power in-flight applications. We are advancing with a few more deliverables to prepare them to purchase a Gemini 2 board. This organization is also a potential funding partner for Plato.

Also, two other aerospace companies are evaluating our technology for image creation and onboard satellite applications for SAR. Now I'll move to the customer and product breakdowns for Q3. In Q3 of fiscal 2025, sales to Nokia were $239,000, or 4.4% of net revenues, compared to $807,000, or 15.2% of net revenues in the same period a year ago, and $812,000, or 17.8% of net revenues in the prior quarter. Military defense sales were 25% of our Q3 shipments compared to 28.2% of shipments in the comparable period a year ago and 40.2% of shipments in the prior quarter. SigmaQuad sales were 39.1% of Q3 shipments compared to 46.9% in Q3 of fiscal 2024 and 38.6% in the prior quarter. Now I'd like to hand the call over to Doug. Go ahead, Doug.

Douglas Schirle (CFO)

We reported net revenues of $5.4 million for Q3 of fiscal 2025, compared to $5.3 million for Q3 of fiscal 2024 and $4.6 million for Q2 of fiscal 2025. Gross margin was 54% in Q3 of fiscal 2025, compared to 55.9% in Q3 of fiscal 2024 and 38.6% in the preceding Q2 of fiscal 2025. The sequential increase in gross margin in Q3 of fiscal 2025 was primarily due to higher revenue, product mix, and severance costs associated with manufacturing workforce reductions in the prior quarter. Total operating expenses in Q3 of fiscal 2025 were $7 million, compared to $9.7 million in Q3 of fiscal 2024 and $7.3 million in the prior quarter.

Research and development expenses were $4 million, compared to $7 million in the prior year period and $4.8 million in the prior quarter. Selling, general and administrative expenses were $3 million in the quarter ended 31 December , 2024, compared to $2.7 million in the prior year quarter and $2.6 million in the previous quarter. Q3 fiscal 2025 operating loss was $4.1 million, compared to an operating loss of $6.7 million in the prior year period and $5.6 million in the prior quarter. Q3 fiscal 2025 net loss included interest and other income of $70,000 and a tax provision of $44,000, compared to $155,000 in interest and other income and a tax provision of $71,000 for the same period a year ago. In the preceding Q2, net loss included interest and other income of $149,000 and a tax provision of $23,000.

Net loss in Q3 of fiscal 2025 was $4 million or $0.16 per diluted share, compared to a net loss of $6.6 million or $0.26 per diluted share for Q3 of fiscal 2024, and a net loss of $5.5 million or $0.21 per diluted share for Q2 of fiscal 2025. Total Q3 pre-tax stock-based compensation expense was $429,000, compared to $649,000 in the comparable quarter a year ago and $663,000 in the prior quarter. On 31 December, 2024, the company had $15.1 million in cash and cash equivalents, compared to $14.4 million at 31 March, 2024. Working capital was $17.9 million as of 31 December, 2024, versus $19.1 million at 31 March, 2024. Stockholders' equity as of 31 December, 2024, was $29.9 million, compared to $36 million as of the fiscal year ended 31 March, 2024.

Previously, 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 continue to work with Lincoln International as our strategic and financial advisor to assist in this process. Operator, at this point, we will open the call to Q&A.

Operator (participant)

Thank you. We will now be conducting a question-and-answer session. If you would like to ask a question, please press star one on your telephone keypad. A confirmation tone will indicate your line is in the question queue. You may press star two to remove yourself from the queue. For participants using speaker equipment, it may be necessary to pick up the handset before pressing these star keys. One moment, please, Nicole, for questions. Our first question comes from the line of Robert Christian, a private investor. Please proceed with your question.

Yes, I'd like to know, is the company utilizing AI to do its algorithms and software, and are we going to get a white paper on the benchmarks for the Gemini 2?

Didier Lasserre (VP of Sales)

Yeah, so the algorithms we're actually writing ourselves with the folks that we have in our Israeli division. As far as a white paper, yes. So as we mentioned, the first algorithm on Gemini 2 will be the YOLO model, specifically YOLO 3. We will also be following that up with the YOLO 5, but we will be publishing benchmarks when that's done. Correct.

Okay, and the second question I have was, when Lincoln was brought on or it was announced back last year in May, it was to enhance the company valuation. We're actually lower now than we were. Are they the right partner?

Douglas Schirle (CFO)

Well, yeah, we've been working with them, and they have brought us opportunities that we've been looking at, and we continue to look at opportunities. I don't know that it's fair to say that they're brought in to increase the company's valuation. The intent was to find strategic alternatives to further grow the company and provide investment into the company.

I was almost positive that's what was stated in May. I could be wrong, but anyway, thank you very much.

Didier Lasserre (VP of Sales)

Thank you, Christian.

Operator (participant)

Thank you. Again, as a reminder, if anyone has any questions, you may press star one to join the queue. And it looks like, yeah, no further questions. Therefore, I will turn the call back over to Lee-Lean Shu for closing remarks.

Lee-Lean Shu (CEO)

Thank you all for joining us. We look forward to speaking with you again when we report our Q4 and the full year fiscal 2025 results. Thank you.

Operator (participant)

Thank you, ladies and gentlemen. This concludes today's conference, and you may disconnect your lines at this time. Thank you for your participation.