GSI Technology, Inc. (GSIT): Business Model Canvas [Dec-2025 Updated]

You're looking at GSI Technology, Inc. right now, and honestly, it feels like a company at a crossroads, trying to bridge a legacy business with a massive technological leap. After pulling the numbers from their fiscal year 2025 filings, the story is clear: they banked $20.5 million primarily from established, high-speed SRAM sales, but they are betting the farm on their proprietary Gemini APU, pouring $16.0 million into R&D while sitting on $13.4 million in cash. This Business Model Canvas breaks down exactly how this defense/AI-focused firm balances its reliable memory business with its ambitious compute-in-memory future, so you can see the precise risks and opportunities in their structure below.

GSI Technology, Inc. (GSIT) - Canvas Business Model: Key Partnerships

You're looking at the network of external entities GSI Technology, Inc. (GSIT) relies on to execute its strategy, especially as it pushes the Gemini-II and Plato chips. These relationships are critical for everything from fabrication to market access.

TSMC for semiconductor manufacturing (fabless model)

GSI Technology, Inc. operates with a fabless model, meaning it designs the chips but relies on external foundries for production. The company has a longstanding partnership with the provider in the semiconductor manufacturing process, which is identified in filings as Semiconductor Manufacturing Company Limited, or TSMC. This relationship is a key resource, as failure by this supplier to meet requirements on a timely basis at competitive prices could cause manufacturing delays, revenue loss, or higher cost of revenues.

Government agencies for Small Business Innovation Research (SBIR) contracts

Government contracts are a vital source of non-dilutive funding and technology validation for GSI Technology, Inc.'s APU family of products. These partnerships strengthen relationships with defense stakeholders and offset development costs. You can see the financial impact:

• A potential U.S. Army SBIR contract was secured, valued up to $250,000, to develop edge computing AI solutions with Gemini-II.
• In January 2024, GSI Technology, Inc. was selected by AFWERX (the Air Force innovation arm) for a Direct-to-Phase II SBIR contract valued at $1.1 million.
• Milestone payments totaling an estimated $1.25 million were agreed upon for the AFWERX contract, with $435,000 received in fiscal 2024 and $318,000 received in fiscal 2025.
• The Space Development Agency awarded an additional $752,000 under an existing SBIR to characterize the Gemini-II die for radiation tolerance.
• Government funding received under SBIRs is recorded as an offset to research and development expense.

Universities (e.g., Cornell) for APU research and benchmarking

Academic partnerships provide crucial, independent validation for the Associative Processing Unit (APU) technology. The results help position the technology against market leaders.

The recent publication of research from Cornell University was an important moment, validating the Gemini-I APU's ability to match NVIDIA's A6000 GPU performance on retrieval-augmented generation tasks while consuming over 98% less energy. This validation directly supports the claims for the next-generation Gemini-II APU.

Potential strategic/financial partners for Plato chip development

The development of the new Plato chip, which targets the large language model (LLM) AI market, is a major strategic focus requiring significant capital. GSI Technology, Inc. is actively pursuing funding for this initiative.

Funding/Development Target	Amount/Status	Date/Context
Plato Chip Development Funding Goal	$50 million	Management pursuit as of Q3 Fiscal 2025.
Registered Direct Offering Gross Proceeds	Approximately $50 million	Closed on or about October 22, 2025.
Use of Proceeds	General corporate purposes, including development of the APU product line (Plato).	October 2025.
Strategic Partner Engagement	Engaging for funding and collaboration on testing/prototyping early versions.	Expected over the next year or so (post-Q2 FY2026).

The company is also engaging with other aerospace companies evaluating Gemini-II for low-power inflight applications, one of which is also a potential funding partner for Plato.

Independent sales representatives and distributors

GSI Technology, Inc. markets its products through a network of independent sales representatives and distributors. While specific commission structures aren't detailed, the financial impact of customer mix and distributor relationships is evident in revenue reporting. For example, the company noted that variable consideration, which primarily results from stock rotation rights and quick pay discounts provided to certain distributors, is not typically significant.

Customer sales data reflects the channels through which revenue flows:

• In fiscal 2025, SRAM business saw increased sales to KYEC for their test and measurement equipment.
• Military/defense sales accounted for 25% of Q3 Fiscal 2025 shipments.
• Sales to Nokia in Q3 Fiscal 2025 were $239,000, representing 4.4% of net revenues.
• SigmaQuad sales represented 62.5% of all shipments in Q1 FY2026, up from 36.3% in Q1 FY2025.

The fiscal year ended March 31, 2025, saw annual net revenues of $20.52 million.

GSI Technology, Inc. (GSIT) - Canvas Business Model: Key Activities

You're looking at the core engine driving GSI Technology, Inc. right now, which is a mix of deep engineering work and disciplined cost management. Here's the quick math on what they are actively doing to keep the lights on and push their specialized tech forward as of late 2025.

Research and development (R&D) of Associative Processing Units (APU)

GSI Technology, Inc. is definitely pouring resources into its APU roadmap, though spending has been tightened. For the full fiscal year 2025, Research and development expenses totaled $16.0 million, a noticeable step down from the $21.7 million spent in fiscal 2024. This reduction was largely a result of cost reductions announced in August 2024. To be fair, government funding under the SBIR programs offset R&D expenses by $1.2 million in fiscal 2025. You saw R&D expenses dip to $3.0 million in the fourth quarter of fiscal 2025, and it was $4.0 million in the third quarter of fiscal 2025.

The focus is clearly on the Gemini-II APU and the next-generation Plato chip. The Gemini-II, a compute-in-memory chip, is positioned for edge AI, delivering AI capabilities at just 15 watts versus data center GPUs consuming up to 2 kilowatts. A Cornell University study even suggested the APU architecture achieves comparable performance while consuming 98% less energy. They delivered an APU Leda-2 board to an offshore defense contractor for proof-of-concept development in the first quarter of fiscal 2026.

The strategic pivot involves Plato, which targets the large language model (LLM) AI market at the edge, with management pursuing $50 million in funding for this development program. If onboarding takes 14+ days, churn risk rises, and similarly, if Plato funding discussions stall, the long-term roadmap faces pressure.

Design and marketing of high-performance Static Random-Access Memory (SRAM)

The SRAM business is currently the revenue workhorse, showing strong momentum tied to AI expansion. For the first quarter of fiscal 2026, SRAM revenue climbed sharply, increasing 35% year-over-year compared to Q1 FY2025. This success is driven by a favorable product mix shift; high-performance SigmaQuad SRAMs accounted for 62.5% of all shipments in Q1 FY2026, up significantly from just 36.3% in Q1 FY2025. This mix helped boost the gross margin to a two-year high of 58.1% in Q1 FY2026.

Overall, GSI Technology, Inc.'s net revenues for the full fiscal year 2025 were $20.5 million, which was a 6% decline from the $21.8 million reported in fiscal 2024. However, the most recent quarter, Q1 FY2026, saw net revenues hit $6.3 million, up from $5.9 million in Q4 FY2025.

Here are the key SRAM sales metrics:

Metric	Q1 FY2026 (Ended June 30, 2025)	Q1 FY2025 (Ended June 30, 2024)	FY 2025 (Ended March 31, 2025)
Net Revenues	$6.3 million	$4.7 million	$20.5 million
Gross Margin	58.1%	46.3%	49.4%
SigmaQuad Shipments Share	62.5%	36.3%	N/A

Software and algorithm development for Gemini-II APU applications

Developing the software stack is just as critical as the silicon itself for the Gemini-II APU. The company is working to support Gemini-II with a Python-supported compiler this year, 2025. They are creating specialized algorithms for the U.S. Air Force Research Laboratory to leverage the compute-in-memory architecture of the Gemini APU. Furthermore, GSI Technology, Inc. is currently developing a multi-modal LLM that targets edge applications, with benchmark results expected by fall 2025. The compiler stack framework lets customers optimize applications by editing APIs provided by GSI Technology, Inc., or by writing their own APIs.

• Gemini-II tape-out targeted for February 2025, with availability in May 2025.
• Preparing for a second spin of Gemini-II for mass production.
• Developing multi-modal LLM with benchmarks due by fall 2025.
• Working to support Gemini-II with a Python supported compiler in 2025.

Operational efficiency and cost-cutting initiatives

You can see the results of the cost discipline clearly in the operating expense line items. Total operating expenses for fiscal year 2025 were $21.0 million, a substantial reduction from $32.3 million reported in fiscal 2024. This structural cost improvement is central to their goal of preserving cash and extending their runway, and management expects to maintain quarterly operating expenses at current levels to minimize cash burn until new funding is secured. For instance, fourth quarter operating expenses were only $5.6 million, down from $7.2 million in the fourth quarter of fiscal 2024.

The reduction in R&D expenses from $21.7 million in fiscal 2024 to $16.0 million in fiscal 2025 was primarily due to cost reductions announced in August 2024. The company reported a net loss of $(10.6) million for fiscal 2025, which meaningfully reduced the net loss by 47% from the $(20.1) million loss in fiscal 2024.

Securing government contracts and defense-related programs

Government programs are a key activity for validating the APU technology and providing non-dilutive funding. As of May 1, 2025, total general payments from their ongoing SBIR programs amounted to $1.6 million. In January 2025, GSI Technology, Inc. announced selection by the U.S. Army for a potential SBIR contract award valued at up to $250,000 to develop AI solutions leveraging Gemini-II. They also anticipate progressing into Phase 2 of the U.S. Army contract, which could be worth up to $2M. Furthermore, they secured an initial production order for radiation-hardened SRAM from a North American prime contractor in March 2025, which carries a significantly higher gross margin than traditional SRAM chips. Still, defense sales are volatile; military and defense customers represented 19.1% of total shipments in Q1 FY2026, down from 31.9% of shipments in the comparable period a year ago.

Finance: draft 13-week cash view by Friday.

GSI Technology, Inc. (GSIT) - Canvas Business Model: Key Resources

You're looking at the core assets GSI Technology, Inc. has built its value proposition upon. These aren't just line items; they are the foundation for their specialized market position in high-performance memory and AI acceleration.

The most tangible financial resource you need to track is the balance sheet position. GSI Technology, Inc. reported cash and equivalents, which stood at $13.4 million at the end of Fiscal Year 2025, March 31, 2025. That figure saw a significant increase to $22.7 million by June 30, 2025, following an ATM program raise.

The intellectual property is substantial, built over two decades. GSI Technology, Inc. holds 129 granted patents, with recent grants in 2025 covering areas like secure in-memory units (Patent Number 12,387,002, August 12, 2025) and Natural Language Processing with k-NN (Patent Number 12,367,346, July 22, 2025). This portfolio is split between Memory Patents and Associative Computing Patents.

The proprietary Associative Processing Unit (APU) technology, specifically the Gemini-I and Gemini-II generations, is central to their future strategy. The Gemini-II second-generation APU silicon is expected to deliver roughly 10x faster throughput and lower latency for memory-intensive AI workloads. The Gemini-I APU demonstrated GPU-class performance but with over 98% lower energy consumption than a comparable GPU on RAG workloads.

Here's a quick look at the performance metrics validated for the APU architecture:

Metric	Gemini-I Result vs. GPU (NVIDIA A6000)	APU Power Consumption vs. GPU
Throughput	Comparable	N/A
Energy Consumption	N/A	Over 98% lower
Power Draw (Example)	N/A	Just 15W vs. conventional 2kW

The high-speed, radiation-hardened SRAM product portfolio is a mature, revenue-generating asset, serving defense and aerospace. These devices are qualified to Class-Q equivalent and Class-V equivalent levels to meet rigorous requirements. The talent pool supports this, with R&D expenses reported at $4.0 million for the third quarter of fiscal 2025 (ended December 31, 2024).

Key specifications for the SRAM product lines include:

• SigmaQuad SRAMs: Operating Frequency Up To 1333 MHz.
• SigmaQuad SRAMs: Densities from 18Mb to 288Mb.
• No Bus Turnaround (NBT) SRAMs: Operating Frequency Up To 400 MHz.
• Rad-Hard products include SigmaQuad-II+, Synchronous Burst, and NBT families.

GSI Technology, Inc. (GSIT) - Canvas Business Model: Value Propositions

You're looking at the core reasons customers choose GSI Technology, Inc. (GSIT) products, which is all about delivering extreme performance where traditional architectures hit power or speed walls. Here's the breakdown based on what the company is delivering as of late 2025.

High-Speed SRAM: High density, bandwidth, and low latency for networking/telecom

GSI Technology, Inc. maintains a strong offering in high-performance static random-access memory (SRAM) for demanding infrastructure. These products are recognized for very high transaction rates, high density, low latency, high bandwidth, and fast clock access times. The value here is providing the foundational speed required for core networking and telecom equipment like routers and switches.

The specific capabilities of their memory lines, which serve military, aerospace, and industrial needs alongside networking, include:

• SigmaQuad SRAMs: Densities from 18Mb-288Mb.
• SigmaQuad SRAMs: Operating Frequency Up To 1333 MHz.
• SigmaQuad SRAMs: Read Latencies of 1.5, 2, 2.5, 3, 5, and 6 clock cycles.
• No Bus Turnaround (NBT) SRAMs: Densities from 4Mb-288Mb.
• Low Latency DRAMs (LLDRAMs): Densities up to 576Mb with Operating Frequency Up To 5330 MHz.

For context on the company's overall scale in this segment, GSI Technology reported Net Revenues of $20.5 million for the full fiscal year 2025.

Radiation-Hardened SRAM: Reliability for military and aerospace applications

The value proposition extends to mission-critical environments through specialized memory. GSI Technology leverages its core SRAM competency to offer Radiation-Hardened and Radiation-Tolerant memory products. This is a high average selling price (ASP), high margin area. A key validation point for this segment was securing an initial order for radiation-hardened SRAM from a North American prime contractor during fiscal 2025, with additional orders anticipated in fiscal 2026.

Gemini APU: GPU-class AI performance at dramatically lower energy consumption

The Associative Processing Unit (APU) is positioned as a paradigm shift, offering true compute-in-memory technology. The value is delivering AI/HPC performance without the massive power draw of incumbent solutions, which is critical for sustainability and deployment in power-constrained environments.

Independent validation by Cornell University researchers, presented at the Micro '25 conference, quantified this advantage for the Gemini-I APU:

Performance Metric	GSI Gemini-I APU	NVIDIA A6000 GPU	Standard CPU
AI Throughput (RAG tasks)	Comparable	Comparable	Significantly Lower
Energy Consumption (Large Datasets)	Over 98% lower than GPU	100% (Baseline)	~80-85% of GPU
Retrieval Task Completion Time	Shortest (up to 80% faster vs CPU)	Short	Longest

The Gemini-I APU targets disruption in the $100 billion AI inference market.

Compute-in-Memory (CIM) for ultra-low power edge AI and HPC

The CIM architecture minimizes data movement, which is the primary source of energy consumption in AI workloads. This efficiency is directly translated into market opportunities for power-constrained applications. The recently released second-generation silicon, Gemini-II, is engineered to further this value proposition.

• The Gemini-II APU is presented as delivering GPU-class performance while operating at only 15W for edge applications like drones.
• Gemini-II is stated to deliver roughly 10x faster throughput and even lower latency for memory-intensive AI workloads compared to Gemini-I.
• The company is targeting the drone market, projected at $2.7 billion by 2030, and the global edge AI processor market, projected at $9.6 billion by 2030.

Specialized solutions for similarity search and Boolean processing

The APU is designed for in-place associative computing, which is highly effective for large database search queries. This capability serves high-growth areas beyond general AI inference.

The value here is efficient search in a compact, low-power format for applications such as:

• Visual search queries for e-commerce.
• Computer vision.
• Drug discovery.
• Cyber security.

The architecture is also explicitly designed for Boolean processing, a core capability for certain defense and data filtering tasks.

Finance: review the Q2 FY2026 cash position of $25.3M against the planned $50M registered direct offering to ensure funding runway for the Plato chip design.

GSI Technology, Inc. (GSIT) - Canvas Business Model: Customer Relationships

You're looking at how GSI Technology, Inc. (GSIT) manages its connections with the people buying its specialized memory and AI chips. It's a mix of supporting long-time partners and deeply engaging with new, high-potential defense and AI clients. Honestly, the relationship strategy seems bifurcated: keep the legacy revenue stream stable while aggressively pushing the new APU technology.

Direct, consultative engagement for APU proof-of-concept and evaluation

For the Associative Processing Unit (APU) portfolio, the engagement is clearly consultative, especially given the paradigm shift in compute-in-memory technology. You see this in the validation work. For instance, Cornell benchmarks confirmed the Gemini-I APU delivered GPU-class throughput comparable to NVIDIA's A6000 on RAG workloads while using over 98% less energy than a GPU. Furthermore, these proof-of-concept engagements are showing tangible results, with customers reporting first-response times up to three times faster than alternative solutions. The next-generation Gemini-II chip is designed to deliver approximately 10x faster throughput. This level of performance difference requires deep technical collaboration to integrate, which is why the sales process is inherently consultative.

• Gemini-I processing time reduced up to 80% versus CPUs on 10GB to 200GB datasets.
• Gemini-II targets complex edge AI capability at 15W.
• The company is actively working to support Gemini-II with a Python-supported compiler in 2025.

Long-term, established relationships with legacy SRAM OEM customers

The foundation of GSI Technology, Inc.'s current revenue still rests on its established relationships, primarily in the Static Random-Access Memory (SRAM) space. These are long-term OEM connections where GSI Technology, Inc. provides high-reliability components. Looking at the fiscal year ended March 31, 2025, total net revenues were $20.5 million, down from $21.8 million the prior year. The relationship health is visible in the quarterly mix. In the fourth quarter of fiscal 2025, sales to KYEC represented 29.5% of net revenues at $1.7 million, while sales to Nokia were 7.5% at $444,000. The focus on higher-margin products like SigmaQuad SRAMs is a direct result of managing this customer base, as SigmaQuad sales made up 39.3% of Q4 FY2025 shipments. The rebound in Q1 FY2026 to $6.3 million in GAAP revenue suggests these legacy customers are resuming orders after inventory depletion.

Metric	Q4 Fiscal Year 2025	Fiscal Year 2025
Net Revenues	$5.9 million	$20.5 million
Gross Margin	56.1%	49.4%
Shipments by Product Type (SigmaQuad)	39.3%	N/A

Dedicated sales and technical support for complex defense/aerospace projects

For the defense and aerospace segments, the relationship is characterized by high-trust, specialized support, often involving government contracts. Military/defense sales accounted for 30.7% of Q4 FY2025 shipments. This segment requires specialized, high-quality parts, such as radiation-tolerant and radiation-hardened SRAMs. A major milestone in this relationship track was securing an initial production order for radiation-hardened SRAM from a North American prime contractor in March 2025, with follow-on orders anticipated in fiscal 2026. Furthermore, GSI Technology, Inc. is engaged through multiple DoD SBIR contracts and ongoing evaluations for applications like drones and satellites. Two other aerospace companies are evaluating the technology for onboard satellite applications for Synthetic Aperture Radar (SAR) image processing.

• Defense/aerospace applications drive demand for GSI Technology, Inc.'s highest truly random transaction rate in the industry: 1866 million transactions per second (MT/s).
• The new radiation-hardened SRAM sale carries a significantly higher gross margin than traditional SRAM chips.
• The Gemini-II APU is specifically positioned to deliver capability at 15W for defense systems.

Ongoing strategic review to enhance shareholder value

The management team's interactions with the investment community reflect an ongoing strategic review aimed at improving financial metrics, which directly impacts shareholder relationships. The operating loss for fiscal 2025 narrowed significantly to $(10.8) million, down from $(20.4) million in the prior year. This was supported by disciplined cost management, including reducing R&D expense to $16 million in FY2025 from $21.7 million in FY2024. A concrete action to improve the balance sheet was the sale and leaseback of the Sunnyvale headquarters, which provided a $5.8 million gain. More recently, the company completed a significant equity raise on October 22, 2025, netting $47 million (net after fees) to fund the Gemini-II deployment and Plato development. This capital infusion directly addresses working capital needs, which stood at $16.4 million as of March 31, 2025.

Finance: draft 13-week cash view by Friday.

GSI Technology, Inc. (GSIT) - Canvas Business Model: Channels

You're looking at how GSI Technology, Inc. gets its products-the specialized memory and AI chips-into the hands of its customers as of late 2025. The channel mix shows a heavy reliance on specific large customers and government contracts, which is typical for this niche, but also highlights channel volatility.

For the fiscal year ended March 31, 2025, GSI Technology, Inc. reported total net revenues of $20.52 million. The fourth quarter of fiscal year 2025 (three months ended March 31, 2025) saw net revenues of $5.9 million.

The composition of sales by customer segment in Q4 FY2025 shows significant concentration, which directly reflects the effectiveness and risk profile of their current channels:

Customer/Segment	Q4 Fiscal Year 2025 Revenue Amount	Percentage of Net Revenues (Q4 FY2025)	Comparison to Q4 Fiscal Year 2024
KYEC	$1.7 million	29.5%	Up from 10.6% of net revenues year-over-year.
Nokia	$444,000	7.5%	Down from 13.5% of net revenues year-over-year.
Military/Defense Shipments	N/A (Shipment % only)	30.7% of fourth quarter shipments	Down from 35.5% of shipments in the comparable period a year ago.

The first quarter of fiscal year 2026 (three months ended June 30, 2025) shows the defense segment as 19.1% of total shipments.

Network of independent sales representatives

GSI Technology, Inc. uses independent sales representatives, who are compensated on commission based on product shipment. These representatives are viewed as important because their products act as door openers for broader sales opportunities.

Global distributors for memory products

The company utilizes large distributors, such as Avnet, which carries most of the lines. Distributors are compensated based on margin.

Direct sales to key OEM and prime contractor customers

Direct engagement is a key channel, particularly for specialized, high-margin products. GSI Technology, Inc. secured an initial order for its radiation-hardened SRAM from a North American prime contractor in Q4 FY2025, with follow-on orders expected in fiscal 2026. This specific sale carries a significantly higher gross margin than traditional SRAM chips.

Direct engagement with government agencies (SBIR programs)

Direct government engagement is channeled through programs like the Small Business Innovation Research (SBIR) contracts. For the three months ended March 31, 2025, Research and development expenses were reduced by $870,000 due to funding under the SBIR programs. For the three months ended June 30, 2025, cost discipline benefited from government funding under SBIR contracts, helping offset development spending.

You should track the percentage of shipments to military/defense customers, which was 30.7% in Q4 FY2025 and 19.1% in Q1 FY2026.

Finance: draft 13-week cash view by Friday

GSI Technology, Inc. (GSIT) - Canvas Business Model: Customer Segments

You're looking at the core markets GSI Technology, Inc. serves with its high-performance memory and AI solutions as of late 2025. Honestly, the mix is shifting, which is something you need to track closely, especially with the growth in AI-related demand.

The customer base is concentrated, meaning a few large buyers can significantly swing the revenue percentages quarter-to-quarter. For instance, looking at the first quarter of fiscal year 2026 (ending June 30, 2025), the percentage contribution from the defense sector dipped compared to the prior year, while a specific semiconductor design customer became a major revenue contributor.

Here's a breakdown of the key customer groups GSI Technology, Inc. targets with its SRAM and new AI/APU products:

• Military, defense, and aerospace: Focus on satellites, radar, and guidance systems, including recent U.S. Army SBIR contract work for Gemini-II edge AI.
• Networking and telecommunications: Serving equipment manufacturers, though sales to specific large telecom customers like Nokia have shown variability.
• High-Performance Computing (HPC) and AI/ML developers: A rapidly growing segment driven by demand for high-speed memory in AI processors and the introduction of the Gemini-II Associative Processing Unit (APU).
• Industrial, medical (ultrasound), and automotive (smart cruise control) OEMs: These are stated target markets for the company's memory products.
• Key AI Chip Enablers: Customers integral to leading AI chip manufacturing, evidenced by significant design wins and growing SRAM sales to GPU providers.

The financial reality of these segments, based on recent shipment percentages and revenue figures from fiscal year 2026, looks like this. Remember, these are shipment percentages or specific customer revenue snapshots, not necessarily total segment revenue:

Customer Segment / Metric	Latest Reported Period (Q2 FY2026, ending Sept 30, 2025)	Prior Period (Q1 FY2026, ending June 30, 2025)	Year Ago Period (Q2 FY2025, ending Sept 30, 2024)
Military/Defense Sales (Shipment %)	28.9%	19.1%	40.2%
Sales to Cadence Design Systems (Revenue %)	21.6% ($1.4 million)	23.9% ($1.5 million)	0% ($0)
SigmaQuad Sales (Shipment %)	50.1%	62.5%	38.6%
Sales to Nokia (Revenue %)	Not explicitly stated for Q2 FY2026	8.5%	17.8% ($812,000)

The growth in the AI segment is clear when you look at the SRAM performance. SRAM revenue growth in Q1 FY2026 was 35% year-over-year. Furthermore, the company confirmed that its largest customer, integral to AI chip manufacturing, is expected to maintain stable order volume for the rest of the fiscal year, despite current supply chain navigation.

For the defense sector, Gemini-II has been approved for prototyping by an offshore defense contractor for synthetic aperture radar (SAR) applications on drones and edge systems. Also, GSI Technology, Inc. secured an initial order for its radiation-hardened SRAM, which is a key validation point for that customer base.

Finance: draft 13-week cash view by Friday.

GSI Technology, Inc. (GSIT) - Canvas Business Model: Cost Structure

You're looking at the core expenses driving GSI Technology, Inc.'s operations for the fiscal year ending March 31, 2025. Understanding this structure is key to seeing where the cash went.

The company's cost structure is heavily weighted toward development and overhead, even after implementing cost-saving measures. For the full fiscal year 2025, total operating expenses came in at $21.0 million, a significant reduction from the $32.3 million reported in fiscal 2024. This reduction reflects a more disciplined operating structure that management focused on achieving.

The largest component of the operating costs was Research and development expenses, totaling $16.0 million in fiscal year 2025. This figure was down from $21.7 million the prior year. To be fair, the fiscal 2024 R&D number included $2.4 million in pre-production mask costs related to the Gemini-II product, which didn't recur in the same magnitude for fiscal 2025. Also, R&D expenses in fiscal 2025 and fiscal 2024 were reduced by $1.2 million and $440,000, respectively, due to government funding under the SBIR programs.

Selling, general, and administrative (SG&A) expenses for FY2025 were $10.8 million. This was a slight tick up from the $10.6 million reported in fiscal 2024. Honestly, this category holds the day-to-day costs of running the business, from sales commissions to executive salaries.

Cost of revenues (COGS) for semiconductor manufacturing and testing is derived from the reported net revenues of $20.5 million and the fiscal 2025 gross margin of 49.4%. Here's the quick math: COGS was approximately $10.37 million ($20.5 million (1 - 0.494)). This COGS figure directly covers the costs associated with the fabless model, which includes paying third parties for wafer fabrication, assembly, and final testing of the memory and APU products.

The company has been actively managing its cost base, especially concerning its workforce. These efforts directly impact the cost structure, often showing up as severance in quarterly results or as reduced overhead in annual figures.

• Gross margin for Q2 FY2025 was impacted by severance costs.
• Cost reductions announced in August 2024 drove the decline in R&D expenses.
• The decrease in fiscal 2025 gross margin to 49.4% from 54.3% was due to product mix and the effect of lower revenue on fixed costs in COGS.

Here is a breakdown of the primary operating expense categories for GSI Technology, Inc. for the full fiscal year 2025:

Cost Category	FY 2025 Amount (in millions)	FY 2024 Amount (in millions)
Research and Development (R&D)	$16.0	$21.7
Selling, General, and Administrative (SG&A)	$10.8	$10.6
Total Operating Expenses	$21.0	$32.3

It's also worth noting that operating expenses in fiscal 2025 included a non-recurring gain on the sale of assets of $5.8 million from the sales and leaseback transaction of the Company's headquarters building in Sunnyvale, CA. That gain offsets the reported operating expenses, so you have to look past it for the true run-rate cost structure.

Finance: draft 13-week cash view by Friday.

GSI Technology, Inc. (GSIT) - Canvas Business Model: Revenue Streams

You're looking at the core ways GSI Technology, Inc. brings in money, which is heavily weighted toward its established memory products but increasingly pivoting toward its advanced processing units.

The company's financial performance for the most recently completed full fiscal year shows the baseline for these streams. Net revenues for fiscal year 2025 were $20.5 million, compared to $21.8 million for fiscal 2024.

The primary revenue driver remains the Sales of high-speed SRAM products (core legacy business), which includes their SigmaQuad and SigmaDDR lines. Demand for these SRAM solutions, fueled by AI processors, is strong, as seen in recent quarters:

• Q4 Fiscal Year 2025 Net Revenues: $5.9 million.
• First Quarter Fiscal Year 2026 Net Revenues: $6.3 million.
• Second Quarter Fiscal Year 2026 Net Revenues: Approximately $6.44 million.

Here's a quick look at how the recent quarterly revenues stack up:

Period Ending	Net Revenues (Millions USD)	Gross Margin (%)
March 31, 2025 (Q4 FY2025)	$5.9	56.1
September 30, 2025 (Q2 FY2026 Preliminary)	$6.4	54.8
June 30, 2025 (Q1 FY2026)	$6.3	58.1

A significant, higher-margin component is the Sales of radiation-hardened SRAM to defense contractors. GSI Technology, Inc. secured an initial order for this product from a North American prime contractor, with follow-on orders anticipated in fiscal 2026. Military/defense sales represented a notable portion of shipments in recent quarters:

• First Quarter Fiscal Year 2026 Shipments: Military/defense sales accounted for 19.1%.
• Second Quarter Fiscal Year 2026 Shipments: Military/defense sales accounted for 28.9%.

The company also generates Revenue from government SBIR contracts and development milestones. The Space Development Agency awarded an additional $752,000 under an existing SBIR contract specifically to characterize the Gemini-II die for radiation tolerance.

Looking ahead, Future revenue from APU chip and server sales (Gemini-I/II) represents the growth vector. The Gemini-II Associative Processing Unit (APU) chip is confirmed as production-ready. GSI Technology, Inc. delivered an APU Leda-2 board to an offshore defense contractor for proof-of-concept development focusing on synthetic aperture radar (SAR) applications. To accelerate this product line, the company closed a registered direct offering for gross proceeds of $50 million.

Finance: draft 13-week cash view by Friday.