Gsi Technology (GSIT) stock price is $3.86, unchanged on the day, as of Oct 23, 2026. It has a market cap of $148.12Mn, and is classified in the Semiconductor Design industry (Technology sector).
Enterprise Value
Margins
P/FCF
Key Stats
Market Cap148.12 Mn
P/E-9.11
P/S5.89
Revenue Growth (1y) (Qtr)0.46
Add ratio to table…
About
GSI Technology provides in place associative computing solutions for artificial intelligence and high performance computing markets, including natural language processing, computer vision and cyber security. The company also designs and sells high speed synchronous SRAM products for networking, telecommunications, defense, aerospace, test and measurement and medical applications. GSI Technology operates a fabless model, relying on third party foundries for wafer fabrication…
GSI Technology provides in place associative computing solutions for artificial intelligence and high performance computing markets, including natural language processing, computer vision and cyber security. The company also designs and sells high speed synchronous SRAM products for networking, telecommunications, defense, aerospace, test and measurement and medical applications. GSI Technology operates a fabless model, relying on third party foundries for wafer fabrication while focusing internal resources on research, development and marketing.
Revenue comes from the sale of associative processing unit products such as the Gemini I and Gemini II chips and from related services that provide APU as a service through connections with Amazon Web Services, Microsoft Azure and Google Cloud. Additional revenue is derived from the sale of high speed synchronous SRAM devices, including radiation hardened and radiation tolerant versions, to original equipment manufacturers, distributors and government contractors. Revenue also includes licensing fees for the APU intellectual property to third parties that integrate the technology into their own semiconductor designs.
The company operates through the following segments.
In place Associative Computing: This segment designs and sells associative processing units such as the Gemini I and Gemini II chips, offers APU as a service via cloud connections, and provides software tools and SDKs for AI and high performance computing applications.
High Speed Synchronous SRAM: This segment supplies SyncBurst, NBT, SigmaQuad and SigmaDDR family products, including radiation hardened and radiation tolerant variants, for use in networking, telecom, defense, aerospace, test and measurement and medical equipment.
In the SRAM market, GSI Technology holds a niche position as a supplier of high speed, high density memory parts, competing with larger firms such as Infineon Technologies, Integrated Silicon Solution and Renesas Electronics. For its associative computing products, the company faces competition from NVIDIA and Intel, which offer GPUs and CPUs for AI workloads, but GSI differentiates itself through its compute in memory architecture that delivers low latency and low power consumption. The company's fabless business model, strong patent portfolio and expertise in radiation hardened semiconductors provide additional competitive barriers.
GSI Technology sells its SRAM products to original equipment manufacturers including Airbus, BAE Systems, KYEC, Lockheed Martin, Northrop Grumman and Rockwell Collins, and to distributors such as Avnet, Mouser and Digi Key. The company also supplies radiation hardened SRAM and associative computing solutions to U. S. government agencies such as the Space Force, the Air Force Research Laboratory and the Army. In addition, GSI Technology provides APU as a service to customers who access the technology through Amazon Web Services, Microsoft Azure and Google Cloud.
Sector:TechnologySector rationaleGSI Technology designs and sells semiconductors, specifically associative processing units (APUs) and high-speed synchronous SRAM, which fall under Semiconductor Design and Electronic Components. The company also generates revenue from APU-as-a-service via cloud platforms and licensing its semiconductor intellectual property, all of which are core Technology sector activities.Industries:Semiconductor DesignSemiconductor DesignPrimaryGSI Technology operates a fabless model designing and selling integrated circuits, specifically associative processing units (Gemini I and Gemini II chips) and high-speed synchronous SRAM products. The company also generates revenue by licensing its APU intellectual property to third parties for integration into their own semiconductor designs.AI PlatformsAI PlatformsSecondaryThe company sells AI/ML capabilities as a standalone product through its 'APU as a service' offering, which allows customers to access its associative computing technology via Amazon Web Services, Microsoft Azure, and Google Cloud.Classified using BQ-MICSCIK: 0001126741
Bull & bear
Investment Thesis
▲ Bull case
GSI Technology's SRAM business has established a resilient and growing financial foundation, with fiscal 2026 revenue increasing 22.4% to $25.1 million and gross margins expanding to 54.5% from 49.4% in the prior year. This performance is driven by sustained demand from customers supporting high-performance AI chip design and simulation, which provides recurring and stable cash flow to fund APU development without dilution. The strengthening balance sheet, highlighted by $67.2 million in cash and zero debt at the end of fiscal 2026—a significant improvement from $13.4 million the prior year—results from the October 2025 registered direct offering and provides ample runway to advance both Gemini II commercialization and PLATO tape-out, expected in late fiscal 2027. This financial stability reduces near-term financing risk and allows management to focus on execution rather than survival, a critical advantage for a small-cap innovator in a capital-intensive sector.
The Gemini II architecture is demonstrating tangible technical validation in real-world, power- and latency-constrained edge environments where it outperforms conventional alternatives, creating a clear path to design wins. In the Sentinel drone surveillance proof-of-concept, Gemini II achieved a time to first token of approximately three seconds at just 30 watts of system power on Gemma 312B multimodal workloads—a metric directly tied to real-time decision-making in defense applications. This performance advantage, combined with the U.S. Army SBIR award progressing from Phase One to Phase Two and the newly awarded Phase One smart city project leveraging the same core technology, indicates that Gemini II is not merely a laboratory curiosity but a deployable solution addressing urgent national security and public infrastructure needs. The ability to transfer software and system knowledge between defense and smart city applications creates a repeatable commercialization model, reducing redundant development costs and accelerating market entry across adjacent verticals.
The PLATO chip design is progressing on schedule for tape-out in the first quarter of calendar 2027 (early fiscal 2028), with software hardware alignment improving over Gemini II, which should accelerate time to revenue. Unlike Gemini I, where software development lagged significantly behind hardware, PLATO benefits from reusable software assets developed for Gemini II, allowing for tighter integration and faster validation. Management emphasized that the PLATO timeline has not been delayed, with tape-out targeted for early calendar 2027 and first wafers expected in summertime 2027—consistent with prior guidance. This synchronization between hardware and software reduces integration risk and positions PLATO to enter the market with a more complete value proposition, particularly for applications requiring even higher compute density and efficiency than Gemini II. Successful PLATO commercialization could significantly expand the company’s addressable market beyond current edge AI use cases into next-generation inference accelerators.
GSI Technology's SRAM business has established a resilient and growing financial foundation, with fiscal 2026 revenue increasing 22.4% to $25.1 million and gross margins expanding to 54.5% from 49.4% in the prior year. This performance is driven by sustained demand from customers supporting high-performance AI chip design and simulation, which provides recurring and stable cash flow to fund APU development without dilution. The strengthening balance sheet, highlighted by $67.2 million in cash and zero debt at the end of fiscal 2026—a significant improvement from $13.4 million the prior year—results from the October 2025 registered direct offering and provides ample runway to advance both Gemini II commercialization and PLATO tape-out, expected in late fiscal 2027. This financial stability reduces near-term financing risk and allows management to focus on execution rather than survival, a critical advantage for a small-cap innovator in a capital-intensive sector.
The Gemini II architecture is demonstrating tangible technical validation in real-world, power- and latency-constrained edge environments where it outperforms conventional alternatives, creating a clear path to design wins. In the Sentinel drone surveillance proof-of-concept, Gemini II achieved a time to first token of approximately three seconds at just 30 watts of system power on Gemma 312B multimodal workloads—a metric directly tied to real-time decision-making in defense applications. This performance advantage, combined with the U.S. Army SBIR award progressing from Phase One to Phase Two and the newly awarded Phase One smart city project leveraging the same core technology, indicates that Gemini II is not merely a laboratory curiosity but a deployable solution addressing urgent national security and public infrastructure needs. The ability to transfer software and system knowledge between defense and smart city applications creates a repeatable commercialization model, reducing redundant development costs and accelerating market entry across adjacent verticals.
The PLATO chip design is progressing on schedule for tape-out in the first quarter of calendar 2027 (early fiscal 2028), with software hardware alignment improving over Gemini II, which should accelerate time to revenue. Unlike Gemini I, where software development lagged significantly behind hardware, PLATO benefits from reusable software assets developed for Gemini II, allowing for tighter integration and faster validation. Management emphasized that the PLATO timeline has not been delayed, with tape-out targeted for early calendar 2027 and first wafers expected in summertime 2027—consistent with prior guidance. This synchronization between hardware and software reduces integration risk and positions PLATO to enter the market with a more complete value proposition, particularly for applications requiring even higher compute density and efficiency than Gemini II. Successful PLATO commercialization could significantly expand the company’s addressable market beyond current edge AI use cases into next-generation inference accelerators.
GSI Technology remains heavily dependent on its legacy SRAM business, which, while growing, represents a niche and cyclical segment of the semiconductor market that may not sustain long-term growth as AI workloads shift toward specialized accelerators. Despite 22.4% year-over-year revenue growth in fiscal 2026 to $25.1 million, the absolute revenue base remains extremely small, limiting economies of scale and making the company vulnerable to order fluctuations from a concentrated customer base supporting chip design and simulation. The CFO acknowledged variability in customer orders and quarter-to-quarter sales swings, noting that the business’s stability hinges on the continuation of underlying AI chip demand—a macroeconomic factor outside the company’s control. If AI chip development slows due to capital constraints or a shift toward fewer, more powerful models, SRAM orders could decline rapidly, undermining the cash flow that currently funds APU development and exposing the company to liquidity pressure despite its current cash balance.
The path from Gemini II’s technical demonstrations to meaningful commercial revenue remains unclear and protracted, with management acknowledging they are still in the early stages of engaging customers and have not yet entered design-level discussions that typically precede production commitments. While the Sentinel drone POC, U.S. Army SBIR, and smart city project validate technical feasibility, these are all non-revenue-generating or minimally compensated engagements—such as SBIR awards and proof-of-concepts—that do not equate to purchase orders or volume production. The company has not disclosed any signed design wins, licensing agreements, or supply contracts for Gemini II, and the reliance on leveraging prior work across applications (e.g., drone to smart city) assumes that customer integration cycles are short and replicable, which may not hold true in complex, regulated environments like defense or municipal infrastructure where sales cycles can extend over multiple years and require extensive certification.
The PLATO program, while positioned as the next-generation APU, carries significant execution risk due to its reliance on external foundry timelines and the unproven nature of its architecture in real-world systems, with no guarantee that performance advantages will translate to market acceptance. Although tape-out is expected in early calendar 2027, the subsequent steps—wafer fabrication, packaging, testing, and system integration—typically add six to twelve months before first customer samples are available, meaning meaningful revenue from PLATO is unlikely before fiscal 2029 at the earliest. Furthermore, the company continues to invest heavily in R&D, with operating expenses rising to $31.2 million in fiscal 2026 (up from $21 million in 2025), primarily driven by PLATO development, and the offset from non-dilutive funding remains modest at $1 million annually. Without a clear near-term revenue inflection point from either Gemini II or PLATO, the company faces the risk of depleting its cash reserves before achieving sustainable profitability, particularly if development timelines slip or if competing edge AI solutions from larger semiconductor firms gain traction faster than anticipated.
GSI Technology remains heavily dependent on its legacy SRAM business, which, while growing, represents a niche and cyclical segment of the semiconductor market that may not sustain long-term growth as AI workloads shift toward specialized accelerators. Despite 22.4% year-over-year revenue growth in fiscal 2026 to $25.1 million, the absolute revenue base remains extremely small, limiting economies of scale and making the company vulnerable to order fluctuations from a concentrated customer base supporting chip design and simulation. The CFO acknowledged variability in customer orders and quarter-to-quarter sales swings, noting that the business’s stability hinges on the continuation of underlying AI chip demand—a macroeconomic factor outside the company’s control. If AI chip development slows due to capital constraints or a shift toward fewer, more powerful models, SRAM orders could decline rapidly, undermining the cash flow that currently funds APU development and exposing the company to liquidity pressure despite its current cash balance.
The path from Gemini II’s technical demonstrations to meaningful commercial revenue remains unclear and protracted, with management acknowledging they are still in the early stages of engaging customers and have not yet entered design-level discussions that typically precede production commitments. While the Sentinel drone POC, U.S. Army SBIR, and smart city project validate technical feasibility, these are all non-revenue-generating or minimally compensated engagements—such as SBIR awards and proof-of-concepts—that do not equate to purchase orders or volume production. The company has not disclosed any signed design wins, licensing agreements, or supply contracts for Gemini II, and the reliance on leveraging prior work across applications (e.g., drone to smart city) assumes that customer integration cycles are short and replicable, which may not hold true in complex, regulated environments like defense or municipal infrastructure where sales cycles can extend over multiple years and require extensive certification.
The PLATO program, while positioned as the next-generation APU, carries significant execution risk due to its reliance on external foundry timelines and the unproven nature of its architecture in real-world systems, with no guarantee that performance advantages will translate to market acceptance. Although tape-out is expected in early calendar 2027, the subsequent steps—wafer fabrication, packaging, testing, and system integration—typically add six to twelve months before first customer samples are available, meaning meaningful revenue from PLATO is unlikely before fiscal 2029 at the earliest. Furthermore, the company continues to invest heavily in R&D, with operating expenses rising to $31.2 million in fiscal 2026 (up from $21 million in 2025), primarily driven by PLATO development, and the offset from non-dilutive funding remains modest at $1 million annually. Without a clear near-term revenue inflection point from either Gemini II or PLATO, the company faces the risk of depleting its cash reserves before achieving sustainable profitability, particularly if development timelines slip or if competing edge AI solutions from larger semiconductor firms gain traction faster than anticipated.