GSI Technology reported net revenues of $6.1 million for the quarter ended December 31, 2025, a 12% increase from $5.4 million in Q3 FY 2025 and modestly below the $6.4 million achieved in Q2 FY 2026. Gross margin was 52.7%, down from 54.0% a year earlier and 54.8% in the prior quarter, reflecting a less favorable product mix. Sales to key customers included KYEC at $1.1 million (17.9% of revenues), Nokia at $675,000 (11.1%), and Cadence Design Systems at $233,000 (3.8%). Military and defense comprised 28.5% of shipments, while SigmaQuad accounted for 41.7%. Operating expenses rose to $10.1 million, driven by R&D investment of $7.5 million—up from $4.0 million a year ago—to support Gemini-II and the new Plato IP, and SG&A of $2.6 million. The operating loss widened to $6.9 million, though non-cash adjustments in interest and other income of $3.6 million narrowed net loss to $3.0 million, or $0.09 per diluted share. The company expects Q4 net revenues between $5.7 million and $6.5 million, with gross margin around 54–56%.

GSI ended the quarter with $70.7 million in cash and equivalents, up from $13.4 million at the end of Q4 FY 2025, propelled by $46.9 million in net proceeds from a Registered Direct Offering in October 2025. Operating activities consumed $7.9 million, investing activities used $0.3 million, and financing activities contributed $53.5 million, resulting in working capital of $71.7 million and stockholders’ equity of $83.6 million. The expanded cash disclosures aim to give investors greater clarity on quarterly cash generation and consumption, highlighting the impact of ongoing R&D spend on Gemini-II and the newly launched Plato hardware development.

During Q3, GSI initiated a proof-of-concept engagement with G2 Tech and two government agencies, targeting an expected $1 million in government funding. The company also published independent benchmark results demonstrating that Gemini-II achieved a 3-second time-to-first-token (TTFT) on a multimodal edge inference workload at approximately 30 watts system power—significantly lower power than competing GPU-based systems, which consume over 100 watts for similar latency. These results underscore Gemini-II’s low-power, low-latency profile ideal for defense-oriented applications such as autonomous perimeter security drones and other edge deployments.

Leveraging the recent bench- mark validation and strengthened cash position, GSI is accelerating its APU roadmap into the Plato hardware platform and enhancing the Gemini-II software stack. The company is actively pursuing initial design wins in defense programs for drones and unmanned systems, as well as select commercial edge markets. Management believes energy-efficient multimodal inference at power-constrained edges will open new opportunities in “physical AI” segments, including smart cities and industrial automation, laying the groundwork for recurring revenue streams beyond traditional SRAM solutions.