GSI Technology reported net revenues of $6.1 million for fiscal Q3 2026, up 12% from $5.4 million in the year-ago quarter and slightly below the $6.4 million recorded in Q2. Gross margin contracted to 52.7% from 54.0% a year earlier and 54.8% in the prior quarter, a decline the company attributed to product mix shifts. Operating expenses climbed to $10.1 million, driven by R&D spending of $7.5 million—up from $4.0 million in Q3 2025—largely related to IP purchases and consulting for the Plato program. SG&A expenses totaled $2.6 million, modestly below the $3.0 million reported in both comparable periods. The result was a $6.9 million operating loss, widening from $4.1 million a year ago, while net loss improved slightly to $3.0 million, or $0.09 per diluted share, aided by $3.6 million of interest and other income tied to warrant-valuation adjustments and a $251,000 tax benefit.

As of December 31, 2025, GSI held $70.7 million in cash and cash equivalents, up sharply from $25.3 million at the end of Q2 and $13.4 million at fiscal-year end March 31, 2025. Net cash used in operating activities during Q3 was $7.9 million, reflecting continued investment in Gemini-II and Plato development. Investing activities consumed $296,000, while financing provided $53.5 million, principally from $46.9 million of net proceeds from the October 2025 registered direct offering. Working capital expanded to $71.7 million, and stockholders’ equity rose to $83.6 million, reinforcing the company’s financial runway through its next hardware tape-out milestone.

Management highlighted third-party benchmarking of Gemini-II’s multimodal APU, which delivered a three-second time-to-first-token (TTFT) on combined text-and-video inference at approximately 30 watts system power. This performance represented up to a threefold faster TTFT versus competing platforms under similar workloads, emphasizing Gemini-II’s suitability for power-constrained edge applications such as drones and unmanned systems. GSI plans further TTFT optimizations over the next five months in advance of a Sentinel demonstration, underscoring its strategy to prioritize performance-per-watt and low latency over peak data-center training metrics.

GSI finalized a proof-of-concept agreement with Israel-based G2 Tech for the Sentinel autonomous perimeter security program, which integrates drones and real-time video analytics. The company expects to receive over $1 million in government funding, to be recorded as an offset to R&D expense, supporting software development for the upcoming Gemma 3 12B implementation on Gemini-II. A competitive evaluation by G2 Tech selected GSI based on its lowest-power, lowest-latency TTFT performance. A successful POC could pave the way for design wins in defense-oriented drone and unmanned markets and broader edge AI applications in smart cities and robotics.