GSI Technology (NASDAQ:GSIT) reported fiscal second-quarter revenue of $6 million and a net loss of $7 million, as inventory shortages delayed some customer shipments and higher manufacturing costs pressured gross margin.
On its Nov. 5 earnings call, management forecast revenue of $5 million to $6 million for the fiscal third quarter and gross margin of approximately 38% to 40%, down from 48.5% in the second quarter. The company also outlined progress on its Gemini-II artificial intelligence processor and its next-generation Plato chip.
Manufacturing costs and delayed shipments
Chief Financial Officer Douglas Schirle said second-quarter revenue was within the company’s guidance range but slightly below its midpoint. Insufficient inventory prevented GSI from filling some orders, while extended manufacturing lead times have made replenishment more difficult, he said.
Schirle said some larger customers had not placed backlog orders despite the company’s warnings about lead times. Vice President of Sales Didier Lasserre added that lead times for substrates and assembly had lengthened.
Second-quarter gross margin fell below the company’s guidance range of 53% to 55%, primarily because of higher manufacturing costs tied to constraints affecting wafers and outsourced assembly, burn-in and testing, Schirle said. Those pressures, along with expected product mix, are affecting the third-quarter margin outlook.
GSI plans to raise customer prices immediately on new orders and effective Jan. 1, 2027, on existing backlog. It is also working with manufacturing partners to improve inventory availability. Schirle said the measures are expected to support a recovery in revenue and gross margin beginning in the fiscal fourth quarter, although replenishment will take time.
Sales to KYEC represented 19.7% of net revenue in the second quarter, up from 16.1% in the preceding quarter, while GSI made no shipments to Cadence during the period. Military and defense customers accounted for 46.7% of shipments, compared with 31.3% in the preceding quarter and 28.9% a year earlier.
Spending and cash use increase
Operating expenses rose to $10.4 million from $6.7 million a year earlier and $8.8 million in the preceding quarter. Schirle attributed the year-over-year increase in research and development spending primarily to external design services for Plato, costs he expects to continue through the planned March 2027 tape-out.
Selling, general and administrative expenses included approximately $2 million in non-routine legal and professional fees related to an internal assessment of certain vendor matters in GSI’s foreign operations. A $600,000 gain from assets received in a vendor settlement partially offset those fees, producing a net impact of about $1.4 million. Schirle said the assessment concluded during the quarter and the company does not expect significant additional expenses related to the matters.
The company’s operating loss widened to $7.4 million from $5.4 million in the preceding quarter. Its net loss was $0.18 per diluted share.
GSI ended the quarter with $69.2 million in cash and cash equivalents, down from $77 million at June 30, 2026, and had no debt. Operating activities used approximately $8 million in cash during the quarter, above the company’s previously discussed expectation of about $4 million per quarter. Schirle said GSI continues to anticipate cash usage of approximately $4 million quarterly under its current operating plan, though actual usage may vary.
Gemini-II projects and software tools
Chairman and Chief Executive Officer Lee-Lean Shu said the initial stage of the Smart Guardian Eye Phase One project in Taiwan was accepted after the quarter ended. The system is providing live AI video monitoring and incident reporting.
Lasserre said Phase One covers approximately 20 cameras in Hsinchu County. A potential second phase is expected to expand the deployment to approximately 80 cameras and add automatic speech recognition. GSI plans to submit a proposal when the government issues a quote request, but any additional work depends on competitive procurement and the county’s decision.
GSI is also developing an AI-assisted software development kit called GRIN to help customers adapt models to its processor. Lasserre said some development tasks that previously took about a month can now be completed in less than a week, although times vary by model and application. The company expects to release GRIN to initial customers before the end of December 2026 and commercially in the first half of calendar 2027.
For the Sentinel program with G2 Tech, GSI has completed chip and software deliverables for the early phase. A field demonstration led by G2 Tech could take place in the first half of calendar 2027, Lasserre said, with timing dependent on the partner and participating defense agencies. Further funding and commercial orders would depend on subsequent program decisions.
Plato remains on schedule
GSI expects to complete Plato’s tape-out in March 2027 and have first-spin chips ready for initial sampling by late summer 2027. The processor is designed to run AI inference at less than 10 watts, according to management.
Lasserre cautioned that it remains uncertain whether another chip revision will be necessary before commercial production. He said GSI is in discussions with potential strategic partners that could help it enter markets it does not currently address, but no agreement has been signed. An agreement might be reached by the end of 2026 or in early 2027, he said.
About GSI Technology (NASDAQ:GSIT)
GSI Technology, Inc is a semiconductor company that develops and supplies high-performance memory and processing solutions. Its product portfolio has historically included static random-access memory (SRAM) devices, including synchronous SRAM products designed for applications requiring fast data access and high bandwidth.
The company also develops associative processing technology for data-intensive computing. Its Gemini Associative Processing Unit (APU) platform is designed to accelerate workloads such as artificial intelligence inference, natural-language processing, similarity search and other applications that involve searching large data sets.
