Gemini-II delivers ultra-low power and industry-leading time-to-first-token performance, ideal for real-time drone workloads
GSI Technology, Inc. (Nasdaq: GSIT), the inventor of the Associative Processing Unit (APU), a paradigm shift in artificial intelligence (AI) and high-performance compute (HPC) processing providing true compute-in-memory (CIM) technology, today announced its defined edge strategy for the Gemini-II APU. GSI Technology’s focus on high-growth AI edge processor markets where its architecture delivers a decisive advantage in performance and power efficiency, beginning with drones, represents a segment projected to reach $2.7 billion by 20301.
The APU architecture is applicable to both edge and data center inference. However, the data center market is saturated with large, well-capitalized incumbents, where power consumption approaches 2kW per GPU. The edge needs an entirely different compute architecture. Gemini-II delivers complex edge AI capability at 15W, which is required for drones, defense systems, robotics, and mobile platforms where competitors simply cannot compete.
“Following our $50 million equity raise, GSI is advancing its roadmap to capture opportunities in edge markets, where our CIM architecture delivers meaningful improvements in power efficiency, latency, and on-device intelligence,” said Lee-Lean Shu, Chairman and Chief Executive Officer of GSI Technology. “At the edge, Gemini-II delivers GPU-class performance at a fraction of the power, enabling real-time responsiveness in power- and size-constrained environments. Its faster time-to-first-token offers a strong advantage in drone and defense applications, where milliseconds and mission endurance matter. Results from our current proof-of-concept engagements reflect this advantage, with customers seeing first-response times up to three times faster than alternative solutions.”
The global edge AI processor market is projected to reach $9.6 billion by 2030, according to third-party research2. As AI transitions from broad data center deployments to purpose-built workloads at the edge, a new growth frontier is developing across markets that demand high volumes of compact, power-efficient devices. Leveraging its established relationships with defense agencies and contractors, and the unique advantages of its architecture for AI applications in these domains, GSI is prioritizing early edge AI deployment in the drone and military vehicle markets, where it sees immediate need and demand.
Mr. Shu continued, “Our next-generation APU, Plato, will position GSI to participate in the broader wave of AI deployment, further penetrating embedded edge AI applications. We believe our progress with Gemini-II establishes a strong foundation for long-term growth and shareholder value creation.”
A recent Cornell University study confirmed that GSI’s APU architecture achieves GPU-class performance with more than 98% lower energy consumption through its memory-centric design. The research also demonstrated that the APU performs retrieval tasks several times faster than conventional CPUs, reducing total processing time by up to 80%—underscoring its potential to transform power-sensitive AI workloads.
Explore AITechPark for the latest advancements in AI, IOT, Cybersecurity, AITech News, and insightful updates from industry experts!