Ever heard of it?
GSI Technology is in a high-risk transformation phase. The company is transforming itself from a niche supplier of memory chips (SRAM) to an innovator in AI hardware.
1. technological spearhead: Gemini & Plato
The core of the analysis is the Associative Processing Unit (APU). In contrast to Nvidia's GPUs, GSI uses a "compute-in-memory" concept.
- Gemini-II: The current generation, which, according to a study by Cornell University (Oct. 2025), offers the performance of high-end GPUs with 98 % less energy consumption energy consumption.
- Plato chip: The next big target. A chip that is specially designed for Large Language Models (LLMs) at the edge (locally on devices). The "tape-out" (start of production of the design) is planned for early 2027.
2. areas of application: Where GSI wants to win
GSI avoids the direct battle in data centers and focuses on edge AI marketswhich, according to market forecasts, could grow to over USD 260 billion by 2035:
- Defense & Space: Real-time object recognition in drones and satellites (SAR imaging), often in challenging conditions without GPS.
- Cybersecurity: High-speed vector search to detect attack patterns in milliseconds.
- Autonomous systems: Robots and drones that rely on the extreme energy efficiency of the APU due to limited battery capacities.
@Multibagger
@Tenbagger2024
@EpsEra
@Klein-Anleger
@Aktienfox
That's what Gemini spit out for me once.
The last capital increase was in October
Break-even should be reached in 2027.
I find the whole thing really exciting if it can really hold a candle to an Nvidia chip, at least in some disciplines.
Perhaps also exciting for a takeover of a hyperscaler.
The figures will be published on January 22.