USC and TetraMem Develop Memristor Chip Operating at 700°C for Extreme AI Computing
Researchers at the University of Southern California have developed a memristor memory chip capable of operating reliably at 700 degrees Celsius, significantly exceeding the thermal limits of previous electronics used in extreme environments like Venus. Published in Science, the device utilizes a unique three-layer structure of tungsten, hafnium oxide, and graphene to prevent atomic migration and short-circuiting under intense heat. The chip demonstrated stability for over 50 hours without data refresh and survived billions of switching cycles. TetraMem, the startup commercializing this technology, is already producing room-temperature AI inference chips on 300mm wafers in partnership with SK hynix, supported by the CHIPS Act. Unlike conventional GPUs that suffer from the von Neumann bottleneck, these memristors perform in-memory computing, executing matrix multiplication physically through Ohm’s Law. This architecture offers vastly superior speed and energy efficiency for AI workloads. While the high-temperature capability opens possibilities for deep-space exploration and industrial applications, the immediate commercial focus remains on revolutionizing AI inference efficiency. The breakthrough highlights a significant advancement in semiconductor materials science, leveraging standard industry materials alongside emerging graphene technology to overcome longstanding thermal and architectural limitations in computing hardware.
Editorial responsibility
- No named human review is recorded for this page.
- Reports are grouped by semantic similarity and deterministic rules. Language models may assist titles, summaries, translation and cross-source analysis; the page itself is projected from evidence records.
- Current automated evidence projection
Wire timeline
USC and TetraMem Develop Memristor Chip Operating at 700°C for Extreme AI Computing
Researchers at the University of Southern California have developed a memristor memory chip capable of operating reliably at 700 degrees Celsius, significantly exceeding the thermal limits of previous electronics used in extreme environments like Venus. Published in Science, the device utilizes a unique three-layer structure of tungsten, hafnium oxide, and graphene to prevent atomic migration and short-circuiting under intense heat. The chip demonstrated stability for over 50 hours without data refresh and survived billions of switching cycles. TetraMem, the startup commercializing this technology, is already producing room-temperature AI inference chips on 300mm wafers in partnership with SK hynix, supported by the CHIPS Act. Unlike conventional GPUs that suffer from the von Neumann bottleneck, these memristors perform in-memory computing, executing matrix multiplication physically through Ohm’s Law. This architecture offers vastly superior speed and energy efficiency for AI workloads. While the high-temperature capability opens possibilities for deep-space exploration and industrial applications, the immediate commercial focus remains on revolutionizing AI inference efficiency. The breakthrough highlights a significant advancement in semiconductor materials science, leveraging standard industry materials alongside emerging graphene technology to overcome longstanding thermal and architectural limitations in computing hardware.