Following the success of EdgeThought, Skymizer’s revolutionary LPU for on-device LLM inferencing, the company has launched HyperThought, its next-generation AI accelerator IP. HyperThought is engineered to meet the increasing demands of real-time, multimodal, and agent-based AI, offering unmatched scalability, performance, and deployment flexibility for developers.
- Build Custom AI Chips with HyperThought LPU IP
- Empowering SoC Teams: HyperThought is not just an accelerator; it’s a fully integrable subsystem that allows developers to design custom AI silicon tailored for their unique applications.
- Domain-Specific Intelligence: Developers can integrate domain-specific intelligence with their host CPU and edge subsystems to create highly optimized AI solutions.
- HyperThought’s Features
- LISA v3 Architecture: HyperThought is powered by Skymizer’s Language Instruction Set Architecture (LISA) v3, supporting multimodal processing, dynamic graph control, and advanced memory operations.
- Multimodal Processing: Seamless execution of text, vision, and instruction-based models across shared compute resources enables efficient use of resources.
- Concurrent Multi-Model Support: The LPU supports parallel execution of multiple quantized LLMs, ensuring low-latency responses across a wide range of applications.
- Agentic AI Infrastructure: HyperThought is designed to handle persistent, goal-driven agents capable of interactive and contextual reasoning, enabling real-time, autonomous decision-making.
- Optimized Software Stack: The co-designed compiler-runtime platform strikes a perfect balance between compute throughput, memory, and bandwidth constraints.
- Scalability and Integration
- Modular and Compact IP Block: HyperThought is highly scalable, with a modular IP block that facilitates easy integration into mobile, embedded, and automotive-grade SoCs.
- Expert Insights
- William Wei, CMO of Skymizer: “HyperThought makes the edge intelligent. With LISA v3, we’re enabling intelligent agents that can reason, interact multimodally, and adapt in real time—paving the way for truly autonomous edge AI.”
- Luba Tang, CTO of Skymizer: “LISA v3 allows us to support real-time intelligent systems that need to react, infer, and adapt on the edge. HyperThought is a foundational step toward truly autonomous, on-device AI.”
- Join Skymizer at COMPUTEX 2025
- Showcasing HyperThought: Skymizer will be showcasing HyperThought at COMPUTEX Taipei from May 20-23, 2025. Attendees can experience FPGA-verified IP powering multimodal inference and agent-based workflows.
- Technical Sessions and Previews: The company will also host technical sessions on LISA v3 and SoC integration, along with previews of HyperThought’s software tools and developer stack.
- 1F Power Chip Booth: I-1111 (PSMC Booth)
- 4F RISC-V Booth: L-0425 (RVTA Booth)
HyperThought represents a major leap forward in AI accelerator technology, empowering developers to build custom, intelligent edge AI systems with real-time multimodal capabilities and agent-based workflows. Skymizer’s latest offering combines cutting-edge hardware and software tools for the next generation of autonomous AI.
