“From Chips to Drones: 11 Startups Accelerate Robotics and Data Center Innovation in IC Taiwan Challenge”
- ashley19241
- 10 hours ago
- 3 min read
The National Science and Technology Council (NSTC) has announced the winners of the 4th batch of the IC Taiwan Grand Challenge (ICTGC). A total of 11 startups from the United States, the United Kingdom, France, Germany, Japan, Israel, and Taiwan were selected this round. Their expertise spans across chip design verification, IP development, key components, data center infrastructure, as well as specialized intelligent applications in areas such as smart buildings and autonomous vessels.
These innovations comprehensively align with the AI industry value chain and are expected to inject strong momentum into Taiwan’s initiative to drive innovation across industries.
Selected Teams Overview:
Chip Design Verification
France-based Aniah focuses on transistor-level verification for advanced analog and mixed-signal SoCs. Its vectorless verification engine, OneCheck, performs comprehensive analysis of all electrical states within power domains, enabling rapid detection of critical corner-case errors. Built on OneCheck, the large language model-based Amigo AI interprets issues, suggests fixes, and improves first-pass silicon success rates.
U.S.-based Leafy Lab leverages AI to enhance analog and mixed-signal chip design processes. By integrating silicon physics with explainable AI models, it significantly reduces iterative trial-and-error cycles and verification costs, shortening design iterations from months to just days. Its platform predicts device behavior within approximately 2% error and automates analog layout and optimization, aiming to reduce chip development time by 50% and improve first-pass success rates.
Israel's Caesarea Labs specializes in automating post-silicon validation (PSV). As advanced nodes and AI chips increase complexity, PSV has become a major bottleneck due to its reliance on manual processes, long debugging cycles, and rising costs. Caesarea Labs introduces modern EDA automation into this critical yet underdeveloped stage, transforming circuit editing—from fragile and slow workflows into reliable, repeatable, and efficient industry standards.
U.S.-based Vellex Computing develops programmable analog AI accelerator IP that integrates with standard MCU cores such as Arm and RISC-V. It enables real-time AI optimization and on-device adaptive learning for edge devices such as drones, wearables, and sensors under low-power conditions, reducing dependence on cloud computing and high-power GPUs.
Key Components
Japan's TopoLogic develops TL-RAM, a next-generation magnetic memory based on topological materials. With ultra-low power consumption and high-speed performance, it is particularly suitable for AI processors and high-performance computing (HPC), addressing the growing energy demands of AI and data centers.
U.S.-based Oculi introduces AI vision sensors and processors (SPU) that transform each pixel into an intelligent programmable sensing unit. By processing only essential data, it significantly reduces power consumption, bandwidth requirements, and system costs—enabling ultra-low latency AI vision in edge devices such as smartphones, smart home sensors, robotics, and industrial automation.
Taiwan's ROCKCORE TECHNOLOGY CO., LTD offers an AMD CPU + FPGA heterogeneous computing platform (Advanced AI-FPC). By integrating CPU, GPU, and FPGA into a unified architecture, it addresses challenges in defense and advanced semiconductor equipment requiring micron-level precision, ultra-low latency, and high integration—such as weapon tracking and lithography alignment systems.
Data Center Infrastructure
UK-based Infiniflux has developed a two-phase direct-to-die liquid cooling system, combining proprietary two-phase cooling technology with direct die cooling. Even at chip power levels exceeding 1000W, it maintains GPU temperatures at 15–20°C, boosts computing performance by 15%, significantly reduces energy and maintenance costs, and extends chip lifespan by at least two times.
Germany's Linque develops all-optical circuit switching technology based on large-scale photonic integrated circuits. By eliminating power-intensive electro-optical conversion, it enables ultra-fast, energy-efficient switching with nanosecond reconfiguration—supporting high-bandwidth, low-power AI-driven data center networks.
AI Applications in Specialized Fields
U.S.-based ThermoVerse utilizes advanced phase change materials (PCM) with precise parameter control to develop thermal controllers capable of managing charge/discharge cycles. Through programmable thermal storage, it reduces HVAC energy consumption by 20–60%.
Compared to autonomous vehicles on land, U.S.-based STR8 Industries focuses on autonomous navigation software for surface vessels and onboard system integration. It aggregates data from distributed multi-node sensors to build a unified real-time environmental model, continuously tracks targets, assesses collision risks, and generates COLREGs-compliant avoidance actions with full audit trails.
Award Ceremony & Technology Showcase
To recognize the achievements of the 4th batch winners and showcase their innovative technologies and future application potential, ICTGC will host a themed pavilion at InnoVEX 2026 (Booth S0724) at Taipei Nangang Exhibition Center Hall 2, 4th Floor.
The Award Ceremony and Innovation Demo Day will take place on June 2 (Tuesday), 15:00–17:30, at the InnoVEX Center Stage. The event will bring together industry experts and investors, providing a platform for interaction with startups and creating opportunities for exposure and collaboration.
Event page:
Source of Information: Internet of Intelligent Agent Association (IIAA)
Comments