Chip startup Taalas raises $169 million to help build AI chips to take on Nvidia

Toronto-based AI chip startup Taalas has secured $169 million in funding to develop specialized chips designed to run AI applications faster and at a lower cost than Nvidia’s offerings. Founded about two years ago, Taalas focuses on creating customized silicon where AI models are essentially “hard-wired” directly into the chip’s transistors. This approach bypasses the conventional flexibility of software-driven compute engines, allowing for significant performance and efficiency gains in AI inference workloads. By tailoring chips specifically for particular AI models, Taalas aims to deliver higher throughput—achieving up to 10 times the tokens per second compared to current mainstream solutions—with drastically reduced production costs, reportedly 20 times lower in some cases.

The company is targeting large language models such as Meta’s Llama 3.1 and DeepSeek’s AI systems, with plans to have a 20-billion parameter model hardcoded onto its HC chip by mid-2026. Taalas’ innovation addresses the growing demand for faster, more energy-efficient AI hardware as models scale rapidly in size and complexity. The startup’s novel chip design strategy challenges dominant players like Nvidia, which currently leads the AI hardware market with GPU-based accelerators.

This fundraising success and unique technology position Taalas as a notable contender in the fiercely competitive AI chip landscape, highlighting the ongoing shift toward specialized semiconductor solutions tailored for AI workloads rather than general-purpose hardware.