Taalas, an AI chip startup with its headquarters in Toronto, Canada, has proudly declared the successful completion of a $169 million financing round. This brings the company's cumulative funding to exceed $219 million. Simultaneously, Taalas has unveiled its inaugural AI inference chip, the HC1. This chip is specifically tailored for Meta's Llama 3.1 8B model. Manufactured using TSMC's advanced 6nm process, the HC1 chip leverages 'hardwired' technology to embed model weights directly into the silicon. As a result, it achieves an impressive inference speed of 17,000 tokens per second, all while maintaining a remarkably low power consumption of just 200 watts per card. Looking ahead, Taalas has ambitious plans to release a processor capable of supporting large models like GPT-5.2 by the end of 2026. By eliminating the traditional barriers between memory and computation, Taalas's innovative technology substantially cuts down on hardware costs and energy consumption.