Taalas Raises $169M to Bake AI Models Directly Into Silicon
Toronto chip startup Taalas hardwires AI models into custom chips for faster inference, pulling in $169M in fresh funding.
Toronto-based Taalas just pulled in $169 million in new funding, pushing its total raise to $219 million. The startup's pitch is bold: instead of running AI models on general-purpose chips, it hardwires them directly into custom silicon.
The approach targets inference speed — the part of AI that actually matters when models are deployed in production. By baking a model's architecture into the chip itself, Taalas aims to blow past the performance ceilings of conventional hardware.
It's a big bet. Custom silicon means less flexibility, but potentially massive gains in speed and efficiency for specific AI workloads. With nearly a quarter-billion dollars in total funding, Taalas clearly has investors convinced the tradeoff is worth it.
The company says it has already developed a working chip. Now comes the hard part: proving it scales.