Next in line to challenge Nvidia: Taalas hardwires Llama into silicon, claims 17,000 tokens per second

Levi Li, DIGITIMES Asia, Taipei Wednesday 25 February 2026 0

Credit: Taalas

Toronto-based AI chip startup Taalas says it can hardwire a large language model directly into silicon to accelerate inference beyond what conventional GPUs can deliver. Founded in 2023, its first product — the HC1 inference chip — generates...

The article requires paid subscription. Subscribe Now

Email address

Password

Save my User ID and Password

Save my User ID and Password

Some subscribers prefer to save their log-in information so they do not have to enter their User ID and Password each time they visit the site. To activate this function, check the 'Save my User ID and Password' box in the log-in section. This will save the password on the computer you're using to access the site.

Note: If you choose to use the log-out feature, you will lose your saved information. This means you will be required to log-in the next time you visit our site.

NEW individual users please REGISTER first.

Enterprise user first-time LOGIN.

I forgot my ID and/or password?

MEMBERSHIP BENEFITS

Full access to articles dating back to 2000.
Real-time access to the news as it breaks.
Access to premium content.
Full access to Tomorrow's Headlines.
Asia Supply Chain 250 dataset access.