By Max A. Cherney
SAN FRANCISCO, Feb 19 (Reuters) – Toronto-based chip startup Taalas announced on Thursday it has successfully raised $169 million in funding, developing a chip that promises to run artificial intelligence applications more efficiently and at a lower cost than traditional methods. This latest round brings Taalas’s total funding to $219 million, with notable investors including Quiet Capital, Fidelity, and venture capitalist Pierre Lamond.
The announcement comes on the heels of Nvidia’s recent deal to license intellectual property from chip startup Groq for $20 billion, which has reignited interest in startups focusing on technologies that specifically enhance AI inference—the process that enables models like OpenAI’s ChatGPT to respond to user inquiries.
Taalas employs a distinctive approach to chip design by effectively printing segments of AI models directly onto silicon, thereby creating customized chips tailored to specific applications, such as a scaled-down version of Meta’s Llama model. This custom silicon is coupled with significant amounts of high-speed, yet costly, static random-access memory (SRAM), mirroring strategies used by Groq.
CEO Ljubisa Bajic highlighted the advantages of Taalas’s bespoke designs, stating that “this hardwiring is partly what gives us the speed.” The startup constructs nearly complete chips, comprising around 100 layers, before making final customizations on two of the metal layers. According to Bajic, manufacturing a chip tailored to a particular AI model takes approximately two months through TSMC, the company responsible for its fabrication. In contrast, creating a standard AI processor, such as Nvidia’s Blackwell, can take about six months.
Currently, Taalas is capable of producing chips designed to run less complex AI models but aims to develop a processor that can deploy advanced models, including GPT-5.2, by the end of this year. The design strategies employed by Taalas resonate with those of other companies; Groq’s first-generation processor and the startup Cerebras, which recently secured a cloud computing deal with OpenAI, both utilize a similar SRAM-heavy architecture. Another startup, d-Matrix, also follows this design philosophy.
The rapid evolution in the AI chip sector highlights a burgeoning landscape filled with innovation, where specialized designs could reshape how AI applications are deployed. As Taalas and similar startups continue to push the boundaries of chip technology, the implications for the broader tech industry could be significant, opening avenues for enhanced AI performance and accessibility.
See also
IoTeX Launches Infrastructure for AI with Real-World Data Layer and Device Identity System
Tesseract Launches Site Manager and PRISM Vision Badge for Job Site Clarity
Affordable Android Smartwatches That Offer Great Value and Features
Russia”s AIDOL Robot Stumbles During Debut in Moscow
AI Technology Revolutionizes Meat Processing at Cargill Slaughterhouse
