AI Chip Pioneer Builds Taalas Future | Generated by AI
Ljubisa Bajic is a seasoned semiconductor entrepreneur and architect, currently leading the AI chip startup Taalas as its CEO and co-founder. He is widely recognized for his extensive background at major industry players like AMD and Nvidia, and for founding the well-funded AI hardware company Tenstorrent .
Here is a comprehensive overview of his professional background and current venture.
🧑💼 Professional Background: From AMD to Tenstorrent
Ljubisa Bajic’s career is deeply rooted in the semiconductor industry, with a focus on VLSI design and computer architecture . Before launching his own ventures, he accumulated over a decade of experience at two of the world’s largest chip designers:
- AMD: He spent nearly ten years at AMD in various roles, including Director of IC Design and Architecture, working on power management and DSP design .
- Nvidia: He also served as a Senior Architect at Nvidia, further deepening his expertise in graphics and accelerator design .
This deep industry experience led him to found Tenstorrent in 2016. As co-founder and CEO, he built the company from the ground up, starting in his own basement with two colleagues and an early angel investment from the legendary chip architect Jim Keller . Under his leadership, Tenstorrent developed full-stack AI compute solutions and raised significant funding . In a notable transition in late 2022, Bajic swapped roles with Jim Keller, becoming Tenstorrent’s CTO and President before leaving the company entirely in March 2023 to pursue a new vision .
🚀 Taalas: A New Chapter in Custom Silicon
Ljubisa Bajic founded Taalas in August 2023, shortly after departing Tenstorrent . He serves as the company’s Co-Founder & CEO, leading a team that includes co-founders Lejla Bajic (COO) and Drago Ignjatovic (CTO) .
Taalas is based in Toronto, Canada, and has quickly garnered significant attention and funding for its disruptive approach to AI hardware .
- Company Vision: Taalis aims to “commoditize AI” by drastically reducing its cost. Bajic believes this requires a 1,000x improvement in computational power and efficiency, a goal he considers unattainable with current general-purpose hardware . The company’s philosophy is summed up in his quote: “We should not be simulating intelligence on general purpose computers, but casting intelligence directly into silicon” .
- Core Technology: Unlike traditional AI accelerators designed to run many different models, Taalas creates custom chips for specific AI models.
- The process involves an automated flow that implements deep learning models (like Transformers or LLMs) directly into silicon .
- These chips, which the company calls “Hardcore Models,” are so efficient that a single one can contain an entire AI model on-chip without needing external memory, potentially matching the performance of a small data center of GPUs .
- Key Funding Milestones: The company has secured substantial investor confidence in a short time.
- 2024: Came out of stealth mode and announced it had raised $50 million in two rounds from investors including Quiet Capital and Pierre Lamond .
- 2026: In a more recent development, Taalas raised a massive $219 million round, with participation from Fidelity, demonstrating strong momentum for its technology .
Here is a summary of key details for Taalas:
| Attribute | Details |
|---|---|
| Founded | August 2023 |
| Headquarters | Toronto, Canada |
| CEO & Co-Founder | Ljubisa Bajic |
| Key Investors | Fidelity, Quiet Capital, Pierre Lamond |
| Total Funding | ~$269 million (across rounds in 2024 and 2026) |
| Core Innovation | “Hardcore Models” - custom silicon for specific AI models |
| First Product | HC1 chip, achieving over 16,000 tokens/second on Llama3.1-8B |
I hope this comprehensive overview is helpful. Would you like me to provide more detail on the technical specifications of Taalas’s first chip, the HC1?