Taalas Lands $50 Million to Revolutionize AI with Advanced Silicon Tech

Led by seasoned experts in the field of silicon technology, Taalas Inc. is set to transform artificial intelligence (AI) by dramatically reducing costs through advanced direct-to-silicon processes, allowing a single chip to surpass the capabilities of a small traditional GPU data center.

Taalas Inc., a pioneering firm in the AI silicon industry, has emerged from the shadows with an impressive $50 million in funding amassed over two financing rounds overseen by industry veterans Pierre Lamond and Quiet Capital.

In recent times, AI has transitioned from conceptualization to widespread practical use, revolutionizing various sectors. The computational demands of deep learning models, which are integral to AI, have grown exponentially, resulting in systems that are both financially and energetically demanding, and are limited by the current GPU technology.

"The ubiquity of AI is comparable to the indispensability of electric power—it's becoming a necessity. To democratize AI, we're looking at needing a thousandfold increase in processing power and efficiency, beyond what incremental enhancements can offer. That's why we are shifting our approach from emulating intelligence on conventional computing systems to embedding intelligence directly into silicon," explains Ljubisa Bajic, CEO of Taalas.

Taalas is in the process of developing a unique automated system that will rapidly integrate various deep learning models directly into silicon. This method eliminates the need for external memory by housing complete AI models within Taalas’ chips themselves. The intrinsic efficiency of these specialized chips is set to exceed the performance of smaller GPU data centers, unlocking the potential for a cost improvement in AI by three orders of magnitude.

"Our support for Taalas' 'direct-to-silicon' foundry is rooted in our belief that it represents three major advances: a radical reduction in AI's current cost structure, the facilitation of substantial model growth, and the ability to operate sophisticated models on consumer electronics locally," states Matt Humphrey, a partner at Quiet Capital.

Taalas plans to have its first large language model chip ready by the third quarter of 2024 and to start offering it to select customers in early 2025.

"Creating sophisticated chips is a complex and high-stakes pursuit that demands a blend of expertise, creativity, and tenacity. The founders of Taalas—esteemed for their long history of producing innovative chips and systems—are unparalleled in the industry," remarks Pierre Lamond, revered for his foundational impact on the semiconductor field.

The establishment of Taalas is the work of Ljubisa Bajic, Drago Ignjatovic, and Lejla Bajic who bring to the table a wealth of experience in AI processors, GPUs, and CPUs with previous stints at acclaimed companies like Tenstorrent, AMD, and NVIDIA.

AI, funding, technology