The company's goal is to change the math on the chip, a significant task that involves using a logarithmic approach instead of a floating point approach. This change could make processing on large language models less expensive. The software part of the stack is expected to be available in Q3 next year, with the hardware to follow in the coming years.
Key takeaways:
- Lemurian Labs, a startup founded by Google, Intel and Nvidia alumni, is aiming to build a chip that provides similar power to Nvidia's GPU chips but at a lower cost.
- The company recently announced a $9 million seed investment led by Oval Park Capital.
- Lemurian's approach involves changing the math on the chip and minimizing the distance data has to travel, essentially moving compute to the data instead of the other way around.
- The company plans to release the software part of the stack first, with a general availability targeted for Q3 next year, and the hardware to follow in the coming years.