Cerebras WSE-3: Compute Acceleration for very large LLMs

And how LLMs are becoming more and more prevalent in biotech

AI supercomputer firm Cerebras announced its next generation of waferscale AI chips that can double the performance of the previous generation while consuming the same amount of power. The Wafer Scale Engine 3 (WSE-3) contains 4 trillion transistors, 50+ percent increase over the previous generation. It is the same waferscale concept but with better silicon technology.

The WSE-3 can be used for the new generation of AI computers, e.g. the ones being installed in a datacenter in Dallas to form a supercomputer capable of 8 exaflops. Cerebras also announced that they have signed a joint development agreement with Qualcomm to boost the performance for AI inference 10-fold.

This is the third iteration of the waferscale chip concept: square-shaped with 21.5 centimeters to a side, it uses nearly an entire 300-millimeter wafer of silicon to make one chip.