Jun. 13 at 3:32 PM
$AIP is another interesting one. Core thesis here is AI is creating a big bottleneck in data-movement inside chips, and Arteris is in the middle of it.
$AIP’s Network-on-chip (NoC) IP is designed to move data efficiently between CPU’s GPU’s NPU’s HBM Memory and Chiplets. Every hyperscaler is developing custom silicon
$GOOGL - TPU
$AMZN - trainium
$MSFT - Maia
$META - MTIA. Every one of these chips needs an on chip network that moves data efficiently. The more custom chips are built. The more TAM
$AIP gets. New age AI chips contain massive memory systems, Chiplets, interconnects. Highly complex and needs a sophisticated internal architecture to move data efficiently. This is where
$AIP comes in. There is a surging demand for its NoC platform. semi conductor companies moving away from single die architecture to multi die (Chiplets) architecture which create even more complex communications inside chips.