This Tuesday, AI chip startups secured over a billion dollars in funding, proving that investor enthusiasm remains strong despite AI bubble fears.
Leading the charge is MatX, a startup launched in 2022 by former Google engineers Reiner Pope and Mike Gunter. They clinched $500 million in series B funding, predominantly from Jane Street and Situational Awareness LP. MatX plans to release its LLM-optimized chip, the MatX One, this year, promising a versatile solution that handles pre-training, reinforcement learning, and inference.
While details on the chip are limited, MatX touts its unique split systolic array, promising high FLOPS per mm2 and scalability to vast chip networks.
In addition, Dutch firm Axelera raised $250 million for its low-power RISC-V AI accelerators targeted at edge computing and data centers. Unlike its peers, Axelera aims to deliver solutions for power-constrained environments like computer vision and robotics, with ambitions to scale towards comprehensive AI/ML functionality.
Elsewhere, SambaNova attracted $350 million from Intel among others, to advance its dataflow accelerators. Their SN50, set for deployment by SoftBank in Japan, exemplifies the broad industry support for non-GPU based AI solutions.