Upstart chipmakers keep challenging Nvidia (www.cnbc.com)

D-Matrix, a startup based in Silicon Valley, has announced the launch of its Corsair chip, claiming it can perform AI inference workloads up to 10 times faster and five times more energy-efficient than Nvidia's standalone GPUs. This assertion highlights a growing trend where smaller companies are attempting to carve out their niche in the competitive AI chip market, traditionally dominated by Nvidia. D-Matrix's innovative approach integrates memory and compute on a single chip using SRAM technology, allowing for low-latency inference, particularly for applications like chatbots and voice agents, while addressing a significant challenge in the market — the shortage of DRAM. With a valuation of around $2 billion and significant backing from investors, including Microsoft, D-Matrix is set to ship its chips to notable hyperscale customers this month. While the Corsair chip excels in efficiency and speed for smaller workloads, it faces limitations handling large models with trillions of parameters, a realm where Nvidia continues to excel. The introduction of D-Matrix’s Corsair sets the stage for an evolving battle in AI hardware, as evidenced by its strategic partnerships, including potential collaborations with Nvidia GPUs to enhance performance, signaling a rising tide of innovation and competition in the AI/ML landscape.