Riding the sting artificial intelligence (AI) boom, startup Cima attracted large-scale investment. Based on this, the plan is to hurry up the event of multimodal edge AI chips.
TechCrunch reported on the 4th (local time) that Cima, a startup developing AI processors for edge devices, secured $70 million (about 100 billion won) in recent funding.
In accordance with this, Dell Technology Capital, Amplify Partners, and renowned chip industry executive Libbutan participated on this investment led by Maverick Capital. Cima, now in its fifth 12 months of business, has gathered $270 million (roughly 360 billion won) through this funding.
Cima is specializing in developing 'MLSoC', a system-on-chip that could be used to equip edge computing devices with AI functions. He has now released the primary generation of MLSoC chips focused on computer vision, and plans to make use of the brand new funding to develop the second generation MLSoC.
The primary generation MLSoC combined six different computing modules right into a single package, performing encoding or running computer vision algorithms on as much as eight video streams directly.
As well as, the corporate also provided a code-free software tool called 'Palette' that makes it easier to jot down software for MLSoC chips. First-generation MLSoC achieved the best results across closed, edge and power categories within the recent MLPuff Inference 4.0 benchmark.
Second-generation MLSoC supports any framework, network, model, or sensor for edge AI applications, in addition to all modalities including audio, voice, text, and pictures. It’s described as a single edge AI platform that runs all AI models, including computer vision, transformers, and multimodal generative AI.
The chip, scheduled to be released in the primary quarter of 2025, can be manufactured using TSMC's 6-nanometer process.
Meanwhile, as demand for AI chips has recently expanded from GPUs for existing data centers to chips for on-device AI or edge AI, development competition amongst edge semiconductor firms appears to be intensifying.
Not only NVIDIA's major competitors akin to Intel and AMD, but in addition AI chip startups akin to Halo are specializing in developing semiconductors for the sting.
Reporter Park Chan cpark@aitimes.com