A number of semiconductor companies are working feverishly to jam Artificial Intelligence functions onto a single chip.
Recently, Israeli startup NeuroBlade joined the fray. The company announced last week it had completed a Round A of financing for $23 million. Marius Nacht, co-founder of Check Point Software, led the round with participation of Intel Capital and previous investors StageOne Ventures and Grove Ventures.
The new funding will be used with $4.5 million previously raised to ramp up hiring and marketing to bring the first generation of the NeuroBlade AI chip to market.
The company was founded in 2017 by CEO Elad Sity and CTO Eliad Hillel, both graduates of the technological unit of Israel’s Intelligence Corps and former employees of SolarEdge.
AI chips are already used today in a number of apps, including image and speed recognition, video analytics, and autonomous driving. Nvidia and Intel are the current leaders, but at least 15 smaller companies are working to develop processors for different needs. Even large enterprises are developing their own specialized AI processors, according to analysts and NeuroBlade.
Some of the latest work with AI puts the intelligence at the edge of enterprise or carrier networks, so that decisions and inferences about data can be made at the edge instead of sending unneeded data to a centralized data center or even to the cloud. One example is how AI might help decide the most pertinent video in data-rich video stream seen by security or monitoring cameras. If a camera is constantly surveying a parking garage and a person walks by, AI and sensors can retain the video of the person for those few seconds that the person walked by for further analysis, rather than all the video data for hours or days.
RELATED: “AI at the Edge”
NeuroBlade has designed its AI chip to overcome pricing and performance concerns of existing chips. “It maintains a strong performance level despite being smaller and less expensive to make,” the company said in a statement. It is designed to run several neural networks and multiple complex algorithms at the same time.
“The chip developed by NeuroBlade will allow running AI algorithms on par with the performance of market-leading chips, but at a smaller size or with significantly better performance at similar throughput and size,” said Tal Clobodkin, partner at StageOne VC. NeuroBlade’s technology is targeted at servers, but is also relevant to devices like laptops, cars, and camers, he added.