Intel Invests In AI Chip Startup To Boost Machine Learning In Edge Devices

Intel is the lead investor in an artificial intelligence chip startup building a low-power semiconductor that can bring machine learning to edge devices.

Syntiant, based in Irvine, Calif., announced on Tuesday it has raised a Series A financing round led by Intel to help commercialize its Neural Decision Processor, which uses "custom analog neural networks to optimize deep learning functions at the transistor level." CEO Kurt Busch declined to comment on the round's amount but confirmed that the Series A corresponded with a U.S. Securities and Exchange Commission filing from last fall showing that the company had raised $5 million.

Intel's investment in Syntiant comes as the Santa Clara, Calif.-based semiconductor giant faces growing competition from large companies and upstarts alike that are building chips specifically suited for AI applications. Facebook and Amazon, for instance, are reportedly building their own AI chips while dozens of semiconductor startups are raising hundreds of millions.

"We believe ultra-low-power, analog neural networks like the ones Syntiant provides could dramatically boost the adoption of distributed AI," Wendell Brooks, president of Intel Capital, said in a prepared statement.

Intel has already cashed in on the AI chip startup boom to a certain extent. In 2016, the company acquired Nervana Systems, which was developing a specialized chip for deep learning and neural networks.

This isn't the first time Busch has worked with Intel. In the late 1990s, he worked for the company through its acquisition of a Digital Equipment Corporation business.

In an interview with CRN, the CEO said Syntiant is targeting its Neural Decision Processor for always-on, battery-powered devices that need to perform AI functions such as image recognition or keyword spotting without sending information to the cloud for processing. Syntiant's target devices include high-end mobile phones, fitness trackers, hearing aids, drones and security cameras.

"We don’t see anyone bringing this much neural capability to the machine learning space," he said.

Syntiant plans to sell the chip in two ways: as an application-specific standard product and as an application-specific integrated circuit. Busch said the ASSP has a larger opportunity because its designed for multiple use cases while the ASIC would be custom-designed for a single application.

Beyond providing financial assistance, Intel has also introduced Syntiant to "more than a dozen tier-one companies" that could end up using the startup's AI chip in their devices, Busch said. The company expects the first Syntiant-powered devices hit the market next year.

"They can offer market acceleration for any company they invest in," he said.

Jamie Shepard, a managing director at Accenture, told CRN "there's a lot of need and a lot of use cases" for edge devices to handle AI computations. For example, he said, AI at the edge could help a security camera system process images faster and provide recommendations on how to react in certain events.

"It's allowing us to drill in on very specific intelligent data in a lot less time," he said.

Healthcare is one area that could particularly benefit from AI edge computing, according to Shepard. For instance, if a patient had a wearable that could interpret symptoms at the edge rather than in the cloud, it could potentially give that patient a faster diagnosis.

"That's where we see a lot of benefits of embedding AI in devices where we can intelligently collect data, not just monitor it, but it will give us better data in real time," Shepard said.