Chip technology company Syntiant Corp. in Irvine has added a second-generation artificial intelligence processor to its lineup for audio and sensor applications in battery-powered devices, as the closely watched firm, backed by heavyweight tech investors, aims for a “significant amount” of revenue this year.
Syntiant’s low-power chips respond to voice and speech, and can wake up a device like an Amazon Alexa or have a device perform a specific function.
The chips are used to control smartphones, earbuds, smart speakers, laptops, AR glasses, wearables, TV remote controls and dozens of other smart products—an industry expected to boom with new Internet of Things applications.
The new NDP120 deep learning processor introduced last week performs functions, with minimal battery power consumption, for battery-powered devices at a performance level that most people would expect from devices plugged into a regular wall socket, Chief Executive Kurt Busch told the Business Journal last week at the time of the new product’s release, calling that the “real point” of the new development.
“We think voice is the next-generation interface,” Busch said. “And pervasive artificial intelligence will bring it about.”
Busch is optimistic about the potential for the NDP120. NDP stands for “Neural Decision Processor.”
“The 2022 market size, we think for this, is north of a billion dollars,” he said.
“We had millions of dollars in revenue last year and we’ll have a significant amount this year,” Busch told the Business Journal, without providing specific figures.
Syntiant’s Goals
Syntiant, formed in 2017, is aiming to be Orange County’s next big semiconductor company, with chips that make it extremely easy to add voice control to any Internet of Things connected device.
Busch explained that the earlier model known as the NDP100 is “ultra, ultra low-power and super simple.”
The new model, the NDP120, doesn’t just respond to voice and phrases, “it can [also] do noise cleanup, noise reduction, speech enhancement,” as well as run multiple sensors, he said.
The NDP120 “brings 25 times the performance and allows for parallel processing,” he said, with other uses including mobile phones, laptops, “hearable” products and medical devices.
Innovator of Year
Busch, a former president and CEO of Irvine’s Lantronix Inc. (Nasdaq: LTRX), was one of five winners of the Business Journal’s Innovator of the Year Awards last year.
In August, Syntiant said it had received $35 million in a Series C financing round led by Microsoft’s venture capital fund M12 and Applied Materials Inc.’s venture capital arm.
It has raised $65 million to date. Other investors include Intel Capital, Robert Bosch Venture Capital, the Amazon Alexa Fund and Atlantic Bridge Capital.
“We’re not looking for [additional] financing at this point,” Busch said on Jan. 5.Â
Industry observers expect great things from the company.
Syntiant’s innovation “around powerful, cost-effective, and yet power-efficient embedded silicon systems that bring machine learning capabilities right to the edge has been fascinating,” Arun Rajasekaran, head of technology strategy, wearables and sensors at Santa Cruz-based Poly (NYSE: PLT), a communications tech company valued at about $1.3 billion, said at the time of the new product announcement.
“By optimizing its silicon for its deep learning models, Syntiant is making neural processing the pervasive technology to interface between people and machines,” Rajasekaran said.
$6 Per Unit
Syntiant said in August it had shipped more than a million units of its processors for voice-activated control of electronic devices.
Busch said the number shipped now is “lots more than that” with an update expected this month.
“The Syntiant NDP120 is sampling now and will be shipping in production volumes in summer 2021,” the company said.
The processors cost $6 per unit for large-scale orders.
The NDP120 runs multiple applications simultaneously with minimal battery power consumption, including echo-cancellation, beamforming, noise suppression, speech enhancement, speaker identification, keyword spotting, multiple wake words, event detection, and local commands recognition, the company notes.
