Home PC News Syntiant raises $35 million for AI speech-processing edge chips

Syntiant raises $35 million for AI speech-processing edge chips

Syntiant, a startup growing AI edge {hardware} for voice and sensor options, as we speak closed a $35 million spherical. CEO Kurt Busch says the funds will probably be used to ramp up manufacturing all through the rest of 2020.

According to a report printed by Meticulous Research, the speech and voice recognition {hardware} market is anticipated to succeed in $26.eight billion by 2025. That’s as a result of units like good audio system, good shows, telephones, headphones, listening to aids, and laptops require low-power chips to course of utterances. While some system-on-chip choices sport coprocessors to deal with voice recognition, they’re usually not in a position to accommodate a number of kind elements.

Three-year-old Syntiant, which is headquartered in Irvine, California, gives {hardware} that merges machine studying with semiconductor design for always-on voice purposes. The firm’s Neural Decision Processors with Alexa Voice Service firmware supply wake phrase, command phrase, and occasion detection in a bundle with a claimed “near-zero” energy consumption. In truth, Syntiant says its processors make units roughly 200 instances extra environment friendly by offering 20 instances the throughput of competing microcontrollers.

Syntiant’s N100 and N101 processors measure about 1.four millimeters by 1.eight millimeters and may run fashions with over half one million parameters. Packing a general-purpose ARM Cortex-M0 processor paired with 128KB of RAM, the chips eat lower than 140 microwatts and allow onboard firmware safety and authentication, key phrase coaching, and as much as 64 output classifications.


The N100 and N101 — which initially focused efficiency of round 20 TOPS (trillion floating-point operations) per watt — use tons of of hundreds of flash reminiscence NOR cells that learn and write information one phrase or byte at a time. The processor-in-memory structure was proposed by CTO Jeremy Holleman, a researcher on the University of North Carolina in Charlotte, way back to the 2014 International Solid-State Circuits Conference. Syntiant asserts that the structure is right for executing massively parallel operations in deep studying at low energy.

Syntiant first demonstrated a prototype at Infineon’s OktoberTech convention in 2018. The firm started transport buyer samples later that 12 months and shipped its first merchandise in September 2019, a month after receiving Amazon Alexa {hardware} certification.

Syntiant’s Neural Decision Processors ship with a software program growth equipment (SDK) that integrates with nearly any built-in growth atmosphere and a “training development kit” (TDK) that lets builders use Google’s TensorFlow machine studying framework and different machine studying toolsets. (The TDK helps the Neural Decision Processors’ distinctive {hardware} traits. Developers prepare AI fashions on Google’s AI cloud service and obtain the outcomes it produces to the chip by way of the TDK.) The SDK and TDK embody a {hardware} mannequin that may be built-in right into a coaching pipeline, together with step-by-step directions for server configurations and Alexa wake phrase neural community fashions.

M12 (previously Microsoft Ventures) and Applied Materials’ Applied Ventures led this week’s sequence C funding spherical, with participation from new buyers Atlantic Bridge Capital, Alpha Edison, and Miramar Digital Ventures. This brings the corporate’s whole raised to over $60 million. As part of the spherical, Applied Ventures funding director Michael Stewart joined Syntiant’s board, alongside Atlantic Bridge Capital’s David Lam, who will function a board observer.

Sign up for Funding Weekly to begin your week with VB’s prime funding tales.

Most Popular

Recent Comments