It Calls Itself an Edge-AI Company.
About 95% of Its Revenue Is Microphones.

On July 6, 2026, Syntiant filed its S-1 with the SEC. It is a company that makes edge-AI inference chips, backed by Intel Capital and Microsoft. Its banner is “Physical AI” — not the cloud, but AI that perceives sound and vibration inside a device and decides on the spot.

But a few pages into the S-1, the business looks different. Of the roughly $64.5 million in revenue for the most recent quarter (January–March 2026), the AI segment is just 5.1%. The remaining ~95% comes from a sensor business — mostly microphones — acquired at the end of 2024.

The engine of that jump is the acquisition. In 2024, before the deal, Syntiant’s own revenue was only about $13.6 million. In 2025 it reached about $271.8 million — roughly a 20x jump in a single year. Almost all of the increase is attributable to the acquired business.

At the same time, its earning power thinned. Gross margin fell from 46.6% before the acquisition to 19.7% after. A high-margin, small AI business gained a low-margin, large microphone business, and the shape of the whole company changed. That shape is what the S-1’s numbers show.

Ten questions. We lay out only what the S-1 says.

Question 1

What is this company, really?

In one line: a maker of edge-AI inference chips — or so it was supposed to be.

Syntiant was born in 2017 in Irvine, California, a semiconductor startup. What it raises as its banner is Physical AI: AI that perceives real-world sound and vibration inside a device and decides on the spot.

“

Syntiant is a semiconductor and software company focused on enabling Physical AI, which we define as on-device sensing and neural inference that allow devices to perceive, process, and respond to real-world inputs.

SYNTIANT S-1 (JULY 2026) — PROSPECTUS SUMMARY

At the core is the NDP (Neural Decision Processor) — a small brain for making decisions. Unlike the giant AI of the cloud, it runs on milliwatts of power and completes inference entirely inside a battery-powered device.

The company’s origin is a line from CEO Kurt Busch’s youngest son: “We just talk to it.” That single remark, the founder letter says, became a conviction that computing was moving toward natural conversation by voice.

But the identity of this company shifts a few pages into the S-1. In December 2024, Syntiant bought the MEMS microphone business of a maker called Knowles for about $114.4 million. A MEMS microphone is a microphone the size of a grain of rice that goes into phones and earbuds. In the S-1 the segments are called Digital Sensors and Analog Sensors, but the central product is this MEMS microphone, so this article refers to those businesses collectively as the microphone business.

Since then, the microphone accounts for about 95% of revenue in the most recent quarter. A company that raises the banner of AI inference chips now makes most of its revenue from microphones.

It was founded by four people — Busch, Vorenkamp, Bailey, and Holleman. All four remain active nine years later, an unusual founding team.

A company flying the flag of AI inference chips now makes ~95% of its revenue from microphones.

Question 2

What does it compete on?

Its battlegrounds: power efficiency, a full stack, and a customer base it bought.

Syntiant’s way of competing comes down to three pillars: outstanding power efficiency, vertical integration from sensor to software, and the bundle of customers it gained by buying the microphone business.

Running on milliwatts: power efficiency

The NDP uses a proprietary at-memory design that minimizes data movement. In the comparison shown, it claims 7.4x the compute efficiency and 5.9x the energy efficiency for wake-word detection, and 11.8x the compute efficiency for noise suppression.

On this point alone, Syntiant sits at the opposite pole of AI semiconductors. If Cerebras, decoded earlier in this series, is the wafer-scale giant chip that handles training in the data center, Syntiant is its exact opposite — a milliwatt inference chip that runs on a battery.