The silicon that runs AI on the device

Every AI model needs a chip to run on, and the chips that train giant models in a data center are a poor fit for a battery-powered gadget. Edge AI chips are the processors built to run inference on the device itself, small, low-power, and fast enough to answer without a cloud round trip. They are what makes a phone understand speech, a camera recognize a face, and a drone navigate on its own.

George Gilder’s latest pitch from the George Gilder Report is a bet on exactly this category. The promo, headlined “This Company Is 40,000 Times Smaller Than NVIDIA… But It’s Set To Power All Of ‘Ambient AI,’” points to one small semiconductor company as the way to own the trend, with the familiar hook of an “early stake in a $6 stock” ahead of a “$1 trillion wealth explosion.”

Three ways to build an edge AI chip

There is more than one way to put intelligence on a device, and the differences matter for investors. A dedicated application-specific chip, or ASIC, is custom-built for one job and is the most efficient but the least flexible. A general-purpose processor with AI accelerators bolted on, the approach Qualcomm and Apple use in phones, balances flexibility with power. And a field-programmable gate array, or FPGA, is a chip whose logic can be reprogrammed in software after it ships, useful when a device may need to run updated models or adapt to new workloads.

QuickLogic (QUIK), the company the Ambient AI promo resolves to, sits in the third camp with a twist. Rather than sell a standalone FPGA, it licenses “eFPGA” intellectual property, the “e” for embedded, so a customer can bake a small programmable fabric directly into their own system-on-chip. The result uses less power and less board space than a standalone chip, which is exactly what battery-powered and defense-grade devices want.

The company behind the tease

QuickLogic is a fabless semiconductor firm in San Jose with about 51 employees and a market value near $192 million. The chipset image in the ad is a QuickLogic ArcticPro eFPGA, which is how the “secret” was identified. The company has been public since 1999 and a Gilder recommendation since December 2019, and its customers span aerospace and defense, industrial infrastructure, and edge computing, consistent with the promo’s claim of a U.S. military deal for Ambient AI chips in next-generation weapons.

We walk through the company’s business in more detail in our QuickLogic Corporation explainer, and we put the whole on-device shift in context in our AI at the edge explainer.

The fine print on the pitch

The headline numbers need a second look. “40,000 times smaller than NVIDIA” is a size comparison, not a valuation argument: at a $192 million market cap, that multiple would put NVIDIA near $7.7 trillion. The “$1 trillion” figure describes the entire edge-AI market, not QuickLogic’s slice of it. And the “$6 stock” language is stale, since the shares were near $8 to $9 when the ad was re-teased in March 2026 and closed at $10.61 on September 2, 2026.

The competitive picture is the part the promo omits. AMD, which now owns Xilinx, and Lattice Semiconductor both sell low-power FPGAs and run their own eFPGA and edge-AI programs, and the biggest edge-silicon buyers favor suppliers with long qualification cycles and deep support staff. A 51-person company is real, but it is competing against far larger engineering teams for the same design wins.

The power budget is the real constraint

Everything in edge silicon bends around one number: the power budget. A device running on a battery has a strict envelope, and a chip that draws too much power shortens battery life or generates heat the device cannot shed. That is why efficiency, not raw performance, is the design goal for edge AI chips, and why programmable logic that sips power can win sockets a data-center GPU never could. It is also why the marketing around these chips leans on watts and milliwatts rather than teraflops.

The honest read

Edge AI chips are a genuine and growing category, and QuickLogic is a legitimate, if tiny, participant through its eFPGA licensing. The technology is real and the theme is sound, but the specific stock carries the execution risk of a microcap in a crowded field. Investors should evaluate QuickLogic on its fundamentals, its revenue growth, and its position against AMD and Lattice Semiconductor, not on the strength of a “$1 trillion” headline.

NewsletterVetter is an independent publication. We receive compensation from some of the services we review through affiliate links. Nothing on this site is investment advice. Always do your own research.