Pioneering artificial intelligence chip startup Mythic said today it has raised $70 million in a new round of funding that follows the launch of its first products last year.

The Series C round was led by BlackRock and Hewlett Packard Pathfinder and brings Mythic’s total amount raised to $165.2 million since it was founded back in 2012.

Mythic is an interesting company that has built its AI chips on older analog computing technology in an effort to make AI more power-efficient and therefore more accessible for devices at the network edge. It claims that modern digital graphic processing units sold by the likes of Nvidia Corp. use far too much energy when performing calculations. That means they’re not only expensive but also difficult to use with devices such as drones that typically have a very short battery life.

To remedy that, Mythic’s chips rely on analog currents and built-in flash memory that enables them to perform inference, or the process of running a trained AI model, more efficiently. The company reckons its analog AI chips can do this using 100 times less power than digital GPUs

The chips are small too, which makes them ideal for drones, smartphones, robots, virtual reality headsets and even wearable devices. It means that AI workloads can be performed on-device, at the edge of the network, rather than sending data back to a cloud data center and performing the computations there.

“We’re unlocking the limitations of AI so companies can cost-effectively deploy AI at an unprecedented scale,” said Mythic co-founder and Chief Executive Mike Henry.

Tim Vehling, Mythic’s senior vice president of product and business development, told SiliconANGLE that the company’s first-generation analog AI chips are somewhat comparable with Nvidia’s miniaturized Xavier AGX GPUs, which are low-powered processors also designed for edge devices that launched in 2018. But he said Mythic’s chips are more powerful and less expensive.

“The M1108 can achieve a peak performance of 35 trillion operations per second with typical deep neural network workloads, consuming only about four watts,” Vehling said. “It’s a 40-nanometer chipmaking process technology that requires no external memory.”

Mythic has taken a long time to bring its chips to market, but it finally sent out its first evaluation samples and kits to prospective customers last year. The M1108 Mythic Analog Matrix Processor will be available in a choice of single-chip or various PCIe card form factors, such as M.2 M Key and M.2 A+E Key configurations, to enable more design flexibility.

Vehling told SiliconANGLE that the company expects to start its first production shipments by the end of the year.

“We have had strong interest in high-performance video analytics applications such as video surveillance cameras for object detection and classification,” he said. “Other applications include industrial machine vision for high-resolution inspection, defect detection, augmented reality applications for low-latency pose and object detection, and drone applications like object detection, depth estimation and classification.”

Mythic said the Series C funding will be used primarily to accelerate mass production of its chips and better support its “growing customer base” that spans Asia-Pacific, Europe and the U.S. The company will also use the money to develop the next generation of its chips and build out its software portfolio.

Constellation Research Inc. analyst Holger Mueller told SiliconANGLE that it’s well-known that analog chip designs have a big advantage in terms of energy consumption, heating and the physical space they take up. “It’s not surprising that Mythic has scored this round of funding, as there is ample room for these kinds of AI inference chips in ‘internet of things’ and edge devices,” he said.

Image: Mythic