Highlights:

  • The primary offering from SiMA.ai is a system-on-a-chip (MLSoC) that may be used to add artificial intelligence capabilities to edge computing devices.
  • With the capital raised from its recently disclosed investment round, SiMA.ai plans to release an updated version of MLSoC.

SiMA Technologies Inc. secured USD 70 million in fresh funding. The startup company develops processors for industrial robots and other edge computing devices.

The chipmaker said that Maverick Capital was the leading investor in the round when it made the announcement recently. Other institutional investors, including Dell Technologies Capital and well-known chip industry leader Lip-Bu Tan, participated in the round. Up to now, SiMA.ai has raised USD 270 million.

Maverick Capital Managing Director Andrew Homan said, “The computational intensity of generative AI has precipitated a paradigm shift in data center architecture. The next phase in this evolution will be widespread adoption of AI at the edge.”

The primary offering from SiMA.ai is a system-on-a-chip (MLSoC) that may be used to add artificial intelligence capabilities to edge computing devices. The processor is a single unit that integrates approximately six distinct processing components. Additionally, customers can write applications for the chip more easily with the help of a suite of software tools called Palette.

A collection of AI-optimized circuits, known as MLSoc circuits, can compute eight-bit integers—a type of data that AI models frequently use—with 50 TOPS of performance. A TOP is equivalent to one trillion calculations per second. The circuits may store the data they process internally instead of in slower external RAM since they are integrated with a 25-megabyte memory pool.

Four vector processing units designed to run computer vision algorithms support AI-optimized circuits. Modules for video encoding, or the act of converting the file format of media data processed by a computer vision algorithm, are also included in the MLSoC. According to SiMA, a device running on its processor can handle up to eight video streams concurrently due to the modules.

The MLSoC can handle general-purpose computing workloads with a quad-core central processing unit. The CPU runs Yocto Linux, an operating system version intended to run on linked devices, and is based on the Cortex A65 core design from Arm Holdings plc. To lessen the chances of hacking, a cybersecurity subsystem examines the MLSoC’s onboard software for security flaws each time it boots.

The chip comes with a set of development tools called Pallet from the manufacturer. The compiler that transforms AI models into a format that can operate effectively on the semiconductor is one of its many components. Additionally, several tools are available to developers to lessen the work required to create machine learning applications.

With the capital raised from its recently disclosed investment round, SiMA.ai plans to release an updated version of MLSoC. The next chip will be produced utilizing the six-nanometer node of Taiwan Semiconductor Manufacturing Co. Ltd. and is scheduled to ship in the first quarter of 2025. It will have processors for computer vision built using designs from Synopsys Inc., a significant supplier of semiconductor design software.

SiMA.ai Chief Business Officer Elizabeth Samara Rubio and Chief Executive Officer Krishna Rangasayee stated, “The second-generation SiMa.ai MLSoC will enable any framework, any network, any model or sensor, as well as any modality (as in audio, speech, text, image, and more) for edge AI applications.”