Analog Inference: Lower Power, Higher Performance

Analog Inference is enabling powerful edge AI using analog computing

Written by

AV Communications Team

Published on

Read

2 min

As machine learning and artificial intelligence continue disrupting many industries, companies are increasingly turning to compute-intensive applications. However, achieving and maintaining the extremely large amount of processing these workloads require is extremely difficult due to the inadequacy of existing hardware that can’t manage low power needs — especially in edge markets such as surveillance, smart retail, smart city, mobile phones, and other edge devices.

Alumni Ventures portfolio company Analog Inference is helping deliver the high-performance and low-power edge AI needs with next-generation technology that is radically superior to current digital computing solutions. The company’s AI chips have far better performance per watt than traditional computing options, which is critical for enabling proper AI computation at the edge of networks and devices.

A Momentous Market and Superior Product

Analog Inference is designing AI coprocessor chips that boast exceptional performance while costing less and requiring less power than current solutions. These chips have distinct advantages over incumbents in the marketplace, such as:

  • Up to 100X better AI performance per watt than traditional solutions
  • Mature process node enables GPU performance at Edge cost
  • Large capacity enables data-center grade neural nets at the edge

One of Analog’s initial target markets is video, specifically working with surveillance players and other companies needing to run video analytics at the edge. However, the company’s technology has the potential to integrate with other fast-growing markets, including autonomous vehicles, edge-cloud inference, and mobile phones. The company already has engagements with prominent companies across these industries. 

Knowledgeable Lead Investor and Experienced Team

Analog is backed by an established investor syndicate, including Khosla Ventures. Khosla led the company’s previous seed round and holds two board seats. Vinod Khosla is one of the most respected Silicon Valley entrepreneurs and venture capitalists. He has founded multiple companies (Daisy Systems, Sun Microsystems, Khosla Ventures), incubated others (Juniper Networks, Nexgen), and is now acting as an advisor to Analog Inference.

Analog’s team has deep technical expertise in semiconductors, AI, and analog computing.

  • CEO Carey Kloss has been involved with the AI accelerator sector since 2014 and has deep expertise in AI and neural computing. After being employee #1 at the leading AI company Nervana Systems, Carey led full-stack engineering for Intel’s data-center NNP AI products, joining Analog Inference in 2020.
  • Founder, President, & COO Vishal Sarin has led the creation of groundbreaking analog and AI technology at Micron, Fast-Chip, Information Storage Devices, SST/Microchip, and National Semiconductor. He holds over 100 patents.

How We Are Involved

Alumni Ventures’ was introduced to the deal through a connection with Khosla. Blue Ivy Ventures (for the Yale community) sponsored Alumni Ventures’ participation in Analog’s Series A1 round. Sibling funds Green D Ventures (for the Dartmouth community), Strawberry Creek Ventures (for the UC Berkeley community), and Triphammer Ventures (for the Cornell community) also participated in the round, along with AV’s Deep Tech Fund and Total Access Fund