Science & Technology

Flex Logix Raises $ 55 Million to Design AI Chips for Edge Enterprise Applications

Join Transform2021 for the most important themes of enterprise AI and data. learn more..

FlexLogix, a startup designing reconfigurable AI accelerator chips, announced today that it has completed a $ 55 million round of funding led by Mithril Capital Management. CEO Geoff Tate said the funding will enable the company to build software, engineering, and customer support teams to accelerate hardware and software availability for edge enterprise applications.

AI accelerators are a type of specialized hardware designed to accelerate AI applications, especially neural networks, deep learning, and various forms of machine learning. They focus on low-precision arithmetic or in-memory computing, which can improve the performance of large-scale AI algorithms and deliver cutting-edge results in natural language processing, computer vision, and other domains. I can do it. Perhaps that’s why they’re expanding their share of edge computing power and are projected to account for 70% of that by 2025. according to For a recent survey by Statista.

Founded in 2014 and based in Mountain View, California, FlexLogix claims that the AI ​​inference chip Infer XX1 is the fastest and most efficient. According to Flex Logix, the Infer X1 outperforms Nvidia’s Xavier NX in the popular computer vision benchmark YOLO v3 and the “real customer model”, giving the company a 10 to 100 times better price-to-performance ratio than existing edge inference solutions. He states that he is aiming.

“FlexLogix is ​​for FPGAs and Arm is for processors,” Tate told VentureBeat in an email. “We believe that this original eFPGA business can grow as large as Arm over time, while the second business line drives state-of-the-art AI inference capabilities to large numbers of applications. We are growing the market to the billions of dollars that market forecasters predict. “

The InferX X1 also features nnMax, what Flex Logix calls a reconfigurable tensor processor, with 64 processors combined with SRAM that can be reprogrammed in 1/4 million seconds. In machine learning, a tensor is a generalization of vectors and matrices, a representation of data inputs, outputs, and transformations in neural networks. Flex Logix claims that nnMax is 3-18 times more efficient in terms of throughput per square than the average Nvidia GPU.

“”[The nnMax] Reconfigure 64 [processors] We repeat this layer layer by layer with RAM resources to efficiently implement a layer with a dedicated data path for full bandwidth, such as an ASIC, “FlexLogix explains on its website. “”[We use] Less than half the silicon area of ​​traditional mesh interconnects, fewer metal layers, higher utilization, higher performance, new breakthrough interconnect architectures … Easily scale up your architecture to compute of any size Can provide winging capabilities … Uses a patented tiling architecture with interconnects at the edges of tiles to automatically form larger arrays of any size. “

On the software side, the Flex Logix compiler takes models from machine learning frameworks such as Google’s TensorFlow and ONNX and optimizes them for the nnMax and InferX1 architectures. Performance modelers are currently available and used by “dozens” of customers, and Flex Logix will eventually make available software drivers for servers and operating systems commonly used in real-time scenarios. I plan to do it.

The Flex Logix product isn’t on the market yet, but the company says it will be available in PCIe cards and M.2 format for edge servers and gateways. PCIe boards, including the InferX1 and X1P1, will go into production in the second quarter of 2021, with prices between $ 399 and $ 499, depending on processor speed. A less powerful variant of the chip, the Infer X11KU, costs $ 99 to $ 199 and volume prices range from $ 34 to $ 69.

Flex Logix

Flex Logix is ​​competing in potential markets $ 91.18 billion By 2025. In March 2020, Hailo, a startup that develops hardware designed to accelerate AI inference at the edge, was acquired. $ 60 million At venture capital.Based in California mythology Raised $ 85.2 million to develop a custom in-memory computing architecture. GraphcoreA UK-based startup that creates chips and systems to accelerate AI workloads, but has a treasure trove of hundreds of millions of dollars in war.And Baidu’s growing AI chip unit Recently evaluated For $ 2 billion after funding.

However, FlexLogix investor Ajay Royan points out that Tate’s pedigree is one of the reasons for his continued confidence. Tate previously managed AMD’s microprocessors and logic groups. He raised his first startup, chip licensing company Rambus, from a four-person $ 2 million stake to a Nasdaq IPO and a multi-billion dollar market capitalization. According to Flex Logix, revenue in 2020 will be in the millions of “double digits” and is expected to grow by 50% to 100% this year.

“We were impressed with the architecture that Flex Logix developed on its own intellectual property … which gives us a sustainable competitive advantage in very high growth markets.” Royan said in a press release. “The benefits of this technology enable Flex Logix to support the rapid growth of edge enterprise inference in applications such as healthcare, retail, industry, and robotics. System-on-chip designers can reconfigure into communications and data centers. It’s even more impressive to do this with a small amount of capital and at the same time build a positive cash flow business, as we are considering incorporating. “

Lux Capital, Eclipse Ventures, and Tate Family Trust also participated in Flex Logix’s latest funding round, Series D. This brings the total amount of companies raised to date to $ 82 million.


VentureBeat’s mission is to become a digital town square for technical decision makers to acquire knowledge about innovative technology and trading. Our site provides important information about data technologies and strategies to guide you when you lead your organization. We encourage you to become a member of the community and access:

  • The latest information on the subject you are interested in
  • Newsletter
  • Gated sort reader content and discounted access to valuable events such as: 2021 transformation: learn more
  • Network function etc.

Become a member

Flex Logix Raises $ 55 Million to Design AI Chips for Edge Enterprise Applications Flex Logix Raises $ 55 Million to Design AI Chips for Edge Enterprise Applications

Back to top button