Movidius

Movidius was a semiconductor company established in 2005 by Sean Mitchell, David Moloney, and Val Muresan. Although its origins were in Dublin, Ireland, the company was later based in San Mateo, California. It developed specialized vision processing units (VPUs) designed to perform computer-vision and machine-learning workloads on devices with limited power, thermal capacity, or connectivity. Its technology addressed a growing need to process visual information locally rather than sending camera data to a remote cloud service. The company's central product family was Myriad, a manycore processor architecture optimized for vision workloads. Myriad chips were intended for embedded products such as drones, robots, smart cameras, virtual-reality and augmented-reality systems, industrial machine-vision equipment, security products, and other Internet-of-Things devices. The architecture emphasized parallel processing and energy efficiency, enabling real-time analysis of images and video in compact hardware. This made Movidius relevant to applications including object recognition, person detection, tracking, gesture interpretation, spatial understanding, and autonomous navigation. Movidius raised nearly $90 million in capital between 2006 and 2016. In May 2013, it appointed Remi El-Ouazzane as chief executive officer. A significant development followed in January 2016, when the company announced a partnership with Google. Movidius technology was associated with Google's Project Tango, a computer-vision and spatial-computing initiative, helping demonstrate the role of dedicated low-power vision processors in mobile and mixed-reality systems. Its Myriad 2 VPU became the basis for several products and integrations. The Fathom USB accelerator packaged a Myriad 2 processor so that developers could add local vision and neural-network inference capabilities to systems using ARM processors and other host platforms. Movidius also supplied or enabled technology used in products such as the Google Clips camera, Intel RealSense Tracking Camera T265, consumer drones, programmable drones, and industrial inference hardware. The Intel Movidius Neural Compute Stick extended the platform into an accessible development device for prototyping edge artificial-intelligence applications with frameworks including TensorFlow and Caffe. The later Myriad X represented the third generation of Movidius VPU technology. It added a dedicated Neural Compute Engine intended to accelerate deep-learning inference while retaining the low-power design goals of the earlier architecture. Intel introduced the Neural Compute Stick 2 in November 2018. Based on Myriad X and designed for use with the Intel Distribution of OpenVINO toolkit, it provided developers with a compact USB-connected platform for optimizing and deploying neural-network models without requiring cloud connectivity. Intel announced its planned acquisition of Movidius in September 2016, and the transaction was completed that month. Intel continued selling products using the Movidius name and incorporated the technology into its broader edge-computing, computer-vision, RealSense, and artificial-intelligence initiatives. Movidius therefore became both an acquired technology business and a continuing product brand rather than an independent publicly traded company. The brand's enduring positioning is built around efficient on-device vision processing: bringing specialized artificial-intelligence inference closer to cameras, robots, drones, and other edge devices.

Movidius was founded in 2005 by Sean Mitchell, David Moloney, and Val Muresan. The company originated in Dublin, Ireland, before becoming based in San Mateo, California. Its business was built around specialized processors for computer vision, a field that requires large amounts of parallel computation but is often deployed in products with strict limits on battery life, size, heat generation, and network bandwidth. Rather than positioning its processors as general-purpose central processing units, Movidius developed the Myriad family as vision processing units. These chips were designed to execute image and video workloads efficiently across many processing elements. The approach was useful for embedded systems that needed to interpret camera feeds in real time, including drones, robots, smart cameras, surveillance equipment, augmented-reality products, and industrial machines. Between 2006 and 2016, Movidius raised nearly $90 million in capital. In May 2013, the company appointed Remi El-Ouazzane as chief executive officer. During this period, the company developed the Myriad 2 platform and expanded its presence among device makers and technology partners. A major partnership with Google was announced in January 2016. Movidius technology was active in Google's Project Tango, which explored mobile spatial perception and three-dimensional environmental understanding. The Myriad 2 architecture also supported the Fathom USB accelerator. Fathom packaged a Myriad 2 processor in a peripheral form factor, allowing developers and manufacturers to add computer-vision and neural-network acceleration to host systems based on ARM processors and other platforms. Its intended uses included drones, robots, personal computers, Internet-of-Things equipment, and video-surveillance systems. The processor was designed to deliver substantial vision-processing throughput while using approximately one watt of power, depending on workload and operating conditions. Movidius technology appeared in a range of products and demonstrations. The Google Clips camera used Myriad 2, as did Intel's RealSense Tracking Camera T265. The Myriad 2 was also incorporated into consumer and programmable drone products, including products associated with Mavic and the Ryze Tello platform. Other documented applications included industrial inference hardware and augmented-reality glasses, illustrating the breadth of the company's intended market. In September 2016, Intel announced its plan to acquire Movidius, and the acquisition was completed during the same month. After the transaction, Movidius ceased to operate as an independent company but remained visible as an Intel product and technology line. Intel continued marketing products based on the architecture and connected the technology with its edge-computing and OpenVINO initiatives. The next major product generation was Myriad X, described as Movidius's third-generation VPU. It introduced a Neural Compute Engine, a dedicated accelerator for deep-learning inference. Intel subsequently marketed the Intel Movidius Neural Compute Stick, a USB development device based on Myriad 2, and the Neural Compute Stick 2, introduced on November 14, 2018, at AI DevCon in Beijing. The second-generation stick used Myriad X and was designed to work with the Intel Distribution of OpenVINO toolkit. These products made Movidius technology available not only inside finished devices but also to software developers prototyping real-time, low-power artificial-intelligence applications. Movidius's historical significance lies in its focus on dedicated edge vision processing before on-device AI became a mainstream product category. Its technology helped establish a model in which cameras and embedded devices could perform visual inference locally, reducing dependence on cloud processing and supporting applications where latency, privacy, or unreliable connectivity mattered. Under Intel ownership, the Movidius name continued to identify this specialized processor heritage.

• 2018

  Neural Compute Stick 2 introduced

  [1]Movidiusen.wikipedia.org

  Intel introduced the second-generation Neural Compute Stick, based on Myriad X, at AI DevCon in Beijing.

• 2016

  Google partnership announced

  [1]Movidiusen.wikipedia.org

  Movidius announced a partnership with Google and became active in work associated with Project Tango.

• 2016

  Acquired by Intel

  [1]Movidiusen.wikipedia.org

  Intel announced and completed its acquisition of Movidius in September, after which the technology continued under Intel.

• 2013

  Remi El-Ouazzane appointed CEO

  [1]Movidiusen.wikipedia.org

  The company appointed Remi El-Ouazzane as chief executive officer in May.

• 2006

  Capital-raising period begins

  [1]Movidiusen.wikipedia.org

  Movidius began a period of financing that ultimately brought in nearly $90 million by 2016.

• 2005

  Movidius founded

  [1]Movidiusen.wikipedia.org

  Sean Mitchell, David Moloney, and Val Muresan co-founded the company in Dublin, Ireland.

Low-power edge-AI and computer-vision processing for embedded devices, emphasizing real-time inference, parallelism, and operation under tight energy and connectivity constraints.

• Vision processing unit

  Myriad 2

  Myriad 2 is a manycore vision processing unit designed for power-constrained embedded devices. It supports parallel computer-vision workloads and can be used for tasks such as identifying people or objects. The processor was designed to deliver approximately 80 to 150 GFLOPS while operating at around one watt, depending on implementation and workload. It appeared in cameras, drones, robots, tracking systems, and other smart devices.

• USB vision accelerator

  Fathom

  Fathom packaged a Myriad 2 processor in a USB stick, allowing a host device to add local computer-vision and neural-network acceleration. The product was aimed at developers and manufacturers working with ARM-based systems, personal computers, drones, robots, Internet-of-Things devices, and video-surveillance equipment. Its peripheral design was intended to simplify experimentation with low-power inference.

• Vision processing unit

  Myriad X

  Myriad X is the third-generation Movidius vision processing unit. Its defining addition is a Neural Compute Engine, a dedicated hardware accelerator for deep-learning inference. The architecture retained the brand's emphasis on efficient processing for edge devices while targeting more demanding neural-network workloads in cameras, robotics, augmented reality, industrial systems, and other embedded applications.

• Edge-AI development accelerator

  Intel Movidius Neural Compute Stick

  The original Neural Compute Stick is a compact USB-connected development device using the Myriad 2 VPU. It was designed to help developers prototype deep-learning applications at the network edge, including real-time vision workloads that could run without cloud connectivity. The platform supported development workflows involving frameworks such as TensorFlow and Caffe.

• Edge-AI development accelerator · 2018

  Intel Neural Compute Stick 2

  The Neural Compute Stick 2 is the second generation of Intel's Movidius-based USB inference accelerator. Introduced in 2018, it uses the Myriad X VPU and was designed to provide substantially greater neural-network inference capability than the first-generation stick. It works with the Intel Distribution of OpenVINO toolkit, allowing developers to optimize and deploy models on local edge hardware.

• 2018

  Launch of Neural Compute Stick 2Generation change

  Background

  Developers needed a compact way to prototype and deploy neural-network inference on edge devices without relying on cloud infrastructure.

  Action taken

  Intel introduced the Neural Compute Stick 2 at AI DevCon in Beijing. The product moved from Myriad 2 to the Myriad X VPU and integrated with the Intel Distribution of OpenVINO toolkit.

  Aftermath

  The second-generation development stick offered a newer hardware platform for local AI inference and extended the Movidius architecture's role in Intel's edge-AI developer ecosystem.

  Source

• 2016

  Intel acquisition of MovidiusM&A

  Background

  Movidius had developed low-power processors for computer vision and had expanded into deep-learning inference and embedded edge devices. Intel announced its intention to acquire the company as demand for computer vision and edge computing increased.

  Action taken

  Intel announced the planned acquisition in September 2016 and completed the transaction during that month. Movidius's technology and products continued under Intel ownership.

  Aftermath

  The acquisition ended Movidius's independence while preserving the Movidius product identity within Intel's processor and edge-AI portfolio.

  Source

• 2018

  Intel Neural Compute Stick 2 introduced[1]Movidiusen.wikipedia.org

  Product launch · Product generation

  Intel introduced the second-generation Neural Compute Stick at AI DevCon in Beijing. The device used the Myriad X VPU and was intended to improve edge-AI development and inference performance over the original model.

• 2016

  Movidius announces partnership with Google[1]Movidiusen.wikipedia.org

  Other

  Movidius announced a partnership with Google and participated in work connected with Project Tango, Google's computer-vision and spatial-computing initiative.

• 2016

  Intel announces planned acquisition of Movidius[1]Movidiusen.wikipedia.org

  M&A

  Intel announced that it planned to acquire Movidius, bringing the low-power vision-processing company and its Myriad technology into Intel's portfolio.

1. 1Movidiusen.wikipedia.org