How will AI impact the semiconductor market through consumer applications
KEY FEATURES OF THE REPORT
- Artificial Intelligence (AI) technologies used in consumer applications
- Cloud computing and edge computing for AI
- System on chip including AI units and sound processor/vision processor for AI descriptions, ecosystems and trends
- Focus on smartphones, drones and smart home
- AI software design and companies’ strategies
OBJECTIVES OF THE REPORT
This report’s objectives are to:
1. To provide a scenario for AI within the dynamics of the consumer market and understand the impact of AI on the semiconductor industry
- Hardware for AI revenue forecast, volume shipments forecast
- Systems ASP forecast, revenue forecast, volume shipments forecast
- Focus on consumer applications embedding technologies: smartphone, drones, cameras, smart/robot home
2. To provide in-depth understanding of the ecosystem and players.
- Who are the players? What are the relationships inside this ecosystem? Who will win the battle for controlling the data?
- Who are the key suppliers to watch and which technologies do they provide?
3. To provide key technical insight & analysis about future technology trends and challenges.
- Key technology choices
- Technology dynamics
- Emerging technologies and roadmaps
Table of Contents
Report objectives and methodology 4
Executive summary 13
Market and technology trends 59
- Consumer drones
- Smart/robot home
Market forecasts 99
- Smartphone – 5 years forecast
- Smartphone SoC – 3 years forecast
- Smartphone SoC – Market share
- Smartphone – AI SoC – 5 years forecast
- Consumer drone – 5 years forecast
- Consumer drone – AI hardware – 5 years forecast
- Audio – Sound processor – 5 years forecast
- Audio – Embedded SPU – 5 years forecast
- Audio – AI market value by type – 5 years forecast
- Imaging – Vision processor – 5 years forecast
- Imaging – Embedded VPU – 5 years forecast
- Imaging – AI market value by type – 5 years forecast
- Embedded vs standalone hardware for AI revenue
- Hardware for AI – Introduction
- Hardware for AI – Smartphone
- Hardware for AI – Consumer drone
- Software for AI – Design
- Software for AI – Strategy
- Hardware – Smartphone SoC integration
- Hardware – Cloud computing integration
- Software – Biometrics authentication
- Software – Virtual personal assistant
- Software – Photography and augmented reality
- Software – Gesture recognition
Algorithm review 261
- AI algorithms for imaging
- AI algorithms for audio
FROM THE CLOUD TO THE EDGE; FROM MOORE’S LAW TO SPECIALIZATION
This report is dedicated to the impact of Artificial Intelligence (AI) on the semiconductor industry. This trend is particularly important in this period of slowdown in Moore’s law and the increasing regulation of data privacy. While the 1990s and early 2000s saw the emergence of decentralized computing with increasingly powerful central processing units (CPUs) and graphics processing units (GPUs) mounted on boards, the arrival of the mobile phone created the need to have calculation devices that are more efficient, smaller, and less greedy. The centralization of computing was then revealed as an adequate solution to meet its needs and the system-on-chip (SoC), the technological marker of this centralization, began to spread in these systems. However, AI algorithms, mainly convolutional neural networks (CNNs), involve many operations. While GPUs were adapted to the Cloud, they are less suitable for Edge computing on the device itself for reasons of excessive energy consumption.
This emerging desire to take the calculation to the device itself is rather recent and corresponds to a desire to free devices from the constraints of the Cloud in areas like data privacy, security, cost and latency. However, the constraints on the Edge are also important: devices must consume little power, always stay on and be fast and accurate.
With Moore’s law slowing down, it is necessary for companies to follow these trends today. To deliver all the necessary criteria, they must create hardware dedicated to the desired software task. For AI in imaging, this dedicated hardware corresponds to what Apple has named its ‘Neural Engine’ for example. It provides accelerators dedicated to extremely fast calculations of the weights associated with neurons of the inference network.
In this report, Yole Développement (Yole) look at these trends for two different types of architectures; those embedded in the SoC and dedicated chips. AI devices in imaging, such as Vision Processing Units or Vision Processors (VPU/VP), and in audio, such as Sound Processing Units or Sound Processors (SPU/SP), are completely different in terms of algorithms and resulting hardware. In all cases, in this race for specialization and optimization, software and hardware require common developments.
MAJOR REVENUES SET TO INCREASE
This report focus on three promising markets for this type of hardware: smartphones, drones and systems included in smart homes such as cameras and virtual personal assistants (VPAs), among others. The following figure shows the revenues for VPU/VP and SPU/SP architectures. Based on our assumptions and combining audio and imaging, the artificial intelligence hardware market could reach more than $10 billion for embedded hardware and more than $13 billion for stand-alone chips in 2023. For imaging, specialization has already begun in smartphones or drones and will enter smart homes by 2020. For audio, because hardware and software technology is still in the optimization phase, penetration in these markets will not be effective until the end of 2019 and more likely in the course of 2020.
The most promising market is smartphones. Yole hypothesize that AI for imaging will be embedded in the smartphone Application Processor (AP).
For audio, according to the choices of different companies and their respective technological status, we could see AI embedded in the AP or computed by a specialized standalone chip.
In the same way, the market of drones capable of integrating AI has been studied. This includes midprice drones with a minimum average selling price (ASP) of $500 and high-end drones with an average ASP of $1200. AI technology would be based mainly on imaging around the recognition of gestures for control, obstacles and the environment using massive VPs.
For the smart home market, all architectures could potentially be used. That makes this market disparate but extremely rich and promising. Because data is a profitable resource, this market has a large ecosystem. Promising startups mix with Google, Apple, Facebook, Amazon, Microsoft (GAFAM), Baidu, Alibaba, Tencent and Xiaomi (BATX) in a battle to control this increasingly regulated resource.
DATA IS MONEY
Controlling data provides phenomenal firepower in the field of AI. Data is needed to create more precise, more efficient, more customizable, better algorithms. Getting the best algorithms leads to greater penetration and acceptance by the public and, obviously, higher revenues.
Who are the players in this game? What are their strategies, visions and cultures? This report try to answer these questions and understand future trends in software and hardware. The battle is raging between decentralization companies, like ams, Sony or Knowles and AP manufacturers, like Apple, HiSilicon, Qualcomm or NVidia. GAFAM and BATX will play a central role, as some of them now design their own hardware and are acquiring or investing massively in AI software technologies.
This battle has just begun and Yole are trying to decipher the forces involved, the trends. It’s an exciting exercise – this world is constantly, very quickly moving.
Alibaba, Alphabet, Amazon, AMD, Another Brain, Apple, ARM, Asus, ATI, Baidu, CEVA, Cray, Deephi Tech, DeepMind, Facebook, Google, Graphcore, Hisilicon, Hover Camera, Huawei, IBM, Imagination,Infineon, Instagram, Intel, Kalray, Knowles, Lattice, LightOn, Mediatek, Microsoft, Motorola, Nokia,Nuance, Nvidia, Oppo, Parrot, Qualcomm, Samsung, Skydio, Sensetime, Socionext, Sony, ST Microelectronics, Synopsis, Tencent, Texas Instrument, Videantis, Xiaomi, Xilinx, and many more…
Related Reports & Monitors
Qualcomm 3D Sonic Sensor Fingerprint
Reverse Costing - Structural, Process & Cost Report
Melexis Far Infrared Thermal Sensor MLX90640
Reverse Costing - Structural, Process & Cost Report
Need to discuss?
We are open for discussionContact us