SAMPLE REPORT

Neuromorphic Chipsets
Applications, Competitive Intelligence, and PatentsAnalysis

The Next Revolution in AI Hardware

May 2019

Proprietary and Confidential, Copyright © 2019, Netscribes, Inc. All Rights Reserved
The content of this document is confidential and meant for the review of the recipient.
Disclaimer: The names or logos of other companies and products mentioned herein are the trademarks of their respective owners
 SAMPLE REPORT
Table of Content
Introduction R&D In the Neuromorphic Hardware
 Introduction to the Architecture and Properties Domain
 Implementation of Neural Algorithms on Hardware  Universities active in theDomain
 Spiking Neural Networks Driving Neuromorphic Computing
 Neuromorphic Chipsets Transforming the AILandscape Patent Trend
 Key Features Enabling Industry-wideAdoption Analysis
 Overview of the PatentingActivities
 Top Technology Domains
Industry Adoption of Neuromorphic  Focus Areas ofthe Technology Domains
Chipsets  Assignees across Geographies
 Automotive
 Overview ofAssignees
 IoT Ecosystem
 TopAssignees
 Financial Services
 Emerging Players and TopUniversities
 Cybersecurity
 Other ImportantActivities
 SpaceIndustry
 Medical Systems
Key
Competitive Takeaways
Intelligence
 Key Players Appendi
 Active Research Projects x
 Definitions
 Increased Interest in EmergingEntities  References
 Overview of the StartupEcosystem  About ROGM
 SAMPLE REPORT
Executive Summary
Click to edit Master
(1/3) titlestyle
Technology Industry Adoption

• Neuromorphic chipsets have integrated • Neuromorphic chipsets are still available for
processing and storage blocks, which eliminates prototyping and experimenting, with the initial
the bottlenecks related to speed, power architectures gradually finding applications in
consumption and efficiency that are present in various industries
the traditional Von Neumannarchitecture
• ………….. IoT ….. are near term application of
• Neuromorphic chipsets have the potential of neuromorphic chipsets
transforming the AI chipset market with near
term and long term use cases across various • Massive …….. and inherent …… properties make
industries the neuromorphic architectures suitable for
stock predictions and self-drivingscenarios
• Neuromorphic chipsets are laying the foundation
to achieve artificial general intelligence (AGI) • Long-term applications in ….. and space
capabilities industries

NEUROMORPHIC CHIPSETS 3
 SAMPLE REPORT
Executive Summary
Click to edit Master
(2/3) titlestyle
Frontrunners Future Innovators

• ……….is a highly accessible chipset in the market • Several startups are focusing on ……………
for neuromorphic demonstrations and closest to neuromorphic hardware for edgeoperations
being fully operational
• …………… and …………. are working on always-on-
• …………. Neuromorphic Research Community is voice functionalities
attracting many academic, government and
corporate research groups to develop hands-on, • Prophesee has a future focus towards human
extended neuromorphic solutions augmentations

• ……………. is an emerging entity directly • …………. Lab, ………….., and ……. are focusing on
competing with Qualcomm, Intel, andIBM neuromorphic vision

• ……………. is exploring AGI models for robots

NEUROMORPHIC CHIPSETS 4
 SAMPLE REPORT
Executive Summary
Click to edit Master
(3/3) titlestyle
Patent Trends Research Universities

• Top patent assignees including IBM, Qualcomm, • Researchers from ………….. University have
…………., ………….., ……….. account for about half investigated architectures like ………… and
of the total patentcount Braindrop. These researchers are also working
on a stealth-mode startup,……………..
• Patent publications are related to memory,
memristors, stochastic operations, neurosynaptic • The University of ………….. is targeting real time
cores, plasticity, and many other technology processing of event-based sensorysignals
areas
• …………… is actively collaborating with ……………
• Another Brain, …………, …………., Kneron, and universities for developing 17 core neuromorphic
………… are emergingplayers technologies

NEUROMORPHIC CHIPSETS 5
 SAMPLE REPORT

Introduction to theArchitecture and

Properties
 SAMPLE REPORT
Neuromorphic Chipsets: An Overview of the
Technology
To keep pace with AI software development, considerable efforts are being made to build hardware that can process AI algorithms. It
seems that the future of AI hardware will be defined by biologically-inspired neuromorphic chipsets, which will provide a real time
boost to AI systems. Brain-like chips will help showcase natural intelligence in major AI applications in the long term. These chips
include desirable characteristics for intelligent sensors, and the ultimate aim is to develop process technologies, materials, memories,
and other building blocks for the integration of neuron chips into sensors.

The third generation of neural networks is closer to biological neurons and is being investigated to develop a true neuromorphic
architecture. The future implementations of neuromorphic designs will lead to autonomous self-learning systems and high
performance architectures that can process quintillion calculations per second.

The major features that are driving adoption of neuromorphic chipsets include low power consumption, stochastic operations, pattern
recognition, faster computation, fault tolerance, and scalability. The concept of emulating neurons on a chip could enhance operations
and make business decisions reliable andcost-effective.

Parallel connected neurons can boost AI verticals significantly compared with the conventional processing systems. Non-stop learning
and pattern recognition leveraging the human brain architecture design can help compute signals and data in the form of visual,
speech, and olfactory inputs to perform real time operations and predict outcomes based on detected patterns.

NEUROMORPHIC CHIPSETS 7
 SAMPLE REPORT
Neuromorphic Chipsets: Introduction to the Concept and
Architecture
Biological Neuromorphic
Brain Chips
Data from varioussensory Data requires transformation The neuromorphic chips exhibit
Inspired by
Biological

organs of a human body into spikes between human brain-like capabilities for
Brain

automatically generates synapses that can be used achieving high connectivity,

spikes between synapses to train neural networks parallelism, and real time
– that are sent to and inferenceengines, and operation by collocating memory
neurons, which in turn ultimately perform various
and processing
enables the brain tomake application-specific tasks
decisions

Synaps Connectio
es from
Synapses receive signals ns
Emulated synapses that
Neuromorphic chips are
other neurons as voltagespikes communicate by means ofspikes
obtained by emulating

Emulation of
Activities
the brain’s neural Biological Emulated Neurons

Brain
activities with the help Neurons
Function in asmooth Integrate incoming signals
of artificial synapses, analog pattern of voltage for parallelism
neurons, and axons –
Axons Wires
recreated on a chip
Transmit voltage spikesto Wires are emulated axonsthat
other neurons mimic an axon’soperations

NEUROMORPHIC CHIPSETS 8
 SAMPLE REPORT
How is Neuromorphic Architecture
Different?
Neuromorphic architectures address
challengeslike high power consumption,
low speed, and other efficiency-related
bottlenecks prevalent in the traditional von
Neumann architecture

Von Neumann Neuromorphic

Architecture Architecture
CPU Neuromorphic architectures integrate processing and
Architectu Bottleneck Versu storage, getting rid of the bus bottleneck connecting the
re Memory s CPU and memory

Unlike the von Neumann architecture with sudden highs and

Encoding Versu lows in the form of binary encoding, neuromorphic chips offer
Scheme s a continuous analog transition in the form of spiking signals
and Signals

Devices CPU, memory, Versu Artificialneurons Neuromorphic devices and components are more complex
and logic gates,etc. s and synapses than logic gates
Componen
ts

NEUROMORPHIC CHIPSETS 9
 SAMPLE REPORT
Neuromorphic Chipsets vs.
GPUs
Neuromorphic chipsets are at an early stage of development, and would take approximately 20 years to be at the same level as GPUs. The
asynchronous
operation of neuromorphic chips makes them more efficient than other processing units.
Parameters Neuromorphic Chips GPU Chips
Basic Operation Based on the emulation of the biological nature of neurons Useparallel processing to perform mathematical operations
onto a chip
Parallelism Inherent parallelism enabled by neurons andsynapses Require the development of architectures forparallel
processing to handle multiple taskssimultaneously
Data Processing High High
Power Low Power-intensive
Accuracy Low High
Industry Adoption Still in the experimentalstage More accessible
Software New tools and methodologies need to be developed for Easier to program than neuromorphicsilicons
programming neuromorphic hardware
Memory Integrated memory and neuralprocessing Use of an externalmemory
Limitations • Notsuitable for precise calculations and programming- • Thread limited
related challenges • Suboptimal for massively parallelstructures
• Creation of neuromorphic devices is difficult due to the
complexity of interconnections

NEUROMORPHIC CHIPSETS 10
 SAMPLE REPORT
Neuromorphic Chipsets: Where Does it Fall in the Timeline?
Technology nodes are progressively reducing in size, while boosting the overall performance of the systems. However, Moore’s Law is reaching its limit, and
once 1-nm transistors are successfully implemented, reducing the size of elements further without compromising on the performance side of the ratio
would be a major hurdle. Therefore, neuromorphic computing is likely to be the next level of evolution in processing
architectures.

Quantum
1-nm Transistor Computing
Neuromorphic
Computing
Overall System
Performance

Carbon Nanotube

eMRAM

FPGA 2.0

201 20X 20X

7 X X

NEUROMORPHIC CHIPSETS 11
 SAMPLE REPORT
Spiking Neural Networks Driving Neuromorphic
Computing
Artificial
Intelligence
Neuromorphic computing provides multiple approaches
for developing AI technologies of tomorrow that are
deriving intelligence from real-world scenarios with a Machine Learning

massive amount of data.

Neural Network

Neural Network
Advancements
1st 2nd 3rd Ge
Different generations of neural networks models have been Gen Gen n
devised to develop and understand neuromorphic capabilities Artificial Convolution Spiking
to achieve brain-like efficiency Neural al Neural Neural
Network Network Network
ANN and CNN architectures are Event-driven, functionally
characterized by a single,continuous- similar to biologicalneurons,
valued activation and biologically more
realistic thanANN

NEUROMORPHIC CHIPSETS 12
 SAMPLE REPORT
Neuromorphic Chipsets Transforming the AI
Landscape
Artificial General Artificial Super
Intelligence Intelligence
Self-learning Intelligent Exascale Surpassing Human
Systems Computing Capabilities

Neuromorphic designing and Neuromorphic is among one of the Neuromorphic AI is a radical path
fabrication targeted towards high-performance architectures towards ensuring the exponential
developing self-optimizing and self- critical for the enablement of exa- growth of machine intelligenceand
configuring AIsystems for different FLOPS (1018 FLOPS)computing human augmentation
industries

>202 204 >205

0 0 0

NEUROMORPHIC CHIPSETS 13
 SAMPLE REPORT
Neuromorphic Chipsets: Key Features Enabling Industry-wide
Adoption
Advantages Leading to Ground-Breaking Use Cases

Low Power Fault- Stochastic

Consumption Tolerant Operation
The human brain performs complex computations Neuromorphic chips continue to operate even CurrentAI chipsets are ordered, and operate in a
on a small power budget of about 20 W of power, after the failure of a few components of the chip. calculated manner. However, neuromorphic chips
compared with the supercomputers that require This reduces the cost of production of the chips are stochastic in nature, because of which they
kW or MW power for AI applications. due to lower fabricationtolerances. can be used for allapplications.

Pattern Faster Scalabili

Recognition Computation ty
The low power pattern recognition of The inherent massive parallelism and low latency Neuromorphic architectures can be employed in
neuromorphic hardware helps to classify objects, factor of neuromorphic chips make it possible to edge applications and can also be scaled up for
make predictions, or anticipate conditions. It is perform complex computationsfaster. server applications.
also possible to understand the context of such Neuromorphic chips use less training data
patterns by using neuromorphic chips. compared with other intelligentchipsets.

NEUROMORPHIC CHIPSETS 14
 SAMPLE REPORT

Industry Adoption ofNeuromorphic

Chipsets
 SAMPLE REPORT
Neuromorphic Chipsets: Potential to Disrupt Industries
High Medium Low

IoTEcosystem Automotive Financial Services and SpaceIndustry Medical Systems

Cybersecurity

Adoption Adoption Adoption Adoption Adoption

Impact Impact Impact Impact Impact

Use Case Use Case Use Case Use Case Use Case

Neuromorphic chipsets are Neuromorphic chipsets are Noteworthy features suchas Neuromorphic chipsets for Neuromorphic chipsets could
appearing first at the edge suitable for classification parallel operations, the space industryare provide high performance
segment of the IoT tasks in ………….and, in the autonomous learning, and ……….…………..andare medical systems owing to
ecosystem where there isa short term, can beintegrated prediction of patterns are expected to be their ability to make
need for in conventional hardware expected to drive the commercialized afternon- predictions related to critical
………………………………….for that deals with ………… adoption of neuromorphic space applications. medical analysis based on
various applications. navigations for traffic. chipsets in financial services pattern recognition.
and cybersecurity.

LongTerm
NEUROMORPHIC CHIPSETS 16
 SAMPLE REPORT
Neuromorphic Chipsets: Automotive
Industry
Current Scenario of AI Bridging the Gap with Neuromorphic Future AI Chip
Chips Chips Expectations
Current architectures are good at precise Key requirement is to ensure that neuromorphic
A viable solution for the future of level 4-5
calculations andassure rule-based driving computing with the brain-inspired chip is compatible
self-driving vehicles would be a
characteristics with establishedarchitectures
combination of neuromorphic chips with
Autonomous driving market requires constant Future chips for autonomous connectivity need to be
the Von Neumann architectures, such that
improvement in AI algorithms for high throughput cost-effective, as AI chips account for a major portion of
the neuromorphic chip can be used for
with low power requirement the cost of self-drivingvehicles
object detection and classification tasks,
To increase the performance of objectdetection, As the life cycle of a chip is shorter than the life cycle of
while the Von Neumann architecture is a vehicle, frequent software and hardware updates will
higher model complexity isfollowed
used for precise calculations to ensure be required during one automobile life cycle to ensure
correct rule-based driving behavior. high performance and safety in self-driving vehicles and
to position the IC manufacturers in the value chain

Advantages of Neuromorphic Chips in Automotive Industry

Pattern Stochastic Low Power Faster

Recognition Operation Consumption Computation
Neuromorphic chips are ideal Brain chips are efficient in a Neuromorphic chipshave Spiking neural networks hasthe
for classification tasks andcan noisy environment, e.g., self- higher power efficiency inherent advantage of faster
be used fordifferent scenarios driving vehicles comparedwith compared with existing recognition speed
in autonomous driving static deep learningsolutions solutions

NEUROMORPHIC CHIPSETS 17
 SAMPLE REPORT
Neuromorphic Chipsets: IoT
Ecosystem
Current Scenario of AI Bridging the Gap with Neuromorphic Future AI Chip
Chips Expectations
Chips
The IoT ecosystem is moving toward the edge. Optimized neuromorphic hardware can AI chips will support absolute edge computing that does
Currently, edge devices rely on the cloud for not require the internet to perform operations
largely …………………
computation
…………………… Chip architectures would be general purpose for
Custom, application-specific SoCs are available for meeting ultra-low-power requirements with enhanced
meeting the low-power requirements of IoT devices performance and speed
………………………architectures is building
Wireless interoperability is needed to reduce Future chips will have autonomous learning capabilities
up and neuromorphic chips perfectly fit
fragmentation in the IoTecosystem to cater to differentindustries
the bill.
Design and verification requirements are the
challenging parameters

Advantages of Neuromorphic Chips in IoT Ecosystem

Pattern Low Power Faster Scalabili

Recognition Consumption Computation ty
Neuromorphic chips can Innovation at the edgerequires Real time learning capabilities Neuromorphic chips are
efficiently process voice,image, low-power and energy- will be essential for various scalable to the server level that
and signal data involved in harvesting devices mobility applications in the IoT would benefit IoT scenariosthat
various IoT user interfaces and ecosystem require hybridarchitectures
sensors

NEUROMORPHIC CHIPSETS 18
 SAMPLE REPORT
Neuromorphic Chipsets: Financial
Services
Current Scenario of AI Bridging the Gap with Neuromorphic Future AI Chip
Chips Expectations
Chips
AI chips require huge training data for mathematical Neuromorphic chips will drive the AI solutions with real time processing will be required to
calculations, which delays the overalloperation avoid losses incurred due to delayed outputs
mission-critical situations in the stock
Cost and power consumption are the problems
market. …………. Optimized power consumption is expected without
when it comes to application scaling
………………….. much emphasis on reducing the form factor
Hyperparameter tuning is the current challenge for
, therefore eliminating delays. Focus on designing algorithms that can adaptively
machine-learning algorithms choose and optimize models in response to the
information observed

Advantages of Neuromorphic Chips in Financial Services

Pattern Low Power Faster

Recognition Consumption Computation
………. Neuron chips are suitable ……………
option for predicting
unconventional and high
frequency tradingpatterns

NEUROMORPHIC CHIPSETS 19
 SAMPLE REPORT
Neuromorphic Chipsets:
Cybersecurity
Current Scenario of AI Bridging the Gap with Neuromorphic Future AI Chip
Chips Expectations
Chips
Current solutions sequentially match small chunks of The inherent parallel processing model AI chips will identify patterns in encrypted packets that
data against a library of suspicious patterns
of …………………. could point to malicious or unusual payloads inside the
The nature of current cybersecurity protocols follow ………………………………. traffic
a proactive and counter-responseapproach The aim is to work toward predictive cybersecurity
and predictive alerts associated with
……………….. postures
Training and inference systems for the detection of a
wide range of random anomalous behaviors of
computers and networking systems isrequired
potential threats and attacks.

Advantages of Neuromorphic Chips in Cybersecurity

Pattern Low Power Faster

Recognition Consumption Computation
SNNs can learn on the fly, which Best suited to anomaly
is an advantage in detecting detection in datamining
new attack behaviors orvectors procedures and predict
potential threats

NEUROMORPHIC CHIPSETS 20
 SAMPLE REPORT

Competitive Intelligence
 SAMPLE REPORT
Competitive Intelligence: An Overview of the Industry Players
Several public and private entities are committed towards building brain-inspired hardware for the future of AI. The major players active in
the neuromorphic chipset domain include IBM, Qualcomm, Intel, Brainchip, Samsung, HP, HRLlaboratories, and GeneralVision.
Collaborative efforts are also being made to apply neuromorphic research to real-world applications. Samsung is working in partnership with
universities to develop core neuromorphic technologies. Research projects by DARPA and European Union started years ago, and are
continuously evolving to leverage neurosciences for high performance architecture levels.

Significant increase in the number of startups and investments in the domain showcase the level of interest in the neuromorphic market.
These startups are providing differentiated offerings and are setting new frontiers in the deep learning architectures. While most of these
entities are focusing on edge operations, always-on applications and neuromorphic vision systems, few of them are already pursuing long-
term goals.
 Vicarious is targeting combination of probabilistic models with cognitive science for AGI, with a future aim of incorporating common
sense in robots
 Prophesee plans to use neuromorphic vision system for building medical devices to restore vision to the blind

A step towards the development of neuromorphic hardware could be the availability of design kits for fabless groups, universities, and
industry players.

NEUROMORPHIC CHIPSETS 22
 SAMPLE REPORT
Neuromorphic Chipsets: Key
Players
Corporates such as IBM, Intel, and Qualcomm are actively working on projects on next-generation AI chipsets. The most fully developed neuromorphic
platform is ……..……..

Company Implementatio Market Application Plans

….. Name/Project n low-power CMOS
28-nm, Readiness
Accessible for prototyping Focus
Robotics, medical, Aims Ahead
to lower AI-training
……..: Neurosynaptic processor containing 5.4 billion process, reduced neuron and demonstrations automotive, gesture power and release APIs to
transistors switching by 99% on recognition,machine its ecosystem partners to
average of typicalnetwork learning, mobile,etc. hook up real time sensors
…….. CMOS process, based Prototype demonstration Robotics, image Collaborations for testing
……….: Neural processing unit that is reprogrammable and around NPU, AIAccelerator recognition, bigdata the technology
supports parallelism Chip And softwareAPI processing
Intel 14-nm FinFET process, Prototype AI edge, imagerecognition, Plans to include>100
Loihi: Asynchronous neuromorphic core with 130,000 functional over 0.5V- robotics, etc. billion synapses inLoihi
neurons and 130 million synapses 1.25V system by 2019 and solve
LASSO optimization
problems
………. CMOS process Expected to be in market in Edge applications, fintech, Collaborating withglobal
…………: System-on-chip with 1.2 million neurons and 10 2019 automotive, manufacturers forearly
billion synapses cybersecurity,surveillance adoption ofAkida
and machinevision
Other Other entities are also exploring neuromorphic chipsets including Neuromem (a General Vision Company), Numenta,
Entities HP andSamsung

NEUROMORPHIC CHIPSETS 23
 SAMPLE REPORT
Neuromorphic Chipsets: Increased Interest in
Emerging Entities

NEUROMORPHIC CHIPSETS 24
 SAMPLE REPORT
Neuromorphic Chipsets: Overview of the Startup Ecosystem - (1/5)

Companies Technology Overview Future Focus

…. develops ultra-low-power and fully The company is aiming to support AI use cases
programmable neuromorphic computing and functions particularly for ………….Some of
technology for sensor analytics and machine the applications include autonomous navigation USD 15Mn
learning applications. This brain-inspired silicon platforms for cars, robots, anddrones.
HQ: Paris, France intelligence is targeted towards smart homes, Additionally, GrAI plans to advancetowards
Founded: 2016 automotive, and healthcare. market deployment.

………. is a provider ofneuromorphic The company plans to use the funds received for
…………………………integratedcircuit technology. It building a team and delivering its first product.
has developed a proprietary analog, ultra-low- USD 3.5 Mn
Furthermore, it is ……………….
power, always-on sensing chip. The solutionswill Fund suggesting a potential way forwardto
HQ: West Virginia, US help to overcome the challenges associated with bring ubiquitous voice-first experiences inlow-
Founded: 2015 data and power. power and portabledevices.

……… is a provider of mixed-signal The company aims to build end to end dedicated
neuromorphic processors that are based on a neuromorphic IP blocks, ASICs, and SoCs as full- USD 2.7 Mn
………………………..These processors are based custom computing solutions that integrate
on reconfigurable real-time neural networks neuromorphic sensors and processors. It is also
that enable ultra-low-power and ultra-low- planning to push real time ECG recording to
HQ: Zurich, Switzerland latency AI edge computingapplications. smartwatch-makers.
Founded: 2017

NEUROMORPHIC CHIPSETS 25
 SAMPLE REPORT

R&D in the NeuromorphicHardware

Domain
 SAMPLE REPORT
Neuromorphic Chipsets: Universities Active in the
Domain - (1/2)

Developed Neurogrid and Developed a hierarchicaland mesh-based Jointly developing an automated Jointly developing reconfigurable spike-
Braindrop architectures based on routing methodology for minimizing both mapping and efficient debugging routing architectures for on-chip local
encode- transform-decode memory requirements and latency, while framework for implementing learning in neuromorphic systems based on
mechanism. These architectures are maximizing programming flexibility to DNN onto a neuromorphic chip multiple routing schemes. In addition,
designed using neural engineering support event-based neural network with crossbar array of synapses. pointer-based schemes are applied to
frameworks used to synthesize architectures. In the future, this method global routing for multicore
neuromorphicnetworks. can be used in real time processing neuromorphic clusters which may support
of event-based sensory signals. millions of neurons and synapses.

Developed the Darwinprocessing unit, Developed a full-scale neuromorphic chip Jointly developing a neuromorphic Developed a neuromorphic chip that
fabricated by CMOS technology, simulator that integrates both RRAM chipset to realizeAI functions such as supports online pattern
targetting resource-constrained, low model and NoC model. The deep learning and data processing recognition.It works on STDP
power small embeddeddevices. neurosynaptic core architecture contains capabilities, including image and mechanism and teacher signals
an RRAM crossbar array for storing sound. The project is aimed at supporting both supervised and
synaptic weights and performs weight developing 17 core unsupervisedlearning.
accumulation through a current sum in technologies in the neuromorphic
analog domain. market.

NEUROMORPHIC CHIPSETS 27
 SAMPLE REPORT

Patent Analysis of Neuromorphic

Chipsets
 SAMPLE REPORT

Patent Trend Analysis: An Overview of the Patenting Activities

Despite neuromorphic chipsets being at an early stage of development, the patent filing activity is gaining interest across key
semiconductor companies, R&D centers and universities. The development of autonomous systems is a prominent focus area in the
overall patent study. Other notable area of interest has been investigation of neuromorphic systems or electronic synapses for the
implementation of reinforcement learning. Several patent publications are focusing on factors such as interconnect, fabrication
techniques, material, memory integration andedge devices.

The neuromorphic hardware technology involves more than 190 applicants. IBM has the highest patent filings (162 unique filings),
followed by Qualcomm (85 unique patent filings). The other key IP holders are SK Hynix and Intel. With growing interest, new players
such as Brainchip are also emerging with low-power neuromorphic voice activation systems and autonomous feature extraction using
SNN.

Of the top 20 assignees, 10 assignees are based in US and six are from South Korea. Several key IP collaborations have been identified
between universities and key entities. Start-ups such as Another Brain, Nantero, Neuramatix, Kneron, and Knowm are also filing
patent publications addressing key requirements of neuromorphic chipsets.

The patent landscape documents how innovation in neuromorphic chipset is increasing globally and new insights related to
the filing and technologytrend.

NEUROMORPHIC CHIPSETS 29
 SAMPLE REPORT
Patent Trend Analysis
The patent study has been conducted to understand the evolution of patent publications and countries of patent filings. The key assignees and technology
domains shaping the neuromorphic IP landscape have also been covered to assess upcoming trends.

Analysis of patent data from 1999 to 2018 provides several

insights into the neuromorphichardware landscape:

 700 patent publications were identified specificto

neuromorphic hardware

 There are few patent filings in the first 10 years (1999-

No. of Publications

2008), as the concept of neuromorphic engineering was

only developed in the1980s

 Filing spike started in 2012, when more than 20 patents

were filed by IBM

 Trends in patent filings during 2013–2015have been

stagnant with a very slight increase in 2015

Priority Year  The patent filing trends show that maximum filing
activity was in 2016
*The timeline considered for the study is 1999-2018
NEUROMORPHIC CHIPSETS 30
 SAMPLE REPORT
Patent Trend Analysis: Top Technology
Domains
Neuron-powered chips will become the enabler of many future applications. Some of the major aspects identified from the patent study are
systems/chips or apparatus focusing on brain models, spike generation, neuro-synaptic cores, etc., to facilitate large-scale neural hardware
implementation.

Interconnects Circuits Neurons

(12) for interconnections
Schemes (136)
Low power neural models or (197)
Architectures that support
in reconfigurable/peripheral circuits with high integration parallel processing, stochastic
units in neuronhardware and small chiparea operations, and back
propagation implementation

Synapse Fabrication Edge Devices

(90)
Integration for analog (52) array with efficiency,
Synapse (50)
Autonomous systems,olfactory
computation, dynamic scalability, performance, and auditory pattern classification,
processing andreinforcement manufacturability and image recognition
learning

Memory Memristors Others

(48) unit for synapse or
Memory (36)
Memristive neuron circuitsfor (79)communication, back
Chip
stochastic environment tosuit signal processing, pattern propagation, feedback
neuromorphic applications recognition, and control mechanisms, multi-chipnetwork,
systems etc.

NEUROMORPHIC CHIPSETS 31
 SAMPLE REPORT
Patent Trend Analysis: Top Technology Domains – FocusAreas (1/2)
TechnologyAreas Sub-Technology Focus Areas Patent Assignees
Time synchronization, encoding, modulation,
Neurons replay components, plasticity, analog
computation, neuron processors, feedbackloops,
materials, processing engines
Parallel processing, analog chip,accelerators,
Circuit non-linear characteristics, reduced chip area,
multiplexing of neurosynaptic cores, spike-
timing-dependent plasticity
Dynamic processing, nano-sheets, field
Synapses programmable synapses array, optical synapse,
resistance adjustments, weights of synapses,
crossbar architecture
Crossbar arrays of resistive processing units, 3D
Fabrication structures, synapses, multi-level architectures,
reduced chip area, ambipolar devices,low-power
designs
Autonomous systems, robots, low-power
Edge Devices requirements, context-based mapping,
parallelism, olfactory systems, image recognition,
improved computation
Resistive control, stochastic operations,
Memory thermodynamic RAM, multi-bit characteristics,
reconfigurable mapping, accelerator
architectures, phase change

NEUROMORPHIC CHIPSETS 32
 SAMPLE REPORT
Patent Trend Analysis: Assignees across
Key
Geographies Assignees
• Honda Motors
• Qualcomm
• Intel
Top Key Research Institutes
• Erle Robotics
Assignees
• IBM
Assignees and Universities
• Samsung
• Qualcomm • Waseda University
• HRLLaboratories
• Intel
• SK Hynix
Research Institutes
and Universities
Other Key • National Center forScientific
Brain Corporation
•Assignees Research CNRS
• HRLLaboratories • Ecole Polytechnique
• Samsung • Technical UniversityDresden
• Denso Corporation Key
Assignee Top
Research Institutes • IBM Assignees
and Research Institutes • SK Hynix
The Regents of theUniversity
•Universities and Universities Other Key
of California • Tsinghua University
• Northeastern University Assignees
• Peking University • Samsung
• MIT
• Nanjing University Research Institutes and
• The Charles StarkDraper
• Zhejiang University Postech Academy-IndustryFoundation
•Universities
Laboratory • Beijing Institute ofTechnology
• Boise State University • Korea University Research and Business
• Hefei University ofTechnology Foundation
• Stanford University
• University of Dayton • Myongji University
• Incheon University
Emerging
Assignees
• Applied BrainResearch
• Brainchip

NEUROMORPHIC CHIPSETS 33
 SAMPLE REPORT
Patent Trend Analysis: Overview of Assignees

 Over 190 applicants have filed patents in the

neuromorphic hardware domain
 The top five assignees account for half of the total
IBM
patent count, signifying the dominance ofmajor

Qualcomm entities in the neuromorphic hardware domain

349 out of 700 of the total  As the domain is dominated by the big players,
SKHynix
patent filing activity of emerging entitiesis
patent publications are
considerably less
assigned to the top 5 assignees Intel
 The remaining assignees include a mix of
Brain Corporation semiconductor manufacturers and IP providers,
automotive manufacturers and automotive service
Remaining
parts suppliers, universities and researchinstitutes,
and specialist neuromorphic solutionsproviders

NEUROMORPHIC CHIPSETS 34
 SAMPLE REPORT
Patent Trend Analysis: Top Assignees
 IBM is leading the race with a comprehensive patent
portfolio covering all aspects of the technology
 Qualcomm holds the second place in terms of patent count.

Neurons Circuits Synapse However, there is considerable gap between the top two
patent holders
Fabrication Edge Devices M emory
 More than half of the patent publications assigned to
Memristors Interconnects Others
Qualcomm are focused on neurons and related
architectures and sub-technologies
 SK Hynix’s patent portfolio is primarily focused on synapse-
related features, including enhancing symmetry and
linearity of synapses, with improved data retention
capabilities and newmaterials
0 20 40 60 80 100 120 140 160 180
 Intel’s IP focuses on accelerator circuits andstochastic
Number of Patent Publications operation for reducing delays, and also on reconfigurable
and flexible interconnection schemes
 Brain Corp. has significant number of patent filings in edge
device implementations
NEUROMORPHIC CHIPSETS 35
 SAMPLE REPORT
Patent Trend Analysis: Emerging Players and Top Universities
E erging Players
m
BrainChip has transformed into a specialized neuromorphic chip solution provider that is now
Edge device Associative connections competing with the big players in the domain and has a considerably strong patent portfolio to
implementation with between neuronsin a neural support its position. Apart from BrainChip, other emerging players were also identified focusing on
focus on voice-based network for efficient various aspects of the neuromorphic computing domain, including memories, memristors, edge
recognition and activation, learning and devices, autonomous operations, and new materials for the architecture
low-power neuronaldesigns, autonomous operations
and autonomoussystems

Low-cost, neuromorphic Neural processing units for

Exploring carbon Neuromemristive AI
chipsets and integrated edge devices with
nanotubes for and thermodynamic RAM
memory modules focusedon reduction in the number of
implementing dendrite technology stack utilizing
local, real-time, energy- components used and
circuits that can emulate Anti-Hebbian andHebbian
efficient operations reduced hardwarecost
neuronal behavior for (AHaH) computing
neuromorphic computing
applications

Top Active
Universities

……… University, ……. Academy-IndustryFoundation, and

the Regents of the University of California are the top
universities in the domain with the maximum number of
patentpublications

NEUROMORPHIC CHIPSETS 36
SAMPLE REPORT

Key Takeaways
 SAMPLE REPORT
Key
Click to
Takeaways
Technolo
edit Master
Busine
titlestyle
Partnersh Applicati
gy ss ip on

Companies intending to adopt new Neuromorphic hardware is still at a Strategic partnershipswith Semiconductor companies lookingto
business models relatively nascent stage, companies ……………………….. enter the neuromorphic space need
……………………….. intending to ………………. will allow companies to start by building chipsets suitable
………………. opportunities in SNN- ……………………….. to bring together neuroscience for ………………………..
based research. ……………….ranging from experts and semiconductor ………………. and in the future
architecture, circuit, and system to designers for developing silicon that build capabilities for complex
materials and fabrication techniques canefficiently mimic brain functions computations required for
autonomous systems.
Companies considering entry intoneural Semiconductor players candiversify ……………………….. Neuromorphic chipsets play acritical
chipsets can consider revenue streams, move beyond ………………. focusing on role in ………………………..
……………………….. monetization of silicon, and offer neuromorphic chipsets can leadto ……………….
………………. NN chipsets and follow neuromorphic IP licensing development in neuromorphic computational
it up with SNN-based chips for securing research and future product infrastructure.
the future applications Companies providingneuromorphic development.
………………………..
………………. advantage compared
incumbent semiconductorcompanies

NEUROMORPHIC CHIPSETS 38
SAMPLE REPORT

Appendix
 SAMPLE REPORT
Definitio
ns
• Synapses: A structure for information relay between two neurons that • Exascale Computing: Computing systems capable of at least one exaFLOPS
is being investigated for building neuromorphic platforms. (i.e., a quintillion) calculations persecond.
• Neuromorphic Computing: Engineering approach based on activities • Quantum Computing: Application of quantum-mechanical phenomena
of the biological brain. It improves efficiency of complex computational suchas superposition and entanglement to perform computation.
task related to perception and decision making. • ASICs: A semiconductor design for a specific application that can be programmed
• Edge Computing: Technique that allows the processing of data close to develop custom siliconsolutions.
to the device where it is generated instead of contacting any centralized • System-on-chip (SoC): Anintegrated circuit including varied electronic
cloud. Edge application services lead to reduction in data volume and the components on a single integrated chip that is used in edge devices such as
related network traffic. wearables, smartphones, etc.
• Neural Network: A network of artificial neurons that uses • Carbon Nanotube: Allotropes of carbon with a cylindrical nanostructure that can
computational model for information processing to solve artificial be single-walled or multi-walled and finds application in nanotechnology, optics,
intelligence problems. It is used for predictive modeling, adaptive control and electronics.
and other data processing.
• Backpropagation: A learning algorithm in artificial neural network for the
• Artificial General Intelligence: Representation of generalized minimum value of the error function in weight space using gradient descent.
human cognitive abilities in software to perform any intellectual
task that a human canperform.
• Von Neumann Architecture: John von Neumann first published the
Von Neurmann Architecture in 1945, where an instruction fetch and
data operation occur simulatenously with a shared common bus.
• Artificial Super Intelligence: The concept beyond artificialgeneral
intelligence where a computer’s cognitive ability surpasses human ability.

NEUROMORPHIC CHIPSETS 40
 SAMPLE REPORT
Research
Methodology
Netscribes follows a structured project management approach to deliver deep-dive research analysis. A comprehensive approach is administered to
assess innovative technologies and develop insights related to technology and business.

Technology Patent Report

Research Analysis Format
The research is focused on key technical trends, Patent study is conducted to provide an overall Report is prepared at the end of the research and
solutions, and adoption of the subject technology landscape of the patenting activities in the is aligned as per expectations and quality check.
across sectors. Segregation of the technology focus subject technology. The key trends in the patent An in-house editorial team conducts the final
areas and identification of emerging technology filing are identified including assignees, level QC.
trends and opportunities gives a detailed overview of technology focus areas, geographicaldistribution,
the technology. and emergingentities.

The analysis also includes technical approach taken Critical insights related to the patent analysis are
by the companies to overcome the limitations and incorporated to highlight future product
assessment of the R&D strategy and future plans developmentpointers.

NEUROMORPHIC CHIPSETS 41
 SAMPLE REPORT
Appendi
x
About Netscribes
Netscribes is a global market intelligence and content services provider that helps corporations achieve strategic objectives through a wide range
of offerings. Our solutions rely on a unique combination of qualitative and quantitative primary research, secondary/desk research, social media
analytics, and IP research. For more than 15 years, we have helped our clients across a range of industries, including technology, financial
services, healthcare, retail, and CPG. Fortune 500 companies, as well as small- to mid-size firms, have benefited from our partnership with
relevant market and competitive insights to drive higher growth, faster customer acquisition, and a sustainable edge in their business.

NEUROMORPHIC CHIPSETS 42
 SAMPLE REPORT
Disclaim
er
This report is prepared by Netscribes (India) Private Limited (“Netscribes”), a market intelligence and content service provider

The content of this report is developed in accordance with Netscribes’s professional standards. Accordingly, the information provided herein has been obtained from sources
which are reasonably believed to be reliable. All information provided in this report is on an “as-is" and an "as-available” basis, and no representations are made about the
completeness, veracity, reliability, accuracy, or suitability of its content for any purpose whatsoever. All statements of opinion and all projections, forecasts, or statements relating
to expectations regarding future events represent Netscribes’s own assessment and interpretation of information available to it. All liabilities, however arising, in each of the
foregoing respects are expresslydisclaimed.

This report is intended for general information purposes only. This report does not constitute an offer to sell or issue securities, an invitation to purchase or subscribe for
securities, or a recommendation to purchase, hold, sell, or abstain from purchasing, any securities. This report is not intended to be used as a basis for making an investment in
securities. This report does not form a fiduciary relationship or constitute investment advice. Nothing in this report constitutes legal advice.

The information and opinions contained in this report are provided as of the date of the report and are subject to change. Reports may or may not be revised in the future. Any
liability to revise any out-of-date report, or to inform recipients about an updated version of such report, is expressly disclaimed.

A bonafide recipient is hereby granted a worldwide, royalty-free, enterprise-wide limited license to use the content of this report, subject to the condition that any citation from
this report is properly referenced and credited to Netscribes. Nothing herein conveys to the recipients, by implication or by way of estoppel, any intellectual property rights in the
report (other than the foregoing limited license) or impairs Netscribes’ intellectual property rights, including but not limited to any rights available to Netscribes under any law or
contract.

To the maximum extent permitted by law, all liability in respect of this report and any related material is expressly disclaimed. Netscribes does not assume any liability or duty of
care for any consequences of any person acting, or refraining to act, by placing reliance on the basis of information contained in this report.

All disputes and claims arising in relation to this report will be submitted to arbitration, which shall be held in Mumbai, India under the Indian Arbitration and Conciliation Act. The
exclusive jurisdiction of the courts in Mumbai, India applies to all disputes concerning this report and the interpretation of these terms, and the same shall be governed by and
construed in accordance with Indian law without reference to the principles of conflict of laws.

NEUROMORPHIC CHIPSETS 43
 SAMPLE REPORT
Office Locations & Geographical
Coverage

Get in touch with us:

USA
41 East, 11th Street,
New York
NY10003, USA
Phone: +1-917-885-5983
Mumbai
Office No. 504, 5th Floor, Lodha
Gurgaon Supremus, Senapati Bapat Marg, Railway
Colony, Lower Parel,
Kolkata Mumbai 400013,
New York Maharashtra, India
Phone: +91‐844‐844-
Mumbai
9475
Singapore Kolkata
3rd Floor, SaberwalHouse
55B Mirza Ghalib Street,
Kolkata - 700 016
West Bengal, India
Phone: +91‐844‐844-9475
Gurgaon
Sector 44, Plot No 130,
2nd Floor
Gurgaon - 122003
Haryana, India
Phone: +91‐844‐844-9475
Singapore
Netscribes Global PTE.Ltd.
10 Dover Rise, #20-11,Heritage
Office locations Geographical Coverage
View, Singapore (138680)

NEUROMORPHIC CHIPSETS 44
For further assistance or to request customizations
please contact:

US :1 888 448 4309/ 1 877 777 6569

India: +91 22 4098 7600

[email protected]

Proprietary and Confidential, Copyright © 2019, Netscribes, Inc. All Rights Reserved
The content of this document is confidential and meant for the review of the recipient.
Disclaimer: The names or logos of other companies and products mentioned herein are the trademarks of their respective owners

www.netscribes.com

45
NEUROMORPHIC CHIPSETS