BRN Discussion Ongoing
- Thread starter TechGirl
- Start date
Jacob Kiraz
Regular
News and revenue will only sort this problemOver 25 million shares traded on ASX just to end up with a rise of 0,005 dollar at 19 cents
At least not 18.5
Good afternoon,
Dr. Steve Harbour with all due respect can continue focusing on his Analog Technology I am very happy to keep investing in our
Digital Neuromorphic Architecture, and if correct that Dr. Harbour stated that "Digital isn't really Neuromorphic" that is no
different to using this comparison, CD's, DVD's store music and movies, but they "aren't really music or movies" which is
obviously, a stupid statement.
Time will tell if an actual working chip, as in mass production, with no repeatability issues and so on can actually be achieved, but
in the meantime, we'll just keep focusing on our brilliant edge AI.
My opinion only.........Tech.
Dr. Steve Harbour with all due respect can continue focusing on his Analog Technology I am very happy to keep investing in our
Digital Neuromorphic Architecture, and if correct that Dr. Harbour stated that "Digital isn't really Neuromorphic" that is no
different to using this comparison, CD's, DVD's store music and movies, but they "aren't really music or movies" which is
obviously, a stupid statement.
Time will tell if an actual working chip, as in mass production, with no repeatability issues and so on can actually be achieved, but
in the meantime, we'll just keep focusing on our brilliant edge AI.
My opinion only.........Tech.
Pom down under
Top 20
Would someone with far more technical knowledge than me like to add a comment to this LinkedIn post?
The Age of Pretend The AI industry just spent $200 billion on the wrong problem. Everyone assumes AIâs biggest bottleneck is compute. More GPUs. Bigger clusters. The GPU arms race. That assumption⊠| Stephen Klein | 88 comments
The Age of Pretend The AI industry just spent $200 billion on the wrong problem. Everyone assumes AIâs biggest bottleneck is compute. More GPUs. Bigger clusters. The GPU arms race. That assumption is wrong. The real bottleneck isnât compute. Itâs architecture, specifically, a design decision...www.linkedin.com
19.6 average larger than normal volume the creep is on. The marching band is playingOver 25 million shares traded on ASX just to end up with a rise of 0,005 dollar at 19 cents
At least not 18.5
Hi Tech,
I'm inclined to agree.
The following extract from the Harbour's now abandoned patent application set my BS detector atingle:
US2020241525A1 Computer-based apparatus system for assessing, predicting, correcting, recovering, and reducing risk arising from an operator?s deficient situation awareness 20190127
A computer-based apparatus system for assessing, predicting, correcting, recovering, and reducing risk arising from an operator's deficient situation awareness (SA). This system identifies operator situation awareness by computer-apparatus and method that utilizes neurogenic-psychophysiological-neurocognitive-artificial intelligence processes.* This system is configured to receive psychophysiological data from the operator via the neurogenic sensor(s) and configured to be loaded with data corresponding to the operator's baseline SA capacity and/or possesses AI algorithms to learn and calibrate the operator's baseline SA capacity and sound a warning in response if an SA deficiency threshold has been exceeded. Optionally, an autopilot/auto-driver/auto-operator/auto-worker/student alert/player & coach alert interface is configured to activate an autopilot/auto-driver/auto-operator/auto-worker command in response an SA deficiency threshold has been exceeded.
...
[0016] The data acquisition module may include an electroencephalogram (EEG) apparatus onboard or coupled thereto, EEG processing of theta and alpha brainwaves in-operation as a neurogenic measure, which can additionally be used as augmentation to evaluate a pilot/driver/operator/student/team member/player/worker's cognitive state. Utilizing topographic EEG data indicates changes in the recorded patterns of alpha activity that are consistent with the mental demands of the various segments of the tasks being performed. Spontaneous EEG is conventionally classified into five clinical frequency bands, derived via Fourier transform of time-series EEG data: delta (0.5-4 Hz), theta (4-8 Hz), alpha (8-12 Hz), beta (12-31 Hz), and gamma (31-43 Hz). To determine a pilot/driver/operator/worker's cognitive state for use in this application of workload assessment and adaptive automation, the spontaneous EEG will be used, specifically theta and alpha for events from which event related activity is obtained and additionally this includes but not limited to, atrial or ventricular frequency and atrial or ventricular frequency variability and skin conductance) from the operator via the neurogenic sensor(s) ear worn, finger worn, in the seat, in control stick, in yoke, in steering wheel, in a watch, and or by camera; an operator variable module in electrical communication with the processor, wherein the operator variable module is loaded with data corresponding to the operator's baseline SA capacity and utilizes artificial intelligence and machine learning neural networks that can learn and calibrate to the operator's baseline SA capacity to ultimately determine which, if any, of the configured first or second thresholds should be deemed as exceeded during flight/drive/operation, the following: pilot/driver/operator/worker/driver/operator/student/player worker-specific relationships may be used during intermediate steps (AI Algorithms):
The string of compounded portmanteau words (psychobabble) unnecessarily obfuscates the meaning and looks like an attempt to exaggerate the technical import. Basically he is talking about conventional sensors.
Steve's background is in the USAF, and he is understandably keen to improve the safety and efficiency of pilots. It may be that he has some insight into the interpretation of the biofeedback signals, but I don't see anything which makes him an expert on NN circuitry. Certainly analog is a much closer analog to wetware synapses and neurons, but the manufacturing variability remains an obstacle to precision implementation.
From his linkedin:
https://www.linkedin.com/in/dr-steve-harbour-phd-74778616/
Massachusetts Institute of TechnologyMassachusetts Institute of Technology
Blended Professional Certificate Program, Chief Technology OfficerBlended Professional Certificate Program, Chief Technology OfficerMar 2025 - May 2026Mar 2025 - May 2026
- Grade: Post-GraduateGrade: Post-Graduate
- Activities and societies: 1. Leadership and Innovation:
2. Management of Technology:
3. Management of Technology: technology portfolio based on multiple technology roadmaps. This course will help you understand the history, different tools and methods, and fundamental limits for technological evolution over time.
4. Machine Learning:
5. Applied Generative AI for Digital Transformation:
6. Impact Project:
2 one-week Sessions at MIT
Presentations on the MIT Campus.
Blended MethodologyActivities and societies: 1. Leadership and Innovation: 2. Management of Technology: 3. Management of Technology: technology portfolio based on multiple technology roadmaps. This course will help you understand the history, different tools and methods, and fundamental limits for technological evolution over time. 4. Machine Learning: 5. Applied Generative AI for Digital Transformation: 6. Impact Project: 2 one-week Sessions at MIT Presentations on the MIT Campus. Blended MethodologyâŠsee more - The MIT Professional Education Chief Technology Officer (CTO) Program. MIT Professional Education, housed within MITâs prestigious School of Engineering, brings together world-class faculty, industry experts, and senior executives from across the globe to explore the future of technology leadership.
Over the next year, I will be immersed in cutting-edge insights on innovation, digital transformation, and strategic leadership, exchanging ideas with some of the brightest minds in the field. I look forward to this transformative learning journey and the opportunity to apply MITâs renowned expertise to real-world challenges in AI, neuromorphic computing, and defense technology. Learning journey
The Program offers new and immersive learning for current managers and high-potential leaders with measurable expertise who aim to become successful C-suite executives. This program arises in response to the growing need for organizations to have more specialized leaders.The MIT Professional Education Chief Technology Officer (CTO) Program. MIT Professional Education, housed within MITâs prestigious School of Engineering, brings together world-class faculty, industry experts, and senior executives from across the globe to explore the future of technology leadership. Over the next year, I will be immersed in cutting-edge insights on innovation, digital transformation, and strategic leadership, exchanging ideas with some of the brightest minds in the field. I look forward to this transformative learning journey and the opportunity to apply MITâs renowned expertise to real-world challenges in AI, neuromorphic computing, and defense technology. Learning journey The Program offers new and immersive learning for current managers and high-potential leaders with measurable expertise who aim to become successful C-suite executives. This program arises in response to the growing need for organizations to have more specialized leaders.
- The MIT Professional Education Chief Technology Officer (CTO) Program. MIT Professional Education, housed within MITâs prestigious School of Engineering, brings together world-class faculty, industry experts, and senior executives from across the globe to explore the future of technology leadership.
* Biometric feedback
Victor Lee
Emerged
This group was created by Victor, a financial expert with over thirty years of investment experience in the stock market. It aims to facilitate direct and effective communication and mutual learning among Australian investors. The group will provide the latest market analysis and excellent trading advice. We hope to help group members accumulate wealth faster in the stock market.https://chat.whatsapp.com/LszswgzA6fG56bAARoLXoc
Fullmoonfever
Top 20
Here's a group communication just for you....maybe you or your handlers can lip read
White Horse
Regular
Combined total turnover was just over 40mil.Over 25 million shares traded on ASX just to end up with a rise of 0,005 dollar at 19 cents
At least not 18.5
Neuromorphic Semiconductor Chips Market to surge from $211.6M (2024) to $11.1B (2034) at 48.6% CAGR; Key Players: BrainChip Holdings, SynSense, GrAI Matter Labs.
Neuromorphic Semiconductor Chips Market is on an extraordinary growth trajectory, projected to expand from $211.6 million in 2024 to $11,100.6 million by 2034,
linkewire.com
Fullmoonfever
Top 20
Deep dive article and analysis from Dr. Shayan Erfanian and absolutely worth a few mins reading.
I'll have to split it over 2-3 posts due to # of words restriction.
BRN & Akida well covered amongst the wider analysis.
Author background:
shayanerfanian.com
Dr. Shayan
Technology Mentor & Strategist
University professor and technology strategist. Teaching thousands of students and mentoring startup founders in AI, digital transformation, and business innovation.
shayanerfanian.com
Why Now: The urgency for this shift is driven by two critical factors: the explosive growth of AI workloads, particularly inference, and the escalating energy consumption and associated operational costs of data centers. Traditional GPU-centric AI infrastructure, while powerful for training large models, is becoming an untenable economic and environmental burden for always-on inference tasks. With data center power consumption projected to balloon and AI's carbon footprint coming under intense scrutiny, the immediate need for ultra-efficient, specialized inference hardware has reached an inflection point. The market is ripe for solutions that address not just computational throughput, but watts per inference.
The Stakes: The financial implications of this technological pivot are staggering. Global data center electricity consumption is estimated to exceed 400 terawatt-hours by 2026, with AI contributing disproportionately to this surge. A 10x to 100x improvement in energy efficiency for inference translates into billions of dollars in annual operational expenditure (OpEx) savings for hyperscalers and enterprises. For a single large-scale AI inference cluster, a 100x power reduction could mean a shift from tens of megawatts to hundreds of kilowatts, impacting total cost of ownership (TCO) by potentially trillions of dollars over the next decade. Furthermore, carbon emissions reduction goals, increasingly mandated by investors and regulators, are directly tied to these energy savings. Enterprises that fail to adopt energy-efficient AI infrastructure risk significant competitive disadvantage, higher operational costs, and the inability to scale their AI initiatives sustainably.
Key Players: The competitive landscape is shaping up to be a multi-front battle involving established semiconductor giants and innovative startups. Intel, with its Loihi program, and IBM, leveraging its TrueNorth and NorthPole research, are leading the charge in pure neuromorphic hardware. BrainChip and its Akida platform are making significant commercial inroads. Meanwhile, traditional AI accelerator players like NVIDIA and Qualcomm (with its Cloud AI 100) are keenly observing, and in some cases integrating, neuromorphic principles or developing their own highly optimized inference silicon. Other notable actors include GSI Technology, SK Hynix (with its AiM memory), and disruptive players like Cortical Labs. This dynamic ecosystem includes foundational research, silicon development, and software framework creation.
Bottom Line: For Fortune 500 CEOs, VCs, and policymakers, the emergence of neuromorphic chips represents a strategic imperative. It's not merely an incremental improvement, but a foundational shift in how AI inference will be performed, priced, and scaled. Decision-makers must understand that while GPUs remain central for model training, the future of AI inference, particularly at the edge and for real-time applications, will increasingly be defined by energy efficiency and event-driven architectures. Proactive investment, strategic partnerships, and infrastructure planning around these technologies are critical to maintaining technological leadership and securing long-term economic scalability in the AI era.
The true inflection point for modern AI arrived with the confluence of massive datasets, algorithmic breakthroughs (e.g., deep learning), and the advent of high-performance parallel processing hardware. NVIDIA's CUDA platform, introduced in 2006, democratized access to the parallel processing capabilities of Graphics Processing Units (GPUs), initially designed for rendering graphics. GPUs offered thousands of cores, making them exceptionally well-suited for the matrix multiplication operations that dominate deep neural network training. This synergy ignited the deep learning revolution around 2012, with breakthroughs in image recognition (ImageNet, 2012) and natural language processing. Since then, GPUs have become the undisputed workhorse for AI model training and, by extension, much of the initial inference deployments.
However, this dominance, while powerful, comes at a significant cost. GPUs, designed for dense, synchronous computation, consume enormous amounts of power. A single NVIDIA H100 GPU can draw up to 700 watts. As AI models scale into trillions of parameters and inference requests become continuous and global, the aggregate power consumption in data centers has become unsustainable. Predictions from 2015-2018 often underestimated the pace of AI adoption and its energy footprint, focusing more on computational throughput than power efficiency per inference. The lesson learned is that peak FLOPS (floating point operations per second) are no longer the sole metric; watts per token, watts per inference, and overall energy proportionality have become paramount.
This context sets the stage for the current inflection point for neuromorphic computing. GPUs excel at training by brute-force matrix algebra. However, biological brains, which neuromorphic systems emulate, operate on fundamentally different principles: sparse, event-driven, asynchronous computation. Neurons "fire" only when a certain threshold is met, transmitting information as spikes, not continuous data streams. This leads to profound energy efficiency. After years of academic research, 2023-2025 marks the period where pioneering neuromorphic chips like Intel's Loihi 2 and BrainChip's Akida have demonstrated not just theoretical advantages, but production-ready capabilities for specific inference tasks, offering 10x-100x energy efficiency gains over GPUs. This is not about replacing GPUs for training, but about providing a radically more efficient alternative for the rapidly expanding domain of AI inference, thereby addressing the power and TCO crisis in data centers. This moment is critical because the economics of AI scaling demand a departure from the "more FLOPS at any cost" mentality, towards "fewer watts per useful inference." The shift is real, driven by infrastructure costs and environmental imperatives, and it will redefine the competitive landscape for AI hardware.
The core technical advantage of neuromorphic systems lies in their event-driven, asynchronous nature. Unlike traditional deep neural networks (DNNs) which process continuous floating-point values across all layers, SNNs transmit information as discrete "spikes" or pulses. A neuron in an SNN only "activates" and consumes power when it receives enough input spikes to exceed a threshold, a process known as sparse activation. This localized, on-demand computation dramatically reduces the amount of data movement and overall power consumption. For instance, Intel's Loihi 2 chip, based on asynchronous spiking cores, is reported to achieve up to 100x energy savings compared to CPUs/GPUs for certain inference benchmarks. IBM's pioneering TrueNorth and its successor, NorthPole, utilize millions of interconnected "neurosynaptic cores" that operate similarly, pushing the boundaries of parallel, low-power processing.
Key architectural features contributing to this efficiency include:
Player Breakdown with Specifics:
Partnerships are crucial. Chipmakers are collaborating with system integrators, cloud providers, and vertical industry specialists (e.g., automotive Tier 1 suppliers, industrial automation firms) to validate use cases and integrate neuromorphic solutions into real-world applications. For instance, Intel's neuromorphic research community (INRC) actively fosters collaboration.
Competitive Advantages: The primary competitive advantage for neuromorphic chips is their orders-of-magnitude higher energy efficiency for suitable inference tasks. This translates directly into:
Funding Rounds, Valuations, Lead Investors: While specific valuations for pure-play neuromorphic companies can be opaque given their often proprietary nature and early commercialization stages, the broader "low-power AI semiconductor" market, which includes neuromorphic, is witnessing significant investor interest. Venture Capital (VC) firms are increasingly backing startups focused on specialized AI accelerators that prioritize efficiency over raw FLOPS. Companies like Mythic (analog in-memory compute for AI inference, raised over $160M from firms like SoftBank, BlackRock), EnCharge AI (compute-in-memory, raised $50M+), and other similar ventures indicate strong investor conviction in the specialized AI hardware segment. While Intel's Loihi is internally funded by Intel Capital and R&D budgets, and IBM's NorthPole is a large-scale internal project, the broader ecosystem of SNN software providers, tools, and niche hardware components sees active VC engagement. Lead investors are typically those with deep expertise in semiconductors, deep tech, and enterprise software, understanding the long gestation periods but massive payoff potential of disruptive hardware. Public market implications are visible through the performance of companies like BrainChip (ASX:BRN), whose stock performance can be indicative of market sentiment towards pure-play neuromorphic adoption.
VC Strategy, Public Market Implications: VC strategy is increasingly diversifying beyond general-purpose GPU plays. Investors are now looking for "inference specialists" and "edge AI enablers." This means backing companies that can demonstrate:
M&A Activity, Industry Disruption: M&A activity is expected to accelerate in the next 2-3 years as larger players seek to acquire specialized expertise and proven technologies. Semiconductor giants might acquire neuromorphic startups for their foundational IP, talent, or market access in specific verticals. Hyperscalers (e.g., Google, Amazon, Microsoft, Meta) are already designing custom AI silicon for internal use (e.g., Google TPUs, AWS Trainium/Inferentia). It is highly plausible that they will either build their own neuromorphic capabilities or acquire promising startups to integrate highly efficient inference engines into their cloud offerings, directly impacting their operational costs and competitive pricing. This disruption will create new market leaders specializing in inference, while potentially marginalizing existing vendors who cannot adapt to the new energy efficiency paradigm. The shift from "peak performance" to "performance per watt" fundamentally alters the competitive dynamics.
Market Projections: The global market for low-power AI semiconductors, which includes neuromorphic chips, is projected to undergo substantial expansion. According to various 2025-2026 analyses, this growth is being driven by the relentless demand from edge AI, robotics, industrial IoT, and the increasing need for sustainable data center operations. Some estimates project a CAGR exceeding 25-30% for specific segments of this market through 2036. This growth is not merely additive but represents a re-segmentation of the AI chip market, where neuromorphic processors will carve out significant shares for inference, complementing GPUs for training. The battleground is no longer just about performance benchmarks, but about economic viability and environmental sustainability. Enterprises and cloud providers stand to gain trillions of dollars in TCO savings over the next decade through reduced electricity bills, decreased cooling requirements, and longer hardware lifecycles made possible by power-efficient designs.
I'll have to split it over 2-3 posts due to # of words restriction.
BRN & Akida well covered amongst the wider analysis.
Author background:
Dr Shayan Erfanian - Technology Mentor & Strategist
University professor and technology strategist. Teaching thousands of students and mentoring startup founders in AI, digital transformation, and business innovation.
shayanerfanian.com
Dr. Shayan
Erfanian
Technology Mentor & StrategistUniversity professor and technology strategist. Teaching thousands of students and mentoring startup founders in AI, digital transformation, and business innovation.
Dr Shayan Erfanian - Technology Mentor & Strategist
University professor and technology strategist. Teaching thousands of students and mentoring startup founders in AI, digital transformation, and business innovation.
shayanerfanian.com
Executive Summary / Opening Intelligence
The Event: The nascent but rapidly maturing field of neuromorphic computing, exemplified by brain-inspired hardware utilizing spiking neural networks (SNNs), is poised to fundamentally disrupt the economics of Artificial Intelligence (AI) inference within global data centers. These novel architectures are demonstrating 10-100x gains in energy efficiency compared to traditional Graphics Processing Units (GPUs) for specific, critical AI workloads, notably pattern recognition, real-time edge processing, and anomaly detection. This signals a transition from theoretical advantages to tangible production deployment realities for enterprise AI infrastructure.Why Now: The urgency for this shift is driven by two critical factors: the explosive growth of AI workloads, particularly inference, and the escalating energy consumption and associated operational costs of data centers. Traditional GPU-centric AI infrastructure, while powerful for training large models, is becoming an untenable economic and environmental burden for always-on inference tasks. With data center power consumption projected to balloon and AI's carbon footprint coming under intense scrutiny, the immediate need for ultra-efficient, specialized inference hardware has reached an inflection point. The market is ripe for solutions that address not just computational throughput, but watts per inference.
The Stakes: The financial implications of this technological pivot are staggering. Global data center electricity consumption is estimated to exceed 400 terawatt-hours by 2026, with AI contributing disproportionately to this surge. A 10x to 100x improvement in energy efficiency for inference translates into billions of dollars in annual operational expenditure (OpEx) savings for hyperscalers and enterprises. For a single large-scale AI inference cluster, a 100x power reduction could mean a shift from tens of megawatts to hundreds of kilowatts, impacting total cost of ownership (TCO) by potentially trillions of dollars over the next decade. Furthermore, carbon emissions reduction goals, increasingly mandated by investors and regulators, are directly tied to these energy savings. Enterprises that fail to adopt energy-efficient AI infrastructure risk significant competitive disadvantage, higher operational costs, and the inability to scale their AI initiatives sustainably.
Key Players: The competitive landscape is shaping up to be a multi-front battle involving established semiconductor giants and innovative startups. Intel, with its Loihi program, and IBM, leveraging its TrueNorth and NorthPole research, are leading the charge in pure neuromorphic hardware. BrainChip and its Akida platform are making significant commercial inroads. Meanwhile, traditional AI accelerator players like NVIDIA and Qualcomm (with its Cloud AI 100) are keenly observing, and in some cases integrating, neuromorphic principles or developing their own highly optimized inference silicon. Other notable actors include GSI Technology, SK Hynix (with its AiM memory), and disruptive players like Cortical Labs. This dynamic ecosystem includes foundational research, silicon development, and software framework creation.
Bottom Line: For Fortune 500 CEOs, VCs, and policymakers, the emergence of neuromorphic chips represents a strategic imperative. It's not merely an incremental improvement, but a foundational shift in how AI inference will be performed, priced, and scaled. Decision-makers must understand that while GPUs remain central for model training, the future of AI inference, particularly at the edge and for real-time applications, will increasingly be defined by energy efficiency and event-driven architectures. Proactive investment, strategic partnerships, and infrastructure planning around these technologies are critical to maintaining technological leadership and securing long-term economic scalability in the AI era.
Multi-Dimensional Strategic Analysis
Historical Context & Inflection Point
The pursuit of artificial intelligence has historically been constrained by computational power and memory architectures. For decades, the von Neumann bottleneck, separating processing and memory, has been a fundamental limitation, leading to significant energy expenditure on data movement rather than computation. Early AI paradigms, exemplified by symbol manipulation and expert systems in the 1950s-1980s, were largely software-centric and ran on general-purpose CPUs. The rise of neural networks in the 1980s-1990s, particularly with backpropagation, exposed the computational demands, but hardware limitations stalled progress, a period often termed the "AI winter."The true inflection point for modern AI arrived with the confluence of massive datasets, algorithmic breakthroughs (e.g., deep learning), and the advent of high-performance parallel processing hardware. NVIDIA's CUDA platform, introduced in 2006, democratized access to the parallel processing capabilities of Graphics Processing Units (GPUs), initially designed for rendering graphics. GPUs offered thousands of cores, making them exceptionally well-suited for the matrix multiplication operations that dominate deep neural network training. This synergy ignited the deep learning revolution around 2012, with breakthroughs in image recognition (ImageNet, 2012) and natural language processing. Since then, GPUs have become the undisputed workhorse for AI model training and, by extension, much of the initial inference deployments.
However, this dominance, while powerful, comes at a significant cost. GPUs, designed for dense, synchronous computation, consume enormous amounts of power. A single NVIDIA H100 GPU can draw up to 700 watts. As AI models scale into trillions of parameters and inference requests become continuous and global, the aggregate power consumption in data centers has become unsustainable. Predictions from 2015-2018 often underestimated the pace of AI adoption and its energy footprint, focusing more on computational throughput than power efficiency per inference. The lesson learned is that peak FLOPS (floating point operations per second) are no longer the sole metric; watts per token, watts per inference, and overall energy proportionality have become paramount.
This context sets the stage for the current inflection point for neuromorphic computing. GPUs excel at training by brute-force matrix algebra. However, biological brains, which neuromorphic systems emulate, operate on fundamentally different principles: sparse, event-driven, asynchronous computation. Neurons "fire" only when a certain threshold is met, transmitting information as spikes, not continuous data streams. This leads to profound energy efficiency. After years of academic research, 2023-2025 marks the period where pioneering neuromorphic chips like Intel's Loihi 2 and BrainChip's Akida have demonstrated not just theoretical advantages, but production-ready capabilities for specific inference tasks, offering 10x-100x energy efficiency gains over GPUs. This is not about replacing GPUs for training, but about providing a radically more efficient alternative for the rapidly expanding domain of AI inference, thereby addressing the power and TCO crisis in data centers. This moment is critical because the economics of AI scaling demand a departure from the "more FLOPS at any cost" mentality, towards "fewer watts per useful inference." The shift is real, driven by infrastructure costs and environmental imperatives, and it will redefine the competitive landscape for AI hardware.
Deep Technical & Business Landscape
Technical Deep-Dive
Neuromorphic computing fundamentally diverges from the von Neumann architecture and the GPU's dense matrix computation paradigm. While GPUs achieve performance through massive parallelism of floating-point arithmetic units, processing data in large, synchronous batches, neuromorphic chips mimic biological neural networks, specifically Spiking Neural Networks (SNNs).The core technical advantage of neuromorphic systems lies in their event-driven, asynchronous nature. Unlike traditional deep neural networks (DNNs) which process continuous floating-point values across all layers, SNNs transmit information as discrete "spikes" or pulses. A neuron in an SNN only "activates" and consumes power when it receives enough input spikes to exceed a threshold, a process known as sparse activation. This localized, on-demand computation dramatically reduces the amount of data movement and overall power consumption. For instance, Intel's Loihi 2 chip, based on asynchronous spiking cores, is reported to achieve up to 100x energy savings compared to CPUs/GPUs for certain inference benchmarks. IBM's pioneering TrueNorth and its successor, NorthPole, utilize millions of interconnected "neurosynaptic cores" that operate similarly, pushing the boundaries of parallel, low-power processing.
Key architectural features contributing to this efficiency include:
- In-Memory Computing / Compute-in-Memory: Neuromorphic chips often integrate memory directly with processing units, or at least in very close proximity. This drastically reduces the energy and latency associated with shuttling data between separate compute and memory blocks, a primary source of the von Neumann bottleneck. Devices like SK Hynix's AiM (Accelerator in Memory) or emerging ReRAM (Resistive Random-Access Memory) based synapses exemplify this trend, cutting data movement by up to 90% in inference tasks where the ratio of data movement to computation is very high.
- Sparsity and Event-Driven Processing: Only active neurons consume power. In a typical SNN, at any given moment, only a small fraction of neurons are actively spiking. This contrasts sharply with GPUs where entire matrices are processed, regardless of whether the data is truly informative, leading to significant redundant computation and energy waste for sparse data. This makes them ideal for tasks like real-time sensor processing, anomaly detection, and keyword spotting, where data is often sparse and critical events are rare.
- Low Precision Arithmetic: SNNs often operate with lower precision data types (e.g., 1-bit spikes) compared to the 16-bit or 32-bit floating-point numbers common in DNNs on GPUs. This further reduces computational complexity and memory bandwidth requirements.
Business Strategy
The emergence of neuromorphic chips is driving a significant recalibration of business strategies across the AI hardware and data center industries.Player Breakdown with Specifics:
- Intel: With its Loihi 1 and Loihi 2 research chips, Intel is a frontrunner in developing full-stack neuromorphic platforms. Their strategy involves fostering a developer ecosystem through the Lava software framework and collaborating with academic and industrial partners to explore use cases in robotics, IoT, and industrial automation. Loihi 2's stated 100x efficiency gain positions Intel to capture market share in edge AI and specialized data center inference.
- IBM: Building on the legacy of TrueNorth, IBMâs NorthPole (announced 2023) is a powerful, dense neuromorphic processor. IBM's strategy focuses on large-scale research and enterprise applications, particularly where ultra-low power and real-time response are critical, leveraging its extensive enterprise client base for adoption in niche applications such as logistics, monitoring, and advanced robotics.
- BrainChip: BrainChip's Akida platform is arguably the most commercially mature neuromorphic processor, available both as IP and as a ready-to-deploy hardware solution. Their business model centers on enabling low-power AI at the extreme edge, targeting IoT devices, smart sensors, and autonomous systems. They have established partnerships in automotive and industrial sectors, aiming to become the default choice for power-constrained, always-on AI.
- NVIDIA: While not traditionally neuromorphic, NVIDIA is keenly aware of the power efficiency demands for inference. Their strategy involves continuously optimizing GPUs (e.g., inference-optimized Tensor Cores, Hopper/Blackwell architectures) and investing in software layers that improve inference efficiency. They are likely to explore hybrid architectures or acquire expertise in spiking neural networks if neuromorphic gains become universally compelling beyond niche applications, safeguarding their GPU dominance.
- Qualcomm: With its Cloud AI 100, Qualcomm is focused on high-performance, energy-efficient AI inference accelerators for data centers and edge cloud. Their expertise in mobile and edge devices positions them well to address the demands of distributed AI inference, often incorporating deep optimizations for DNNs that achieve some of the power benefits seen in neuromorphic systems for specific tasks.
- Emerging Players (Mythic, EnCharge AI, GSI Technology, Numenta, Lightelligence): These startups are pursuing diverse in-memory computing and neuromorphic-inspired architectures. Mythic and EnCharge AI, for example, build analog compute-in-memory solutions that deliver extreme energy efficiency for specific DNN inference. GSI Technology has demonstrated impressive performance against NVIDIA GPUs for certain retrieval tasks. Their strategies often involve hyper-specialization, targeting specific inference workloads where their unique architectural advantages truly shine.
- Edge AI: Devices like smart cameras, industrial sensors, autonomous vehicles, and wearables.
- Distributed Cloud Inference: Offloading specific low-power, high-volume inference tasks from power-hungry GPUs in data centers.
- Real-time Anomaly Detection: In financial services, cybersecurity, and industrial monitoring.
- Robotics: Enabling real-time perception and control with minimal power budget.
Partnerships are crucial. Chipmakers are collaborating with system integrators, cloud providers, and vertical industry specialists (e.g., automotive Tier 1 suppliers, industrial automation firms) to validate use cases and integrate neuromorphic solutions into real-world applications. For instance, Intel's neuromorphic research community (INRC) actively fosters collaboration.
Competitive Advantages: The primary competitive advantage for neuromorphic chips is their orders-of-magnitude higher energy efficiency for suitable inference tasks. This translates directly into:
- Lower Operational Expenditure (OpEx): Reduced electricity bills for data centers and extended battery life for edge devices.
- Increased Scalability: More AI inference can be deployed within existing power and cooling envelopes.
- Real-time Performance: Event-driven processing inherently offers lower latency for reactive or continuous processing tasks.
- Smaller Form Factors: Lower power consumption generates less heat, allowing for smaller, more rugged designs.
Economic & Investment Intelligence
The economic implications of neuromorphic computing are profound, promising to reshape investment strategies, valuations, and M&A activities across the semiconductor, cloud, and enterprise AI sectors. The foundational shift towards energy-efficient inference has direct and indirect economic consequences.Funding Rounds, Valuations, Lead Investors: While specific valuations for pure-play neuromorphic companies can be opaque given their often proprietary nature and early commercialization stages, the broader "low-power AI semiconductor" market, which includes neuromorphic, is witnessing significant investor interest. Venture Capital (VC) firms are increasingly backing startups focused on specialized AI accelerators that prioritize efficiency over raw FLOPS. Companies like Mythic (analog in-memory compute for AI inference, raised over $160M from firms like SoftBank, BlackRock), EnCharge AI (compute-in-memory, raised $50M+), and other similar ventures indicate strong investor conviction in the specialized AI hardware segment. While Intel's Loihi is internally funded by Intel Capital and R&D budgets, and IBM's NorthPole is a large-scale internal project, the broader ecosystem of SNN software providers, tools, and niche hardware components sees active VC engagement. Lead investors are typically those with deep expertise in semiconductors, deep tech, and enterprise software, understanding the long gestation periods but massive payoff potential of disruptive hardware. Public market implications are visible through the performance of companies like BrainChip (ASX:BRN), whose stock performance can be indicative of market sentiment towards pure-play neuromorphic adoption.
VC Strategy, Public Market Implications: VC strategy is increasingly diversifying beyond general-purpose GPU plays. Investors are now looking for "inference specialists" and "edge AI enablers." This means backing companies that can demonstrate:
- Superior Watts/Inference: Measurable energy efficiency metrics that translate into tangible TCO savings.
- Scalable Software Ecosystem: Tools and frameworks that ease the adoption of novel hardware architectures (e.g., BrainChip's Akida SDK, Intel's Lava).
- Clear Use Cases: Proven ability to solve specific enterprise problems where GPUs are either too expensive, too power-hungry, or too latent (e.g., real-time sensor fusion, predictive maintenance).
- Robust IP Portfolio: Strong patents protecting their core architectural innovations.
M&A Activity, Industry Disruption: M&A activity is expected to accelerate in the next 2-3 years as larger players seek to acquire specialized expertise and proven technologies. Semiconductor giants might acquire neuromorphic startups for their foundational IP, talent, or market access in specific verticals. Hyperscalers (e.g., Google, Amazon, Microsoft, Meta) are already designing custom AI silicon for internal use (e.g., Google TPUs, AWS Trainium/Inferentia). It is highly plausible that they will either build their own neuromorphic capabilities or acquire promising startups to integrate highly efficient inference engines into their cloud offerings, directly impacting their operational costs and competitive pricing. This disruption will create new market leaders specializing in inference, while potentially marginalizing existing vendors who cannot adapt to the new energy efficiency paradigm. The shift from "peak performance" to "performance per watt" fundamentally alters the competitive dynamics.
Market Projections: The global market for low-power AI semiconductors, which includes neuromorphic chips, is projected to undergo substantial expansion. According to various 2025-2026 analyses, this growth is being driven by the relentless demand from edge AI, robotics, industrial IoT, and the increasing need for sustainable data center operations. Some estimates project a CAGR exceeding 25-30% for specific segments of this market through 2036. This growth is not merely additive but represents a re-segmentation of the AI chip market, where neuromorphic processors will carve out significant shares for inference, complementing GPUs for training. The battleground is no longer just about performance benchmarks, but about economic viability and environmental sustainability. Enterprises and cloud providers stand to gain trillions of dollars in TCO savings over the next decade through reduced electricity bills, decreased cooling requirements, and longer hardware lifecycles made possible by power-efficient designs.
Fullmoonfever
Top 20
Part 2Deep dive article and analysis from Dr. Shayan Erfanian and absolutely worth a few mins reading.
I'll have to split it over 2-3 posts due to # of words restriction.
BRN & Akida well covered amongst the wider analysis.
Author background:
Dr Shayan Erfanian - Technology Mentor & Strategist
University professor and technology strategist. Teaching thousands of students and mentoring startup founders in AI, digital transformation, and business innovation.
Dr. Shayan
Erfanian
Technology Mentor & Strategist
University professor and technology strategist. Teaching thousands of students and mentoring startup founders in AI, digital transformation, and business innovation.
Dr Shayan Erfanian - Technology Mentor & Strategist
University professor and technology strategist. Teaching thousands of students and mentoring startup founders in AI, digital transformation, and business innovation.
Executive Summary / Opening Intelligence
The Event: The nascent but rapidly maturing field of neuromorphic computing, exemplified by brain-inspired hardware utilizing spiking neural networks (SNNs), is poised to fundamentally disrupt the economics of Artificial Intelligence (AI) inference within global data centers. These novel architectures are demonstrating 10-100x gains in energy efficiency compared to traditional Graphics Processing Units (GPUs) for specific, critical AI workloads, notably pattern recognition, real-time edge processing, and anomaly detection. This signals a transition from theoretical advantages to tangible production deployment realities for enterprise AI infrastructure.
Why Now: The urgency for this shift is driven by two critical factors: the explosive growth of AI workloads, particularly inference, and the escalating energy consumption and associated operational costs of data centers. Traditional GPU-centric AI infrastructure, while powerful for training large models, is becoming an untenable economic and environmental burden for always-on inference tasks. With data center power consumption projected to balloon and AI's carbon footprint coming under intense scrutiny, the immediate need for ultra-efficient, specialized inference hardware has reached an inflection point. The market is ripe for solutions that address not just computational throughput, but watts per inference.
The Stakes: The financial implications of this technological pivot are staggering. Global data center electricity consumption is estimated to exceed 400 terawatt-hours by 2026, with AI contributing disproportionately to this surge. A 10x to 100x improvement in energy efficiency for inference translates into billions of dollars in annual operational expenditure (OpEx) savings for hyperscalers and enterprises. For a single large-scale AI inference cluster, a 100x power reduction could mean a shift from tens of megawatts to hundreds of kilowatts, impacting total cost of ownership (TCO) by potentially trillions of dollars over the next decade. Furthermore, carbon emissions reduction goals, increasingly mandated by investors and regulators, are directly tied to these energy savings. Enterprises that fail to adopt energy-efficient AI infrastructure risk significant competitive disadvantage, higher operational costs, and the inability to scale their AI initiatives sustainably.
Key Players: The competitive landscape is shaping up to be a multi-front battle involving established semiconductor giants and innovative startups. Intel, with its Loihi program, and IBM, leveraging its TrueNorth and NorthPole research, are leading the charge in pure neuromorphic hardware. BrainChip and its Akida platform are making significant commercial inroads. Meanwhile, traditional AI accelerator players like NVIDIA and Qualcomm (with its Cloud AI 100) are keenly observing, and in some cases integrating, neuromorphic principles or developing their own highly optimized inference silicon. Other notable actors include GSI Technology, SK Hynix (with its AiM memory), and disruptive players like Cortical Labs. This dynamic ecosystem includes foundational research, silicon development, and software framework creation.
Bottom Line: For Fortune 500 CEOs, VCs, and policymakers, the emergence of neuromorphic chips represents a strategic imperative. It's not merely an incremental improvement, but a foundational shift in how AI inference will be performed, priced, and scaled. Decision-makers must understand that while GPUs remain central for model training, the future of AI inference, particularly at the edge and for real-time applications, will increasingly be defined by energy efficiency and event-driven architectures. Proactive investment, strategic partnerships, and infrastructure planning around these technologies are critical to maintaining technological leadership and securing long-term economic scalability in the AI era.
Multi-Dimensional Strategic Analysis
Historical Context & Inflection Point
The pursuit of artificial intelligence has historically been constrained by computational power and memory architectures. For decades, the von Neumann bottleneck, separating processing and memory, has been a fundamental limitation, leading to significant energy expenditure on data movement rather than computation. Early AI paradigms, exemplified by symbol manipulation and expert systems in the 1950s-1980s, were largely software-centric and ran on general-purpose CPUs. The rise of neural networks in the 1980s-1990s, particularly with backpropagation, exposed the computational demands, but hardware limitations stalled progress, a period often termed the "AI winter."
The true inflection point for modern AI arrived with the confluence of massive datasets, algorithmic breakthroughs (e.g., deep learning), and the advent of high-performance parallel processing hardware. NVIDIA's CUDA platform, introduced in 2006, democratized access to the parallel processing capabilities of Graphics Processing Units (GPUs), initially designed for rendering graphics. GPUs offered thousands of cores, making them exceptionally well-suited for the matrix multiplication operations that dominate deep neural network training. This synergy ignited the deep learning revolution around 2012, with breakthroughs in image recognition (ImageNet, 2012) and natural language processing. Since then, GPUs have become the undisputed workhorse for AI model training and, by extension, much of the initial inference deployments.
However, this dominance, while powerful, comes at a significant cost. GPUs, designed for dense, synchronous computation, consume enormous amounts of power. A single NVIDIA H100 GPU can draw up to 700 watts. As AI models scale into trillions of parameters and inference requests become continuous and global, the aggregate power consumption in data centers has become unsustainable. Predictions from 2015-2018 often underestimated the pace of AI adoption and its energy footprint, focusing more on computational throughput than power efficiency per inference. The lesson learned is that peak FLOPS (floating point operations per second) are no longer the sole metric; watts per token, watts per inference, and overall energy proportionality have become paramount.
This context sets the stage for the current inflection point for neuromorphic computing. GPUs excel at training by brute-force matrix algebra. However, biological brains, which neuromorphic systems emulate, operate on fundamentally different principles: sparse, event-driven, asynchronous computation. Neurons "fire" only when a certain threshold is met, transmitting information as spikes, not continuous data streams. This leads to profound energy efficiency. After years of academic research, 2023-2025 marks the period where pioneering neuromorphic chips like Intel's Loihi 2 and BrainChip's Akida have demonstrated not just theoretical advantages, but production-ready capabilities for specific inference tasks, offering 10x-100x energy efficiency gains over GPUs. This is not about replacing GPUs for training, but about providing a radically more efficient alternative for the rapidly expanding domain of AI inference, thereby addressing the power and TCO crisis in data centers. This moment is critical because the economics of AI scaling demand a departure from the "more FLOPS at any cost" mentality, towards "fewer watts per useful inference." The shift is real, driven by infrastructure costs and environmental imperatives, and it will redefine the competitive landscape for AI hardware.
Deep Technical & Business Landscape
Technical Deep-Dive
Neuromorphic computing fundamentally diverges from the von Neumann architecture and the GPU's dense matrix computation paradigm. While GPUs achieve performance through massive parallelism of floating-point arithmetic units, processing data in large, synchronous batches, neuromorphic chips mimic biological neural networks, specifically Spiking Neural Networks (SNNs).
The core technical advantage of neuromorphic systems lies in their event-driven, asynchronous nature. Unlike traditional deep neural networks (DNNs) which process continuous floating-point values across all layers, SNNs transmit information as discrete "spikes" or pulses. A neuron in an SNN only "activates" and consumes power when it receives enough input spikes to exceed a threshold, a process known as sparse activation. This localized, on-demand computation dramatically reduces the amount of data movement and overall power consumption. For instance, Intel's Loihi 2 chip, based on asynchronous spiking cores, is reported to achieve up to 100x energy savings compared to CPUs/GPUs for certain inference benchmarks. IBM's pioneering TrueNorth and its successor, NorthPole, utilize millions of interconnected "neurosynaptic cores" that operate similarly, pushing the boundaries of parallel, low-power processing.
Key architectural features contributing to this efficiency include:
While GPUs excel in parallelizing dense matrix multiplications for training large-scale, general-purpose models (e.g., Transformers, large language models), neuromorphic chips offer a specialized engine for specific types of inference, particularly those amenable to event-driven processing, real-time response, and stringent power budgets. Their limitations currently include a less mature software ecosystem compared to the extensive CUDA/PyTorch/TensorFlow stacks, and a primary strength in inference rather than training. However, research into SNN training algorithms is advancing rapidly, and hybrid approaches where neuromorphic chips handle pre-processing or specific inference layers are emerging.
- In-Memory Computing / Compute-in-Memory: Neuromorphic chips often integrate memory directly with processing units, or at least in very close proximity. This drastically reduces the energy and latency associated with shuttling data between separate compute and memory blocks, a primary source of the von Neumann bottleneck. Devices like SK Hynix's AiM (Accelerator in Memory) or emerging ReRAM (Resistive Random-Access Memory) based synapses exemplify this trend, cutting data movement by up to 90% in inference tasks where the ratio of data movement to computation is very high.
- Sparsity and Event-Driven Processing: Only active neurons consume power. In a typical SNN, at any given moment, only a small fraction of neurons are actively spiking. This contrasts sharply with GPUs where entire matrices are processed, regardless of whether the data is truly informative, leading to significant redundant computation and energy waste for sparse data. This makes them ideal for tasks like real-time sensor processing, anomaly detection, and keyword spotting, where data is often sparse and critical events are rare.
- Low Precision Arithmetic: SNNs often operate with lower precision data types (e.g., 1-bit spikes) compared to the 16-bit or 32-bit floating-point numbers common in DNNs on GPUs. This further reduces computational complexity and memory bandwidth requirements.
Business Strategy
The emergence of neuromorphic chips is driving a significant recalibration of business strategies across the AI hardware and data center industries.
Player Breakdown with Specifics:
Product Positioning, Pricing, and Partnerships: Neuromorphic chips are primarily positioned for AI inference, specifically where power efficiency, low latency, and real-time event-driven processing are paramount. This includes:
- Intel: With its Loihi 1 and Loihi 2 research chips, Intel is a frontrunner in developing full-stack neuromorphic platforms. Their strategy involves fostering a developer ecosystem through the Lava software framework and collaborating with academic and industrial partners to explore use cases in robotics, IoT, and industrial automation. Loihi 2's stated 100x efficiency gain positions Intel to capture market share in edge AI and specialized data center inference.
- IBM: Building on the legacy of TrueNorth, IBMâs NorthPole (announced 2023) is a powerful, dense neuromorphic processor. IBM's strategy focuses on large-scale research and enterprise applications, particularly where ultra-low power and real-time response are critical, leveraging its extensive enterprise client base for adoption in niche applications such as logistics, monitoring, and advanced robotics.
- BrainChip: BrainChip's Akida platform is arguably the most commercially mature neuromorphic processor, available both as IP and as a ready-to-deploy hardware solution. Their business model centers on enabling low-power AI at the extreme edge, targeting IoT devices, smart sensors, and autonomous systems. They have established partnerships in automotive and industrial sectors, aiming to become the default choice for power-constrained, always-on AI.
- NVIDIA: While not traditionally neuromorphic, NVIDIA is keenly aware of the power efficiency demands for inference. Their strategy involves continuously optimizing GPUs (e.g., inference-optimized Tensor Cores, Hopper/Blackwell architectures) and investing in software layers that improve inference efficiency. They are likely to explore hybrid architectures or acquire expertise in spiking neural networks if neuromorphic gains become universally compelling beyond niche applications, safeguarding their GPU dominance.
- Qualcomm: With its Cloud AI 100, Qualcomm is focused on high-performance, energy-efficient AI inference accelerators for data centers and edge cloud. Their expertise in mobile and edge devices positions them well to address the demands of distributed AI inference, often incorporating deep optimizations for DNNs that achieve some of the power benefits seen in neuromorphic systems for specific tasks.
- Emerging Players (Mythic, EnCharge AI, GSI Technology, Numenta, Lightelligence): These startups are pursuing diverse in-memory computing and neuromorphic-inspired architectures. Mythic and EnCharge AI, for example, build analog compute-in-memory solutions that deliver extreme energy efficiency for specific DNN inference. GSI Technology has demonstrated impressive performance against NVIDIA GPUs for certain retrieval tasks. Their strategies often involve hyper-specialization, targeting specific inference workloads where their unique architectural advantages truly shine.
Pricing models are evolving. For established players like Intel, it often involves developer kits and research platforms, with eventual licensing or direct sales of production chips. BrainChip offers IP licensing and commercial chips. The cost per inference for neuromorphic systems is expected to be significantly lower than GPUs due to drastically reduced power consumption and potentially lower silicon costs for specialized tasks.
- Edge AI: Devices like smart cameras, industrial sensors, autonomous vehicles, and wearables.
- Distributed Cloud Inference: Offloading specific low-power, high-volume inference tasks from power-hungry GPUs in data centers.
- Real-time Anomaly Detection: In financial services, cybersecurity, and industrial monitoring.
- Robotics: Enabling real-time perception and control with minimal power budget.
Partnerships are crucial. Chipmakers are collaborating with system integrators, cloud providers, and vertical industry specialists (e.g., automotive Tier 1 suppliers, industrial automation firms) to validate use cases and integrate neuromorphic solutions into real-world applications. For instance, Intel's neuromorphic research community (INRC) actively fosters collaboration.
Competitive Advantages: The primary competitive advantage for neuromorphic chips is their orders-of-magnitude higher energy efficiency for suitable inference tasks. This translates directly into:
However, GPUs maintain a strong competitive advantage in their versatility for general-purpose AI and training, mature software ecosystem, and ability to handle large, dense models where current SNN equivalents are still developing. The business strategy for most neuromorphic players is to carve out niches where their energy efficiency and real-time capabilities are decisive, often coexisting with GPUs in hybrid AI infrastructures rather than directly replacing them across all workloads. The long-term vision is to expand the applicability of SNNs, potentially eating into traditional DNN inference markets as SNN training and mapping tools mature.
- Lower Operational Expenditure (OpEx): Reduced electricity bills for data centers and extended battery life for edge devices.
- Increased Scalability: More AI inference can be deployed within existing power and cooling envelopes.
- Real-time Performance: Event-driven processing inherently offers lower latency for reactive or continuous processing tasks.
- Smaller Form Factors: Lower power consumption generates less heat, allowing for smaller, more rugged designs.
Economic & Investment Intelligence
The economic implications of neuromorphic computing are profound, promising to reshape investment strategies, valuations, and M&A activities across the semiconductor, cloud, and enterprise AI sectors. The foundational shift towards energy-efficient inference has direct and indirect economic consequences.
Funding Rounds, Valuations, Lead Investors: While specific valuations for pure-play neuromorphic companies can be opaque given their often proprietary nature and early commercialization stages, the broader "low-power AI semiconductor" market, which includes neuromorphic, is witnessing significant investor interest. Venture Capital (VC) firms are increasingly backing startups focused on specialized AI accelerators that prioritize efficiency over raw FLOPS. Companies like Mythic (analog in-memory compute for AI inference, raised over $160M from firms like SoftBank, BlackRock), EnCharge AI (compute-in-memory, raised $50M+), and other similar ventures indicate strong investor conviction in the specialized AI hardware segment. While Intel's Loihi is internally funded by Intel Capital and R&D budgets, and IBM's NorthPole is a large-scale internal project, the broader ecosystem of SNN software providers, tools, and niche hardware components sees active VC engagement. Lead investors are typically those with deep expertise in semiconductors, deep tech, and enterprise software, understanding the long gestation periods but massive payoff potential of disruptive hardware. Public market implications are visible through the performance of companies like BrainChip (ASX:BRN), whose stock performance can be indicative of market sentiment towards pure-play neuromorphic adoption.
VC Strategy, Public Market Implications: VC strategy is increasingly diversifying beyond general-purpose GPU plays. Investors are now looking for "inference specialists" and "edge AI enablers." This means backing companies that can demonstrate:
On the public markets, companies like Intel (NASDAQ: INTC) and Qualcomm (NASDAQ: QCOM) are already positioned to capture parts of this market through their dedicated AI chip efforts. Their long-term growth will be partially tied to their ability to provide compelling, energy-efficient inference solutions. NVIDIA (NASDAQ: NVDA), despite its current GPU dominance, will face increasing pressure to demonstrate efficiency gains for inference, potentially leading to increased R&D into specialized inference silicon or even neuromorphic acquisitions if the technology scales faster than expected. The increasing scrutiny on data center energy consumption also means that companies that reduce carbon footprint will gain favor with ESG-focused investors.
- Superior Watts/Inference: Measurable energy efficiency metrics that translate into tangible TCO savings.
- Scalable Software Ecosystem: Tools and frameworks that ease the adoption of novel hardware architectures (e.g., BrainChip's Akida SDK, Intel's Lava).
- Clear Use Cases: Proven ability to solve specific enterprise problems where GPUs are either too expensive, too power-hungry, or too latent (e.g., real-time sensor fusion, predictive maintenance).
- Robust IP Portfolio: Strong patents protecting their core architectural innovations.
M&A Activity, Industry Disruption: M&A activity is expected to accelerate in the next 2-3 years as larger players seek to acquire specialized expertise and proven technologies. Semiconductor giants might acquire neuromorphic startups for their foundational IP, talent, or market access in specific verticals. Hyperscalers (e.g., Google, Amazon, Microsoft, Meta) are already designing custom AI silicon for internal use (e.g., Google TPUs, AWS Trainium/Inferentia). It is highly plausible that they will either build their own neuromorphic capabilities or acquire promising startups to integrate highly efficient inference engines into their cloud offerings, directly impacting their operational costs and competitive pricing. This disruption will create new market leaders specializing in inference, while potentially marginalizing existing vendors who cannot adapt to the new energy efficiency paradigm. The shift from "peak performance" to "performance per watt" fundamentally alters the competitive dynamics.
Market Projections: The global market for low-power AI semiconductors, which includes neuromorphic chips, is projected to undergo substantial expansion. According to various 2025-2026 analyses, this growth is being driven by the relentless demand from edge AI, robotics, industrial IoT, and the increasing need for sustainable data center operations. Some estimates project a CAGR exceeding 25-30% for specific segments of this market through 2036. This growth is not merely additive but represents a re-segmentation of the AI chip market, where neuromorphic processors will carve out significant shares for inference, complementing GPUs for training. The battleground is no longer just about performance benchmarks, but about economic viability and environmental sustainability. Enterprises and cloud providers stand to gain trillions of dollars in TCO savings over the next decade through reduced electricity bills, decreased cooling requirements, and longer hardware lifecycles made possible by power-efficient designs.
Geopolitical & Regulatory Deep-Dive
The race for neuromorphic dominance is not merely technological or economic; it is deeply embedded within a complex geopolitical and regulatory landscape. The power efficiency and inherent capabilities of neuromorphic computing have significant strategic implications for national security, economic competitiveness, and technological sovereignty.US Policy, EU Regulations, China Strategy:
- United States: US policy encourages innovation in advanced computing, including AI hardware. Funding agencies like DARPA (Defense Advanced Research Projects Agency) and NIST (National Institute of Standards and Technology) have long supported neuromorphic research, recognizing its potential for defense applications (e.g., autonomous systems, real-time intelligence analysis at the edge) and economic leadership. The CHIPS and Science Act, enacted in 2022, allocates billions for domestic semiconductor manufacturing and R&D, a portion of which indirectly benefits advanced AI silicon development. The US aims to maintain its technological lead, fostering an ecosystem where companies like Intel and IBM can innovate. Current regulatory discussions focus on AI safety and ethics, but directly regulating neuromorphic hardware through specific legislation is less common, with the preference being for broad support of R&D and strategic investment.
- European Union: The EUâs regulatory approach is characterized by a strong emphasis on data privacy and ethical AI, exemplified by the upcoming AI Act. While not directly regulating neuromorphic hardware, the Actâs emphasis on transparency, robustness, and energy efficiency for AI systems indirectly favors such technologies. The EU is heavily investing in digital transformation and advanced computing, including initiatives like the European Processor Initiative and various Horizon Europe projects that fund neuromorphic research (e.g., BrainScaleS, SpiNNaker). The goal is to reduce reliance on non-EU tech giants and build sovereign AI capabilities, with energy efficiency as a key driver for sustainability goals.
- China: China views AI and advanced semiconductors as critical strategic areas for national rejuvenation and global leadership. Its "Made in China 2025" initiative and subsequent plans heavily prioritize domestic chip development, including neuromorphic and other AI-specific accelerators. Institutes like Tsinghua University and companies like Baidu and Alibaba are actively researching and developing their own neuromorphic solutions. The Chinese government provides massive state subsidies and incentives for R&D and manufacturing. Their strategy is dual-pronged: achieve technological self-sufficiency and establish global standards for next-generation computing. The low-power nature of neuromorphic chips aligns perfectly with China's push for energy efficiency and sustainable development, especially given its vast data center expansion.
- Technological Sovereignty: Both nations seek to control the entire AI value chain, from chip design to manufacturing and application. The ability to design and produce cutting-edge neuromorphic chips reduces reliance on foreign supply chains, a critical national security concern. The US export controls on advanced AI chips to China, while currently focused on GPUs and high-end logic, could easily extend to advanced neuromorphic processors if they become strategically significant.
- Military and Intelligence Applications: The real-time, low-power capabilities of neuromorphic chips are highly attractive for military applications, including autonomous drones, smart sensors in contested environments, and secure edge AI processing for intelligence gathering, where traditional power-hungry systems are impractical. This makes neuromorphic R&D a matter of national security for both powers.
- Economic Leadership: Dominance in next-generation AI hardware translates into significant economic advantages, securing future high-tech jobs, driving innovation, and capturing market share in nascent but lucrative industries like robotics and advanced IoT.
- 2023-2025: Focus on high-level AI ethics, safety, and data privacy (e.g., EU AI Act implementation). Export controls on advanced semiconductors become tighter, primarily impacting GPUs. Neuromorphic chips are largely in the research or early commercialization phase, operating under existing general hardware regulations.
- 2026-2028: As neuromorphic chips scale and become integrated into critical infrastructure (e.g., autonomous systems, smart cities), regulatory scrutiny will increase. Discussions around hardware-level security, explainability, and potential biases embedded in specialized SNN architectures may emerge. Governments may consider incentives for domestic production or R&D in energy-efficient AI hardware to meet climate goals or tech sovereignty objectives.
- Beyond 2028: If neuromorphic AI becomes pervasive, especially in high-stakes applications, specific industry standards and regulatory frameworks tailored to their unique characteristics (e.g., event-driven processing, robustness against adversarial attacks in SNNs) could be developed. The geopolitical "chip wars" will likely encompass neuromorphic capabilities, leading to strategic alliances or further restrictions on technology transfer. The environmental impact of data centers and pressure to reduce carbon emissions will continue to push for policies favoring energy-efficient hardware, directly benefiting neuromorphic adoption.
Fullmoonfever
Top 20
Part 3
Future Forecasting & Strategic Implications
Near-Term Horizon (6-12 months): Immediate Catalysts
The next 6-12 months will be critical for solidifying the commercial viability and initial market penetration of neuromorphic technology, moving beyond proof-of-concept to early production deployments.Events to Watch:
- Key Commercial Announcements: Anticipate major product announcements from BrainChip and other startups, revealing specific customer wins and larger-scale deployments of their neuromorphic chips, particularly in industrial IoT, automotive (ADAS pre-processing), and smart city infrastructure. These announcements will move discussions from theoretical efficiency to tangible ROI.
- Intel's Loihi Acceleration: Watch for Intel to announce further industrial partnerships and potentially a more aggressive commercialization roadmap for Loihi 2 or its derivatives. This could include showcasing benchmark results from complex real-world inference tasks that definitively demonstrate 50x-100x energy efficiency gains over leading GPUs for specific workloads, compelling enterprises to consider architectural shifts.
- Cloud Provider Pilots: Hyperscalers like AWS, Google Cloud, and Azure are likely to conduct advanced pilot programs, if not outright announcements, regarding the integration of specialized inference accelerators, potentially including neuromorphic or neuromorphic-inspired architectures. Their internal findings on TCO reduction and energy savings will be highly influential.
- SNN Software Maturity: Observe the progress in SNN conversion tools and frameworks (e.g., from PyTorch/TensorFlow to SNN platforms). Improvements in ease of development, model conversion accuracy, and availability of pre-trained SNN models will be a significant catalyst for broader adoption. A 10x improvement in developer productivity for SNNs within this timeframe could unlock substantial enterprise interest.
- Benchmarking Standardization: The industry will need standardized, peer-reviewed benchmarks specifically designed for event-driven, sparse AI, moving beyond traditional FLOPS and establishing "inference per watt" as a primary metric. The emergence of such widely accepted benchmarks will provide clarity and accelerate enterprise decision-making. Initiatives like MLPerf Inference could introduce new tracks for SNNs.
- Small-Scale Enterprise Deployments: Expect initial deployments in specific, high-value, power-constrained applications such as remote sensor networks for critical infrastructure monitoring (e.g., pipelines, bridges) or intelligent surveillance systems where data must be processed at the source with minimal power draw.
- Hybrid AI Architectures: We will likely see more public examples of hybrid AI deployments where neuromorphic chips handle pre-processing (e.g., anomaly detection, feature extraction) and feed compressed, critical data to traditional GPUs for more complex, larger-model inference. This allows enterprises to incrementally adopt neuromorphic technology without a complete architectural overhaul.
- Increased R&D Spending: A noticeable uptick in R&D spending by large enterprises in sectors like manufacturing, aerospace, and healthcare, specifically earmarked for evaluating or developing applications on neuromorphic platforms.
- Cost Leadership: Early adopters can realize substantial OpEx savings on their AI inference infrastructure, translating into lower production costs for AI-powered services or products. A 20-30% reduction in AI inference-related energy costs within the first year would be a major strategic play.
- Performance Edge: For real-time applications (e.g., autonomous driving, high-frequency trading, real-time fraud detection), the ultra-low latency and event-driven processing of neuromorphic chips will deliver a decisive performance edge over competitors relying solely on GPUs. Sub-millisecond latency improvements are critical in these domains.
- Sustainability Credentials: Demonstrating leadership in energy-efficient AI aligns with increasing ESG pressures and can enhance corporate reputation, attracting environmentally conscious customers and investors. Reducing the carbon footprint of AI operations by 15-20% through neuromorphic adoption could be a public relations win.
- Talent Acquisition: Early engagement allows companies to build in-house expertise in a cutting-edge field, attracting top AI and hardware engineering talent. This specialized knowledge will be a crucial differentiator. Strategic plays include designating internal AI innovation labs to specifically explore neuromorphic use cases, forming joint ventures with neuromorphic startups, or making strategic investments in the intellectual property of these nascent companies. Missing this window could mean facing higher operating costs and slower adoption rates later on.
Mid-Term Horizon (2-3 years): Industry Restructuring
Over the next 2-3 years, neuromorphic computing, alongside other specialized AI accelerators, will drive a significant restructuring of the AI hardware industry and the broader data center ecosystem. This period will witness a more pronounced shift from exclusive GPU reliance to a diversified, purpose-built AI infrastructure.Displaced Industries, New Giants:
- Traditional Data Center Infrastructure: The primary displacement will occur in data center power and cooling systems. As inference workloads shift to ultra-efficient neuromorphic platforms, the demand for high-power racks, advanced liquid cooling solutions, and massive power delivery units will decrease for specific AI inference clusters. This might lead to diminished revenue growth for some traditional data center component manufacturers, while those specializing in low-power, high-density edge deployments will see a boost.
- Niche AI Accelerator Vendors: Companies that specialize in highly optimized but still GPU-like architectures for inference, without embracing the fundamental energy efficiency principles of neuromorphic design or in-memory computing, might find their market share eroded.
- New Giants: Expect new specialized AI hardware companies, particularly those focused on compute-in-memory and neuromorphic architectures, to rise in prominence. BrainChip, with its commercial momentum, could solidify its position. Others, currently in stealth or early funding rounds, may emerge as multi-billion dollar entities. Established players like Intel and IBM, if they successfully productize their research, will leverage their scale to capture significant market segments.
- Silicon Designers and Manufacturers: There will be a significant shift in demand towards silicon designers proficient in asynchronous circuit design, SNN architectures, and hybrid chip integration. Foundries capable of producing these specialized, often lower-power chips with integrated memory will gain strategic importance.
- AI Software Engineers: The skill set for AI software engineers will broaden. While deep learning frameworks will remain essential, there will be increasing demand for expertise in Spiking Neural Networks, neuromorphic programming models (e.g., Lava), optimization for sparse data processing, and bridging the gap between traditional DNNs and SNNs. Universities and corporate training programs will adapt to this new curriculum.
- System Integrators: System integrators and cloud solution architects will need to develop expertise in designing heterogeneous AI infrastructure that optimally combines GPUs (for training/generalized inference) with neuromorphic accelerators (for power-efficient, real-time inference). This complex orchestration will create new service opportunities.
- Edge Device Manufacturers: Companies producing IoT devices, industrial automation systems, and robotics will increasingly integrate neuromorphic processors, pushing sophisticated AI capabilities further to the edge without compromising battery life or form factor.
- NVIDIA: While still dominating training, NVIDIA will intensify its efforts in inference efficiency, potentially through acquisitions, further specialization of Tensor Cores, or even incorporating neuromorphic-like principles into its future architectures. Their revenue stream for inference hardware will face pressure from specialized alternatives.
- Intel/IBM: If successfully commercialized, Intel's Loihi and IBM's NorthPole could see significant revenue inflection from specific enterprise and government contracts where their unique capabilities are mission-critical. This would provide a vital new growth engine beyond their traditional CPU businesses.
- BrainChip/Mythic/EnCharge AI: These specialized players will likely see substantial revenue growth as enterprises move beyond pilots to production deployment, showcasing tangible ROI from reduced energy bills. Their valuations will reflect this accelerating adoption curve.
- Cloud Providers: Hyperscalers embedding neuromorphic or similar specialist silicon will lower their internal operating costs, allowing them to offer more competitive pricing for inference services or increase their profit margins. This can create a significant competitive differentiator in the cloud AI market. The "watts per query" metric will become a key competitive battleground.
- Manufacturing and Vision Systems: Industries leveraging real-time vision processing for quality control, drone inspection, or inventory management will see improved efficiency and cost savings with neuromorphic integration.
Long-Term Vision (5 years): Civilizational Impact
Looking 5 years out, the widespread adoption of neuromorphic computing and energy-efficient AI inference will transcend technological and economic shifts, manifesting as a profound civilizational impact, restructuring global economic patterns, geopolitical dynamics, and fundamental human capabilities.Societal Transformation, Economic Structure:
- Ubiquitous, Invisible AI: The ultra-low power consumption of neuromorphic chips will enable AI to be truly ubiquitous, embedded into nearly every aspect of daily life without requiring constant charging or bulky infrastructure. Smart environments (homes, cities, workplaces) will become truly intelligent, with real-time, personalized, and proactive assistance seamlessly integrated. This passive yet powerful AI presence will be a fundamental shift from today's active engagement with AI through screens.
- Sustainable AI Economy: The drastic reduction in data center power consumption for inference will make AI a more environmentally sustainable technology. This will alleviate pressure on energy grids, reduce carbon footprints significantly, and enable the global expansion of AI services into regions with limited energy infrastructure. The economic structure will shift towards valuing "sustainable compute" alongside raw performance, driving innovation in renewable energy integration for smaller, distributed AI inference hubs.
- Decentralized Intelligence: The ability to perform complex AI inference at the extreme edge, on small, battery-powered devices, will lead to a more decentralized intelligence paradigm. This will reduce reliance on centralized cloud services for many tasks, improving data privacy and security. Localized AI will make devices more robust and less dependent on network connectivity, fostering innovation in areas like remote healthcare, precision agriculture, and disaster response.
- Productivity Revolution: Industries suffering from data latency or power limitations will experience a new wave of productivity. Robotics and automation will become more agile and intelligent, able to react to real-world events with human-like speed and efficiency. Manufacturing, logistics, and resource management will be optimized by real-time AI at unprecedented levels, leading to further supply chain efficiencies and cost reductions, contributing to deflationary pressures in goods and services.
- Altered Geopolitical Power Dynamics: Nations that lead in neuromorphic R&D, manufacturing, and deployment will gain a substantial geopolitical advantage. This extends beyond economic leadership to military and intelligence superiority, as low-power, resilient AI systems become critical for defense and cyber warfare. The energy efficiency aspect also plays into energy independence and resilience, further strengthening nations that embrace this technology. The current semiconductor "blocs" (US-led vs. China-led) will solidify around this new generation of AI hardware, potentially leading to even greater technological divergence.
- Enhanced Human Capabilities: Neuromorphic systems, particularly their event-driven, low-latency nature, hold immense potential for human-machine interfaces, neuroprosthetics, and even brain-computer interfaces (BCIs). The ability to process biological signals in real-time with minimal power could lead to breakthrough medical devices that restore sensory functions, control exoskeletons, or augment cognitive abilities, blurring the lines between human and artificial intelligence in ethically profound ways.
- Redefinition of "Intelligence": As hardware moves closer to mimicking biological brains, it will deepen our understanding of intelligence itself. The development of more brain-like AI will stimulate philosophical and scientific inquiry into consciousness, learning, and sentience, potentially altering collective human self-perception.
- AI Ethics and Governance: The ubiquity of stealthy, low-power AI will necessitate even more robust ethical frameworks and governance mechanisms. The challenge of identifying and auditing AI decisions made by systems embedded deep within the environment, operating on event-driven principles, will be substantial. International cooperation on AI safety and interpretability will become even more critical as AI becomes an invisible but ever-present force in society.
Executive Conclusion & Strategic Takeaways
Bottom Line Assessment: Neuromorphic chips represent a credible, disruptive force in the AI hardware landscape, particularly for inference workloads. The demonstrated 10-100x energy efficiency gains over GPUs are not theoretical but are being realized in early prototypes and commercial offerings (Intel Loihi 2, BrainChip Akida). While GPUs will retain dominance for large-scale model training and generalized inference for the foreseeable future, neuromorphic architectures are poised to carve out a significant and strategically vital market segment in energy-constrained, real-time, and edge AI applications. My confidence level in this shift is very high (9/10) over a 2-3 year horizon for specific use cases, and moderately high (7/10) for broader data center inference beyond 5 years as software matures. The economic imperative to reduce data center TCO and meet sustainability goals makes the adoption of such ultra-efficient hardware inevitable.Key Insights Summary:
- Inference Economics Redefined: The primary battleground for AI hardware is shifting from raw FLOPS to "watts per inference," significantly altering data center OpEx and TCO. Neuromorphic chips are leading this shift.
- Strategic Specialization: AI hardware will become increasingly specialized. Neuromorphic chips will not universally replace GPUs, but rather complement them within heterogeneous AI infrastructures, optimized for specific, energy-sensitive inference tasks.
- Geopolitical Race for Efficiency: Dominance in energy-efficient AI hardware, including neuromorphic systems, is a critical component of national technological sovereignty and military advantage, intensifying US-China competition.
- Talent and Ecosystem Maturity: The nascent neuromorphic ecosystem, including software frameworks and developer tools, is rapidly maturing. Investing in SNN expertise and fostering specialized talent is a strategic imperative for enterprises and nations.
- Beyond Cost Savings to New Capabilities: Beyond direct cost reductions, neuromorphic chips unlock new real-time, autonomous, and ubiquitous AI capabilities that are currently impossible with power-hungry traditional architectures.
- M&A and Investment Opportunities: Expect accelerated M&A activity and significant VC investment in neuromorphic startups and related compute-in-memory technologies as larger players seek to acquire capabilities.
- ESG as a Driving Force: Environmental, Social, and Governance (ESG) pressures, particularly regarding data center carbon emissions, will further accelerate the demand for ultra-efficient AI hardware, making neuromorphic solutions highly attractive.
MetaGuard AI Cracks Open Cybersecurity's Black Box at CES 2026 - BriefGlance.com
Debuting at CES, MetaGuard AI's new platform offers an industry-first: full source code access, blending defense-grade AI with neuromorphic tech.
MetaGuard AI Cracks Open Cybersecurity's Black Box at CES 2026
Key Data- $1.65 million: Federal funding from the U.S. Department of Energy for CyberNeuroRT's development.
- 100 Enterprise Ultra customers: MetaGuard AI's 'Transparency Pioneer Program' offers discounts to early adopters.
- Neuromorphic processors: CyberNeuroRT runs on BrainChip's Akida processors, enabling ultra-low-power edge computing.
Expert ConsensusExperts view MetaGuard AI's CyberNeuroRT as a groundbreaking shift toward transparency and accountability in cybersecurity, offering verifiable, auditable AI solutions that could redefine enterprise trust in digital defenses."
Just a reminder that CyberNeuro-RT runs on the Akida Edge Box, which, in its current configuration, contains 2 Akida 1000 chips as co-processors with a NXP processor. Obviously, later versions may switch to Akida 1500, Akida 2, etc."MetaGuard AI Cracks Open Cybersecurity's Black Box at CES 2026 - BriefGlance.com
Debuting at CES, MetaGuard AI's new platform offers an industry-first: full source code access, blending defense-grade AI with neuromorphic tech.briefglance.com
MetaGuard AI Cracks Open Cybersecurity's Black Box at CES 2026
- $1.65 million: Federal funding from the U.S. Department of Energy for CyberNeuroRT's development.
- 100 Enterprise Ultra customers: MetaGuard AI's 'Transparency Pioneer Program' offers discounts to early adopters.
- Neuromorphic processors: CyberNeuroRT runs on BrainChip's Akida processors, enabling ultra-low-power edge computing.
Experts view MetaGuard AI's CyberNeuroRT as a groundbreaking shift toward transparency and accountability in cybersecurity, offering verifiable, auditable AI solutions that could redefine enterprise trust in digital defenses."
https://shop.brainchipinc.com/products/akidaâą-edge-ai-box
- Host CPU: NXP i.MX 8M Plus Quad SOC
- AI/ML Accelerator: 2 x AKD1000 (BrainChip Akida Chip) Over PCIe
- On Board Memory: 4GB LPDDR4, 32GB eMMC
- Supports Linux Embedded OS Version 6.1
- External Storage: Micro SD Card Connector that supports Micro SD
- Display: 1 x HDMI Output with Max Resolutions of 3840 x 2160p30 Pixel clock up-to 297MHZ
- Network Connectivity:
- Wi-Fi Connectivity: 802.11 ac/b/g/n/c (2.4GHz /5GHz)
- Ethernet Connectivity: Two Ethernet Port with10/100/1000mbps
- Ethernet Port supports external camera interface
2 Akida chips is a very powerful chunk of AI, and it will only get better.
I wonder if, with Akida doing the heavy lifting, it can just be cooled by heat sink without a fan?
Last edited:
Are we getting traction? Are we getting meaningful engagements? Are we being acknowledged as a real serious neuromorphic player at the edge? Are we being acknowledged by some big players in a variety of edge AI developing markets? Are we expanding our edge AI footprint? Are we continually improving our Akida technology? Are we currently funded to move forward with our commercial development? Are we about to tape out our Akida 1500 with customers already placing buy orders? Are we moving towards the 2nd and 3rd generations of Akida, with clients engaged? Are we leading this space at the far edge? Are we going to succeed?
If you answered yes, well, the company certainly appreciates your support.
This Australian company started by Peter Van der Made will succeed, why, because Aussies always punch above their weight on the big stage, watch out Jensen, a change is coming...go Brainchip, we love u.
Tech 
If you answered yes, well, the company certainly appreciates your support.
This Australian company started by Peter Van der Made will succeed, why, because Aussies always punch above their weight on the big stage, watch out Jensen, a change is coming...go Brainchip, we love u.
Tech 
Frangipani
Top 20
Adam Taylor from Adiuvo Engineering, who wrote very favourably about benchmarking Akida on Edge Impulse just over two years ago (https://www.edgeimpulse.com/blog/brainchip-akida-and-edge-impulse/), posted about ORA today, âa comprehensive sandbox platform for edge AI deployment and benchmarkingâ:
ORA: A Sandbox Platform for Edge AI Deployment | Adam Taylor posted on the topic | LinkedIn
Something else we have been working on, selecting the best edge AI deployment for your use case. ORA is a comprehensive sandbox platform for edge AI deployment and benchmarking. Our catalogue of edge devices lets you test and evaluate which hardware best suits your application needs. With...
View attachment 90947
View attachment 90948
View attachment 90949 View attachment 90950 View attachment 90951 View attachment 90952
LinkedIn Login, Sign in | LinkedIn
Login to LinkedIn to keep in touch with people you know, share ideas, and build your career.www.linkedin.com
View attachment 90953
In September, I shared a LinkedIn post
by Adam Taylor, Founder and Lead Consultant of Adiuvo Engineering as well as Owner & Organiser of FPGA Horizons, about ORA:âORA is a comprehensive sandbox platform for edge AI deployment and benchmarking. Our catalogue of edge devices lets you test and evaluate which hardware best suits your application needs. With devices spanning Nvidia, Axelera, Hailo, and BrainChip, ORA offers diverse computing architectures. These range from traditional GPUs to cutting-edge neuromorphic processors and specialised AI accelerators. All accessible via a single interface.â
Turns out Adam Taylor recently co-founded Dublin-based space tech startup Setanta Space Systems (of which he is CEO), with fellow co-founders James Murphy (CTO) and Jake OâBrien (Lead Engineer), both of whom used to work for another space tech company based in the Irish capital: RĂ©altra Space Systems Engineering, which âis dedicated to the design, development and manufacture of cost-effective space electronic systems using cutting-edge technologies.â (https://www.linkedin.com/company/realtra/about/)
Setanta Space is âdeveloping edge AI for space applicationsâ:
Something I have been working on with James Murphy and Jake O'Brien for a little while. Setanta Space developing edge AI for space applications. The first products are already in development and⊠| Adam Taylor
Something I have been working on with James Murphy and Jake O'Brien for a little while. Setanta Space developing edge AI for space applications. The first products are already in development and already getting significant interest. Stay Tuned!
Setanta Space
www.setantaspace.com
âAI-Driven Autonomy for the Next Generation of Spacecraft
Setanta Space Systems is an Irish space technology company developing advanced solutions for satellite automation and space electronics. Our expertise spans both software and hardware, enabling us to deliver integrated systems for next-generation missions.
We design AI models for on-board anomaly detection, predictive maintenance, autonomous navigation, space situational awareness and real-time data processing. We also build space-qualified electronics, AI-ready processing platforms, and bespoke systems tailored to mission requirements, from low Earth orbit to deep space.â
Services currently offered comprise Telemetry Monitoring, Space Situational Awareness (SSA) and Autonomous Navigation as well as Earth Observation (EO).
And the above-mentioned Ora sandbox is now listed as one of Setanta Space Systemâs products, alongside Danu, their modular onboard computer, as well as their Galaxia Space Development Board and their Galaxia Space Tile:
Note that Jake OâBrien is concurrently pursuing an âIndustry-based PhDâ in Computer Science with University College Dublin - the industry placement being Adam Taylorâs company Adiuvo Engineering.
In his LinkedIn profile, Jake OâBrien refers to his PhD research as follows:âSpiking Neural Networks for Space Situational Awareness and Autonomous Navigation: Leveraging Neuromorphic Hardware and Dynamic Vision Sensorsâ.
Now add to all this the fact that Alf Kuchenbuch loved Jake OâBrienâs December launch announcement of Setanta Space - the two of them have exchanged several LinkedIn likes in recent months, by the way.
Dot...dotâŠdotâŠ
I wouldnât be in the least surprised if Akida played a part both in Jake OâBrienâs PhD research and in at least one of the edge AI solutions that Setanta Space are currently developing.
Hopefully we will find out more before or on Saint Patrickâs Day.
Thrilled to share the exciting news that together with Adam Taylor and James Murphy, we have launched Setanta Space. Looking forward to taking on the new year and cool projects with this amazing⊠| Jake O'Brien
Thrilled to share the exciting news that together with Adam Taylor and James Murphy, we have launched Setanta Space. Looking forward to taking on the new year and cool projects with this amazing team. Ar aghaidh linn! đźđȘ
And hereâs the third co-founderâs LinkedIn post announcing the birth of their space tech baby last month:
I am excited to share that I have taken a huge step and founded Setanta Space with Adam Taylor and Jake O'Brien. Starting a company is a risk, especially in a sector like space, but it is one I have⊠| James Murphy
I am excited to share that I have taken a huge step and founded Setanta Space with Adam Taylor and Jake O'Brien. Starting a company is a risk, especially in a sector like space, but it is one I have thought about carefully. I believe strongly that this is the right time, and that the...
Three co-founders of an Irish company - that almost demanded the use of a green triquetra (âIrish trinity knotâcf. https://en.wikipedia.org/wiki/Triquetra) as their logoâŠ
I presume Setanta Space Systems was named after SĂ©tanta aka CĂș Chulainn, an Irish warrior hero and demigod in Celtic mythology?
CĂș Chulainn - Wikipedia
Attachments
Similar threads
