What's new in neuromorphic computing?

This blog post has been written by the person who has mapped the neuromorphic computing market in a clean and beautiful presentation

Neuromorphic computing has reached a critical inflection point in 2025, moving from research labs to commercial products with billion-dollar investments and real-world deployments.

Major players like Intel, BrainChip, and Innatera are shipping actual hardware while startups raised over $200M in series A/B rounds, signaling genuine market momentum beyond academic curiosity.

And if you need to understand this market in 30 minutes with the latest information, you can download our quick market pitch.

What are the most significant breakthroughs or commercial product launches in neuromorphic computing since January 2025?

Intel's Hala Point system represents the largest neuromorphic deployment to date, featuring 1.15 billion neurons and achieving over 15 TOPS/W efficiency—more than 12 times better than conventional GPU/CPU systems.

BrainChip launched the Akida Pulsar, the world's first mass-market neuromorphic microcontroller specifically designed for sensor edge applications. This chip delivers 500× lower energy consumption and 100× latency reduction compared to conventional AI cores, marking a shift from research prototypes to production-ready hardware.

Innatera unveiled their SNP (Spiking Neural Processor) at CES 2025, targeting ambient intelligence applications with sub-milliwatt power dissipation and sub-millisecond latency for presence and activity detection. The company also announced partnerships with major automotive suppliers for autonomous vehicle sensor fusion.

Intel's Loihi 2 benchmarks revealed dramatic performance advantages: 75× lower latency and 1,000× higher energy efficiency versus NVIDIA Jetson Orin Nano on state-space model workloads. These aren't marginal improvements but fundamental efficiency gains that justify commercial adoption in power-constrained environments.

Looking for the latest market trends? We break them down in sharp, digestible presentations you can skim or share.

Which startups, companies, or research labs are leading the development of neuromorphic hardware and software as of mid-2025?

Intel maintains the research leadership position with Loihi 2 deployments at Sandia Labs and partnerships spanning from Microsoft Research to Thales, but BrainChip leads in commercial market penetration with actual shipping products and established customer relationships.

BrainChip secured $35 million in Series B funding and expanded their ecosystem through partnerships with Andes Technology for RISC-V integration and Edge Impulse for ML toolchain integration. Their Akida processor now ships in M.2 form factors and embedded modules, moving beyond development boards to production systems.

Innatera raised $20 million and demonstrated their SNP processor's capabilities in real-world ambient intelligence scenarios. Their approach focuses on ultra-low-power applications where conventional processors simply cannot operate within power budgets, creating clear market differentiation.

European players are gaining momentum through the UK Multidisciplinary Centre for Neuromorphic Computing (£5.6 million UKRI funding), involving Queen Mary University, University of Strathclyde, and industry partners including HP Labs and Thales. This represents coordinated ecosystem development rather than isolated research efforts.

Acquisitions signal market maturation: Microsemi acquired GrAI Matter Labs IP to integrate low-power edge AI capabilities, while corporate venture arms from Samsung NEXT, Qualcomm Ventures, and Intel Capital actively invest in neuromorphic startups.

If you want fresh and clear data on this market, you can download our latest market pitch deck here

What specific use cases or industry verticals are showing the most traction for neuromorphic computing in 2025, and what's expected in 2026?

Cybersecurity applications demonstrate immediate commercial viability, with BrainChip's Akida processor showing superior performance in intrusion detection systems while consuming only 1 watt versus 2.5 watts for Loihi 2 in comparable workloads.

Event-based vision systems represent the fastest-growing application area, with Prophesee partnering with BrainChip to demonstrate gesture recognition at Embedded World 2025. These systems excel in scenarios requiring real-time response to motion or changes, such as automotive safety systems and industrial monitoring.

Autonomous robotics deployments are moving from pilots to production, with BMW Research implementing Loihi 2 clusters for real-time traffic sign recognition and Lockheed Martin testing Innatera SNP processors for autonomous drone navigation. The key advantage lies in real-time processing with minimal power consumption—critical for battery-operated autonomous systems.

Medical wearable devices present a massive opportunity for 2026, with several companies pursuing FDA approvals for neuromorphic-powered seizure prediction and continuous health monitoring devices. The ultra-low power consumption enables always-on monitoring without frequent battery replacement.

IoT sensor networks show promise for 2026 mass deployment, where neuromorphic processors can perform local inference on sensor data while consuming minimal power. Industry forecasts suggest neuromorphic chips will be present in 40% of IoT sensor nodes by 2030, driven by edge processing requirements and power efficiency demands.

How have investment trends changed in 2025 for neuromorphic startups, and what funding rounds or acquisitions signal real market interest?

Venture funding in neuromorphic computing exceeded $200 million in Series A and B rounds during 2025, representing a 3× increase from 2024 levels and indicating investor confidence in commercial viability rather than speculative technology betting.

What are the latest benchmarks or performance comparisons between neuromorphic chips (like Loihi 2, Akida, etc.) and traditional GPUs or TPUs?

Intel's Loihi 2 benchmarks on state-space models revealed 75× lower latency and 1,000× higher energy efficiency compared to NVIDIA Jetson Orin Nano, representing the most comprehensive neuromorphic vs. conventional AI comparison to date.

BrainChip's Akida processor demonstrated superior performance in cybersecurity workloads, consuming only 1 watt versus 2.5 watts for Intel's Loihi 2 on comparable intrusion detection tasks while maintaining higher streaming throughput for real-time security monitoring.

The Hala Point system (1,152 Loihi 2 chips) achieved over 15 TOPS/W on AI inference tasks—more than 12× better efficiency than conventional GPU/CPU systems. This represents 20 peta-operations per second throughput with dramatically lower power consumption than equivalent traditional architectures.

Sparse coding benchmarks show Loihi 2 delivering 10×-50× energy savings versus CPU/GPU implementations, particularly valuable for applications requiring continuous operation on battery power. These efficiency gains enable new application categories impossible with conventional processors.

However, neuromorphic chips currently excel in specific workloads (sparse inference, event-driven processing, always-on monitoring) rather than general-purpose computing. Traditional GPUs maintain advantages in dense matrix operations and well-established AI training workflows.

What are the major technical challenges still holding back broader adoption of neuromorphic computing, and who is working on solving them?

Standardization remains the biggest barrier, with no unified benchmarking methodology to compare neuromorphic systems fairly against conventional hardware or even against each other across different architectures and applications.

The IEEE P2800 working group addresses neuromorphic interfaces and programming models, while Neurobench and SNABSuite emerge as standardized benchmarking frameworks covering low-level hardware metrics to high-level application performance. These efforts aim to create industry-wide comparison standards by late 2025.

Programming paradigm gaps persist because neuromorphic systems require fundamentally different approaches than traditional von Neumann architectures. Current solutions include Intel's Lava framework for Loihi 2, Nengo for cross-platform development, and PyTorch-based frameworks like SpikingJelly and snnTorch for GPU-accelerated development.

Device variability and manufacturing yield issues affect analog and mixed-signal neuromorphic implementations. CEA-Leti addresses these through novel materials research, while companies like BrainChip focus on digital neuromorphic architectures that eliminate analog variability concerns entirely.

Wondering who's shaping this fast-moving industry? Our slides map out the top players and challengers in seconds.

Integration challenges with existing systems require hybrid architectures combining neuromorphic processors with conventional CPUs and GPUs. Intel and BrainChip develop reference designs showing how neuromorphic chips integrate into standard computing environments without requiring complete system redesigns.

If you need to-the-point data on this market, you can download our latest market pitch deck here

What level of power savings, latency reduction, or efficiency gains are companies reporting from neuromorphic deployments in edge and IoT devices?

Innatera's SNP processor achieves sub-milliwatt power dissipation with sub-millisecond latency for presence and activity detection, enabling always-on monitoring in battery-powered devices where conventional processors would drain batteries within hours.

BrainChip's Akida Pulsar delivers 500× lower energy consumption and 100× latency reduction versus conventional AI cores in edge inference tasks. These improvements enable real-time AI processing in applications previously impossible due to power and thermal constraints.

Intel's Loihi 2 implementations show 10×-50× energy savings in sparse coding applications compared to CPU/GPU alternatives, particularly valuable for small autonomous robots requiring continuous operation without frequent battery replacement or external power sources.

Real-world deployments report dramatic efficiency gains: BMW's traffic sign recognition system using Loihi 2 clusters operates continuously in vehicles without impacting battery life, while Lockheed Martin's drone navigation systems achieve autonomous operation for extended missions impossible with conventional processors.

Medical wearable prototypes demonstrate continuous health monitoring with monthly rather than daily battery replacement, enabling practical always-on health tracking for chronic disease management and emergency detection applications.

How are governments or regulatory bodies supporting or influencing the neuromorphic computing ecosystem globally in 2025?

The UK leads government support through the UKRI EPSRC £5.6 million award to establish the UK Multidisciplinary Centre for Neuromorphic Computing, bringing together Queen Mary University, University of Strathclyde, and major industry partners for coordinated ecosystem development.

European Union Horizon Europe programs allocated €50 million for brain-inspired hardware and software development, focusing on sustainable neuromorphic ecosystems and integration with existing industrial processes rather than pure research initiatives.

US DARPA maintains funding for neuromorphic sensing and defense applications through the EdgeSense program, emphasizing real-time threat detection and autonomous system capabilities that leverage neuromorphic advantages in power-constrained military environments.

China's government backing supports companies like Aizip through strategic investment funds, focusing on neuromorphic accelerators for edge AI inference to maintain competitiveness in AI hardware development and reduce dependence on foreign GPU architectures.

Standardization efforts include IEEE P2800 working group development of neuromorphic interfaces and programming standards, providing regulatory framework for commercial deployment and interoperability between different neuromorphic architectures and conventional systems.

What open-source frameworks, developer tools, or APIs have emerged in 2025 to make neuromorphic computing more accessible to engineers and startups?

• Lava Framework (Intel): Provides comprehensive development environment for Loihi 2 systems, including high-level application development tools and automatic mapping to neuromorphic hardware architectures
• Nengo Platform: Cross-platform SNN modeling framework supporting multiple neuromorphic backends, enabling developers to write code once and deploy across different hardware platforms
• SpikingJelly: PyTorch-based framework allowing developers to leverage existing deep learning expertise while transitioning to spiking neural networks and neuromorphic deployment
• snnTorch: GPU-accelerated training framework for spiking neural networks, bridging conventional deep learning workflows with neuromorphic deployment requirements
• Open-Neuromorphic Organization: Curated repository of neuromorphic software tools, datasets, and educational resources, providing centralized access to community-developed tools and documentation

Edge Impulse integrated Akida support into their ML deployment toolchain, enabling engineers to train models using conventional tools and automatically optimize for neuromorphic deployment without requiring specialized neuromorphic programming expertise.

The emergence of these frameworks addresses the critical barrier of programming complexity, allowing engineers with conventional AI backgrounds to develop neuromorphic applications without learning entirely new programming paradigms from scratch.

If you want to build or invest on this market, you can download our latest market pitch deck here

What partnerships or pilot programs have been announced in 2025 between neuromorphic players and major corporations in automotive, robotics, or defense?

BMW Research implements Loihi 2 clusters for real-time traffic sign recognition systems, representing the first automotive OEM deployment of neuromorphic processors in production vehicle development programs rather than research experiments.

Prophesee partnered with BrainChip to demonstrate event-based vision systems at Embedded World 2025, showcasing gesture recognition capabilities for automotive human-machine interfaces and industrial safety monitoring applications.

Lockheed Martin tests Innatera SNP processors for autonomous drone navigation, focusing on power-efficient real-time obstacle avoidance and target tracking in GPS-denied environments where conventional processors cannot operate within power and thermal constraints.

BrainChip and Andes Technology announced integration of Akida neuromorphic cores with RISC-V processors on QiLai development boards, enabling hybrid conventional-neuromorphic computing architectures for automotive and industrial applications.

Edge Impulse integrated Akida support into their ML toolchain, allowing major industrial customers to deploy neuromorphic inference without specialized neuromorphic programming expertise, accelerating adoption across multiple verticals simultaneously.

These partnerships demonstrate movement from proof-of-concept demonstrations to actual commercial deployment programs with clear technical specifications and timeline commitments, indicating genuine market traction rather than marketing collaborations.

What are the key predictions for neuromorphic computing adoption between now and 2030, and what milestones need to happen in 2026 to stay on track?

Industry forecasts predict neuromorphic chips will be present in 40% of IoT sensor nodes and 15% of autonomous robots by 2030, driven by power efficiency requirements that conventional processors cannot meet in battery-operated edge applications.

Critical 2026 milestones include standardized neuromorphic benchmarks through IEEE P2800, mass production of neuromorphic microcontrollers at ≥100,000 units per year scale, and first medical device approvals for neuromorphic-powered seizure prediction and continuous health monitoring wearables.

Technical targets for 2030 include achieving ≥1 peta-ops/W efficiency in commercial neuromorphic SoCs, representing 100× improvement over current systems and enabling new application categories impossible with conventional architectures.

Market size predictions suggest neuromorphic computing will reach $45 billion by 2030, growing from approximately $1 billion in 2025, indicating 45× market expansion driven by edge AI deployment and power efficiency requirements.

Planning your next move in this new space? Start with a clean visual breakdown of market size, models, and momentum.

Critical success factors include maintaining development momentum in software frameworks, achieving cost parity with conventional processors in high-volume applications, and establishing clear regulatory pathways for safety-critical applications in automotive and medical verticals.

How can an investor or founder identify the most promising business models in neuromorphic computing—hardware IP, edge applications, SaaS platforms, or custom chips?

Hardware IP licensing represents the highest-margin opportunity, with companies like BrainChip licensing Akida cores for integration into RISC-V and other processor architectures, enabling SoC manufacturers to add neuromorphic capabilities without developing proprietary architectures.

Edge AI SaaS platforms show strong recurring revenue potential, particularly toolchains like Edge Impulse + Lava that offer hosted neuromorphic application development, enabling companies to deploy neuromorphic inference without internal neuromorphic expertise.

Custom chip fabrication targets high-value, low-volume applications in automotive and defense sectors where performance requirements justify premium pricing and customers accept longer development cycles for specialized capabilities.

Sensor+compute bundles present compelling integration opportunities, combining event cameras with neuromorphic processors for robotics applications where system-level optimization delivers superior performance compared to discrete components.

Conclusion

Sources

1. Intel Newsroom - Hala Point Neuromorphic System
2. OpenReview - Loihi 2 Benchmarks
3. Innatera - Pulsar Neuromorphic Microcontroller
4. Edge AI Vision - BrainChip M.2 Form Factor
5. BrainChip - CES 2025 Announcements
6. Electronic Specifier - Innatera CES 2025
7. University of Strathclyde - UK Neuromorphic Centre
8. BrainChip - Cybersecurity White Paper
9. RISC-V - BrainChip Andes Partnership
10. UKRI - Neuromorphic Computing Centre Funding
11. GitHub - Open Neuromorphic Organization
12. Open Neuromorphic - Software Resources
13. PNNL - Loihi 2 Benchmarking Study
14. BrainChip - Embedded World 2025 Demonstrations
15. Globe Newswire - Neuromorphic Market Growth Forecast
16. Precedence Research - Neuromorphic Computing Market Analysis