Slides
Page 1
Quarterly investor update Q2 2026 11 September 2026
Page 2
Agenda Edge AI – an early growth story Why BrainChip? Q2 highlights AKD1500 chips create a new revenue stream Akida powers hybrid computing Investor questions Appendices
Page 3
Edge AI – an early growth story Data centre cost, power and latency constraints increasing AI is moving from the cloud into physical devices (the Edge) 3 Quarterly Investor Update – Q2 2026 Edge AI adoption accelerating Privacy and data sovereignty requirements increasing Demand for AI growing exponentially Physical AI $4B – $26B (+34%p.a.) 3 Global edge AI market Total market change (2026 - 2031): $31B – $69B (+18%p.a.) 1 Defense AI $12B – $22B (+13%p.a.) 2 Wearable AI $55B – $190B (+28%p.a.) 2 Voice AI $2.3B – $3.5B (+17%p.a.) 2 1. MarketsandMarkets; 2. Grand View Research; 3. Acumen Research.
Page 4
Why BrainChip? Strategy built to win BrainChip is the worldwide leader in ultra low power edge AI 4 Quarterly Investor Update – Q2 2026 1 st to commercialise neuromorphic BrainChip in the market with a range of leading products/services enabling edge AI Protected IP advantage 67 patents either secured or under review Customer choice Modules, Chips, IP – solutions tailored to customer requirements Plug and Play Hardware, software and easy - to - run modules P ower , latency + privacy Advanced event - based processor responds quickly at ultra - low power Hybrid compute Even more efficient when tasks routed between CPU, GPU and Akida
Page 5
Why BrainChip? More to offer Quarterly Investor Update – Q2 2026 Akida IP Models Edge AI hardware Developer tools & cloud Reference platforms AKD1500 M.2 PCIe RPi 5 AKD1000 M.2 PCIe RPi 5 More intelligence, products and services than ever – driving greater customer interaction Reference designs Radar, Electronic Warfare and Wearable/Akida Tag Akida 1 and 2 IP Enablement services Hardware, software and models Akida Tag Akida 2 – Field Programmable Gate Array (FGPA) More ways to access BrainChip intelligence More products and services to buy More customer contacts than ever Easier to access More discussions More solutions More products to buy More customers in the pipeline 5 Refer to slide 19 for more details.
Page 6
Top of funnel Demonstration/ Evaluation Solution/ Support Negotiations Why BrainChip? Growing opportunity pipeline Growing interest from a wide range of sectors in the ‘funnel’ 6 Quarterly Investor Update – Q2 2026 SoC Vendor Medical/Healthcare Defense/Aerospace Industrial/Robotics Consumer Automotive/Mobility IoT Prospects
Page 7
Why BrainChip? Key competitive advantages BrainChip leads in ultra low power, ultra low latency and data sovereignty 7 Quarterly Investor Update – Q2 2026 Power consumption Operating performance (TOPS) 500 100 20 1 100 mW 1 W 200 W • Ultra - low power • Ultra - low latency • No cloud fees/tokens • High security • Ultimate privacy
Page 8
8 Space Imaging Safety Drone RISC-V AI Subsystem WiFi Wearables Advanced Radar Medical Wearables Radar & Drones Defense Smart MeteringIndustrial Sensing ASIC Design RF and Signal Processing Quarterly Investor Update – Q2 2026 Why BrainChip? Diversified Customer Base BrainChip has real deployments already across a broad range of market segments and sectors
Page 9
Neuromorphyx relationship expanded Q2 highlights BrainChip’s commercialisation strategy continues to advance 9 Quarterly Investor Update – Q2 2026 BrainBoard1500 Purchased 1,200 chips in Dec - 25 Entered partnership for BrainBoard 1500 development in Aug - 26 “Akida enables Orama.AOI to deploy high - performance, power - efficient inspection solutions built for the demanding conditions of modern manufacturing.” - L. Thomas Heiser, President and CEO of Heiser Industries Orama.AOI industrial defect detection partnership “…lightweight inference tasks can be directed to Akida instead of going by default to a GPU.” - IT Brief Australia, itbrief.com.au IBM Spectrum Symphony Community Akida Bundle launched https://developer.brainchip.com/symphony - community - akida - bundle/ AKD1500 chips arrive New revenue streams Accelerates product evaluation M.2 Board Next generation products progressing AKD2000 (MPW) Dec - 26 tape out on schedule Akida Pico IP ready Sequora available for customer evaluation Q4 2026 Partnerships expand market access in significant semiconductor communities of Korea and Taiwan First Akida 2 RTL (IP) delivery
Page 10
AKD1500 chips create a new revenue stream Physical chips delivery driving revenue and accelerating customer product evaluation 10 Quarterly Investor Update – Q2 2026 ~25% chips sold to date Chip modules Regular, low volume sales of Akida Tag, M.2, PCIe and RPi 5 boards Evaluating Timing and volume of next batch AI - enabled products for defense environments Multi - year partnership advancing with chip order Developed epileptic seizure detecting glasses Partnership to create and distribute developer - friendly BrainBoard1500 Strong interest From customers in the pipeline 1 1. Pipeline customers refers to slide 6. Award - winning - Won Global AI award in Edge AI + IoT Intelligence - Named Truly Innovative Tech by Electronics For You
Page 11
Akida powers hybrid computing Akida strong in hybrid computing Balancing workloads across Akida, CPU and GPU can significantly improve computational efficiency and inference latency IBM Spectrum Symphony Community Akida bundle launched Provides easy - to - test neuromorphic processors as standard compute resource Open - sourced for ease of use – available on GitHub Akida/Symphony combination demonstrations bundled with software Akida/Symphony success can be extended to other workload management solutions (like Kubernetes) Akida enhances computational efficiency for appropriate workloads 11 Quarterly Investor Update – Q2 2026
Page 12
Investor questions
Page 13
Appendices 13
Page 14
Investor calendar 14 Quarterly Investor Update – Q2 2026 Date Detail 11 Sep 10:30am Investor update (Q2) 22 Sep 11:45am AI & Semiconductors: ASX - Listed Stocks Investor Webinar Register here Late Sep Australian broker desk briefing Late Oct Q3 2026 QAR and Appendix 4C Oct/Nov Investor update (Q3) Jan 2027 Q4 2026 QAR and Appendix 4C Feb 2027 Annual result and investor update (Q4) Note: Dates and times are AET and subject to change.
Page 15
More on BrainChip ASX (ASX:BRN) – Financial and operational updates, material disclosures LinkedIn – Key updates, awards, partnerships Instagram – Key updates, awards, partnerships, videos Podcast ( Apple , BrainChip website ) – BrainChip updates, investor questions answered YouTube – Experiments, podcasts (investor and product) X – BrainChip industry events, updates, demonstrations Discord – Developer discussion groups BrainChip communication channels and socials 15 Quarterly Investor Update – Q2 2026 Investors Customers Developers Other communications Media releases - brainchip.com/media/ Product events - brainchip.com/events/ Product demonstrations - brainchip.com/videos - podcasts/ Podcasts (investor and product) - brainchip.com/podcast/ Webinars - brainchip.com/webinars/ Connect Contact us at IR@brainchip.com to be added to our mailing list.
Page 16
Glossary Quarterly Investor Update – Q2 2026 ASIC Application - Specific Integrated Circuit BLE Bluetooth Low Energy CNNs Convolutional Neural Networks – traditional image processing that is computation heavy CPU Central Processing Unit – the ‘Brain’ of a computer Edge Outer - most boundary of a network – a device or sensor – that receives inputs Edge AI Edge computing using neural networks - the processing of information received at the Edge (device or sensor), rather than sending to the cloud (data centre) for pro cessing Foundry Specialized manufacturing plant that builds integrated circuits and microchips designed by other companies FPGA Field - Programmable Gate Array – customisable integrated circuit chip that you can program and re - program. BrainChip has develope d an FPGA with Akida 2 IP GPU Graphics Processing Unit – speciali z ed electronic circuit running in parallel to CPU for high - performance computing, deep learning and AI KWS Key Word Spotting e.g. a wake work like “Hey Siri”, “Alexa” Neuromorphic Neuro (‘brain’) – morphic (‘shaped’). Intelligence that is brain like, and event - driven, not processing everything. Enables very low power usage NPU Neural Processing Unit – speciali z ed computer chip designed to handle AI OEM Original Equipment Manufacturer – typically used for parts used in another party’s finished product RTL Register Transfer Level – a foundational step in chip design and allows engineers to test, simulate and optimise designs for pow er, performance and area (PPA) SoC System on a Chip – functions as an entire computing system on a chip – can include CPU, GPU, memory controllers, Wi - Fi radios an d AI processors SWaP - C Size, Weight and Power (and Cost/Cooling) – Key performance optimisation indicator for computing hardware Tape out The final stage of silicon chip design – where the blueprint is sent to a Foundry for physical silicon fabrication TENNs Temporal Event - based Neural Networks – lightweight AI architecture designed for edge devices with significantly lower computatio nal processing compared to CNNs TOPS Tera Operations Per Second – the measure of computing power of a chip / also GOPS (Giga Operations Per Second) 16
Page 17
Sparse by Design, Always-On by Default Our key differentiator: event-based sparsity that does more with less. vs. What That Buys You 1+ year Always - On Operation On a Battery mW Ultra-Low Power 10X Computational Efficiency <10 ms Real-Time Response How Akida Stays Always-On Quantization Near-memory compute Event-based sparsity 17 Real-Time Intelligence That Never Sleeps
Page 18
Built for Data Sovereignty Conventional AI has to send data somewhere to be useful. Akida doesn’t — every inference runs on-chip, at the point of sensing. Where the Data Goes Cloud-Dependent AI Sensor Network Cloud Decision Every transit is a vulnerability. Every hop is a compliance liability. Akida — Inference On-chip Sensor Decision No data leaves the device — ever. The Result? Zero Data Transmissions Off The Device <1W Inference Power Sub-ms On - Chip Latency Non - negotiable in defense, medical and implantable devices, communications, and critical infrastructure. 18 Your Data Never Leaves the Device Private by architecture, not by policy — a hardware guarantee, not a software promise
Page 19
BrainChip Offerings Dev. Kits Tools Silicon License IP Reference Platforms metaONNX 1 2 AKD1500 Raspberry Pi 5 Dev. Kit AKD1500 PCIe Card AKD1500 Edge AI Co-Processor AKD1000 Edge AI SoC (System-on-Chip) AKD1000 PCIe Card Akida Comms Akida Radar 19 Scalable Platform from IP to Solutions AKD1000 Raspberry Pi 5 Dev. Kit AKD1000 M.2 Card AKD1500 M.2 Card Models AkidaTag
Page 20
20 Multi-tech portfolio ideal for drones, ADAS, space, etc. Specialized models in radar and Signal Intelligence Early customer interest Multi-tech portfolio ideal for wearables Ultra-low power, always-on AI Low-cost Low-power and low-cost Custom LLM with RAG demo Use less memory bandwidth Edge AI Solution Provider Wearables Mobility Voice Assistance Micro-Doppler Classification Adaptive Target Recognition Activity Detection
Page 21
Akida Communication Reference Platform brings ultra - low SWaP - C RF intelligence on - device. Watch Demo >> Watch Demo >> RF Signal Intelligence Radar Object Classification Watch demo >> Real - time drone detection and classification at the edge. AkidaTag Keyword Spotting Self - learning in keyword spotting, bringing intelligence to battery - powered devices. Explore Solutions >> Explore Solutions >> Explore Solutions >> 21 Edge AI Use Cases Watch Demo >> Explore Solutions >>
Page 22
Why BrainChip? Trusted By Industry Leaders Semiconductor, software, supply chain and AI leaders turn ing Akida from licensable core into a production system 22 Partner Ecosystem IP Technology CPU and processor IP validated with Akida Solutions Enablement Models, tools and applications built on Akida Silicon Integration Akida IP into custom silicon and foundry flows OEM Integration Akida built into edge hardware and systems Industry Organizations & Universities Standards bodies and research collaborations * ForwardEdge ASIC is a Lockheed Martin subsidiary. *
Page 23
Getting Started with Akida 23 Developer KitsDeveloper Resources See BrainChip product offerings including boards, cards , and boxes . Integrate event - based neuromorphic AI into your next design. User Guide & Data Sheet Models Tools Register Now Akida Cloud Run Akida just like a chip on your bench. Test, benchmark, and validate your AI models in Akida Cloud . BrainChip University Downloads Resources
Page 24
What It Is Ultra-low-power neuromorphic AI module with BLE connectivity Runs multiple AI models entirely on-chip, battery powered - no cloud Powered by BrainChip AKD1500 processor and Nordic nRF5340 SoC Akida TAG : Always on Edge AI, No Cloud Required Roadmap to Market Q3/4 2026: 15 eval units to beta customers across 3 global sites Nov 2026: 250-unit production build (full assembly with enclosure) Go to Market: Q4 2026 Akida TAG Target Applications Keyword Spotting : voice wake and command Edge Learning on KWS (Key Word Spotting) Human Activity Recognition : Fall vs ADL (Activities of Daily Living) Visual Wake Word : Person / No Person Watch Activity Detection Product Status 10 fully functional AkidaTag boards prototypes on hand 250 for production, 10 pre - orders, 30 brochure requests 2K targeted e - mails sent, high click rate 24
Page 25
What It Is An end - to - end radar + AI platform that enables real - time target classification on low - power edge hardware. The platform integrates radar signal processing, micro - Doppler feature extraction, and neuromorphic AI inference into a single deployable system running on Raspberry Pi - 5 + AKD1500 Roadmap to Market Integrate additional radar AI capabilities (CFAR, beamforming, SAR) Partner with defense and commercial customers for field validation Transition from reference platform to customer product designs Drone detection system on the edge Target Applications Drone vs bird classification and airspace awareness Human activity recognition (fall, motion, presence) Industrial monitoring(Vibration analysis, machinery health monitoring) Automotive sensing (occupancy, gesture, safety, autonomous driving) Smart environments (security, intrusion, monitoring) Product Status Webinar presentation held Apr - 26 Demonstrating full pipeline: radar → AI → classification Demonstrated live at Microelectronics US event Apr - 26 Radar Reference Platform : Enabling Intelligent Sensing at the Edge 25
Page 26
What It Is Compact, ultra - low - power, event - based signal analysis reference platform for real - time RF waveform classification. Powered by BrainChip AKD1500 NPU with 70 MHz – 6 GHz SDR front - end. Communications Reference Platform : Real-Time RF Intelligence at the Edge Platform Roadmap AKD1500 hardware demo (current) AKD2 5 00 MPW demo (Future) Sequora : TENNs & SSM inference (Future) Comms Reference Platform Target Applications Spectrum Monitoring & Signal Intelligence (SIGINT) Electronic Warfare: Classify & Retransmit Threat Detection & Modulation Classification UAV / Airborne RF Awareness Interference Detection Key Specifications 70 MHz – 6 GHz frequency range 2.8 M Samp / Sec continuous BW, 50M Samp / Sec Burst BW Low SWaP_C implementation Raspberry Pi 5 host Low SWaP implementation 26
Page 27
What It Is TENNs model Voice Assistant running on a mobile phone without connection to internet. LLM models using Akida GEN 3 model, showing latency and performance for edge applications. Private Voice Assistant Reference Platform Platform Roadmap Improve mobile chip support and stability with 8GB memory footprint Improve tokens/sec performance with software/model optimization targets Develop Google Play application for evaluation/listing/download Private Voice Assistant Target Applications Private Voice Assistant Fully Functional on Android Mobile app uses same models that will run on Gen3 core Key Specifications Automatic Speech Recognition: TENNs: SSM - based, fp32, 41M/480MB RAGLLM (SSM - based, 1.2B/2.5GB fp16) Text to Speech (TTS): fp32 (326MB) -- best quality, fp16 (163MB), int8 (88MB), Configurable styles/speed, speech/text, improvements with sentence - based inference Low SWaP implementation 27
Page 28
Brainchip.com investor.Brainchip.com Troy Dahms Head of Investor Relations and Financial Communications IR@Brainchip.com 28