Ann: Annual General Meeting Addresses and Presentation, page-2

*BrainChip AGM summary

The AGM message was broadly positive but still acknowledged that 2025 did not convert into the commercial outcomes management wanted. Sean Hehir said BrainChip is now in a stronger position, but the key job is turning technical progress into revenue and larger commercial wins.

Main takeaway: BrainChip is trying to move from being mainly an IP/technology provider into a fuller edge-AI product company, with chips, modules, reference designs, developer tools and complete customer-ready solutions.

Key points:

1. AKD1500 is central to 2026

Management framed AKD1500 as a major turning point. It is now positioned as a volume-production product, not just a reference design or sample. Peter said the next phase begins in the second half of 2026 when AKD1500 production chips become available in large quantities.

2. Parsons is the standout commercial validation

The AGM highlighted the multi-year Parsons supply agreement, including an initial order of 10,000 AKD1500 chips. This was framed as validation of AKD1500 for mission-ready defence platforms and constrained edge environments.

3. Defence, aerospace, medical, wearables and industrial are the target markets

BrainChip is deliberately focusing on sectors where low power, low latency and on-device learning matter most. Examples included Frontgrade Gaisler space SoC, Raytheon/AFRL radar work, ISL radar/electronic defence, Onsor epilepsy glasses, Arquimea drones, Andes RISC-V, Chelpis/Mirle robots and HaiLa low-power RF sensing.

4. Reference designs are a major strategy shift

They admitted customers need easier paths to adoption. So BrainChip is launching reference design platforms for wearables, radar, speech assistants and electronic warfare. That means less “here is the chip, now figure it out” and more “here is a working template you can build from.”

5. TENNs, SSMs and Akida GenAI are now core to the story

Sean said the industry is increasingly recognising State-Space Models, which underpin BrainChip’s TENNs. BrainChip plans to deliver Akida GenAI in the second half of 2026, aiming at on-device generative AI and local LLM-style capability without cloud connectivity.

6. Kevin D. Johnson’s IBM/Symphony work was explicitly showcased

This is notable. Peter used Kevin’s work as an example of AKD1000 being used in larger systems, where Akida processes raw sensor inputs at the edge and sends compact 128-byte observation records into systems like GPFS, IBM Symphony and Palantir. The AGM slides even show the “Solution Courtesy of Kevin D. Johnson (IBM)” diagram.

7. Management is optimistic, but still asking for patience

Peter compared BrainChip’s journey to long deep-tech maturation stories like NVIDIA, Microsoft, Tesla and Netflix. The message was: deep-tech takes time, but BrainChip believes it is approaching an inflection point.

My read

This AGM was much stronger strategically than financially. The company is saying: the tech is ready, the product range is broader, AKD1500 is moving into volume, reference designs will make adoption easier, and defence/aerospace/medical/wearables are becoming real commercial lanes.

The bullish part is Parsons, AKD1500 volume availability, GenAI/TENNs positioning, and the fact Kevin Johnson’s architecture was formally acknowledged.

The cautious part is that management openly admitted execution has not yet translated into the commercial outcomes shareholders expected. So 2026 now becomes a proof year: BrainChip needs to convert pipeline, chip orders, reference designs and IP opportunities into meaningful revenue.
*gpt5.4