BRN Discussion Ongoing

[Deleted member 3781]

Sean in Australia

And…..

without teleprompter…. Impressive

 

[davidfitz]

Well, if it wasn't obvious over the last few months, 20 cents is definitely where someone wants it to be.

Over 15 million traded so far and it has barely moved!

Absolutely ridiculous in my opinion

 

[Cirat]

Well, if it wasn't obvious over the last few months, 20 cents is definitely where someone wants it to be.

Over 15 million traded so far and it has barely moved!

Absolutely ridiculous in my opinion

Have to agree - has been like that for at least the week.

I've never seen the market depth for BRN with such large blocks on the Buyers side not moving the price up.

 

[Baron Von Ricta]

Sean in Australia

Well at least Sean is very positive. I liked this:- When it was put to Sean that Brainchip was sort of a story and not a business, Sean replied, "Watch us now, just watch us now." (4.23)
Well I say to that, "bring it on, just bring it on."

 

[Rskiff]

Well at least Sean is very positive. I liked this:- When it was put to Sean that Brainchip was sort of a story and not a business, Sean replied, "Watch us now, just watch us now." (4.23)
Well I say to that, "bring it on, just bring it on."

Also said re revenue " revenue due about now" Bring it on!!!

 

[Flenton]

Sean in Australia

When do you expect to see revenue.
About now, right about now.

 

[Bravo]

Here's an article published a couple of days ago by Forbes titled “Neuromorphic Finance: When the Trading Floor Thinks Like a Brain”.

I doesn’t mention BrainChip directly, however, all the points are entirely applicable. It mentions on-device inference, real-time detection, temporal modelling, efficient processing, which is exactly where BrainChip's AKida shines.

Just think, if the finance industry adopts neuromorphic computing, then that opens up a whole new addressable market!

It would also validate the idea that neuromorphic processors could be mission-critical in high-frequency, streaming, decision-heavy fields - finance being one of the toughest.

The other thing that struck me in reading this article was that it suggests that fraud detection could be an area that would land first, the example given is an ATM fitted with neuromorphic fraud detection chips.

This reminded me of our cybersecurity work with Quantum Ventura + Lockheed. Naturally I've started to ponder over how Akida might work in this ATM use case.

Because ATMs sit idle most of the time; an event-driven coprocessor like Akida, could watch continuously while the main PC stays asleep.Temporal anomaly detection would be Akida's great strength since fraud signals change over time. And then you could also add Akida's on-device learning capabilities where it could incrementally learn each ATM’s environment without cloud retraining.

Imagine if there were Akida chips inside every ATM on earth. There are apparently upwards of 2.9 million ATM worldwide and banks would need to order two chips (keeping one for redundancy purposes), then that's upwards of 5.8 million chips, just for that one single use case.

Neuromorphic Finance: When The Trading Floor Thinks Like A Brain​

ByAnusha Nerella,
Forbes Councils Member.
for Forbes Technology Council
COUNCIL POST | Membership (fee-based)
Oct 28, 2025, 08:00am EDT
Share
Save
Anusha Nerella leads fintech automation at Global fintech firm, focused on Java Engineering, AI and intelligent automation.

getty
For decades, Wall Street’s advantage was defined by speed. Firms poured billions into faster cables, denser data centers and GPU clusters that consumed enough power to run small cities. The prize was shaving microseconds off trades.
But here’s the hard truth: We’ve hit peak speed, and it’s not enough anymore.
Financial systems today aren’t failing because they’re too slow. They’re failing because they can’t adapt. Black-box AI models churn out decisions regulators can’t explain. Fraudsters pivot faster than detection engines. Risk systems buckle under shocks they weren’t trained to anticipate.

The next competitive edge won’t come from brute force. It will come from systems that think more like markets themselves: event-driven, adaptive and efficient. That’s the promise of neuromorphic finance—a concept still emerging but already showing potential in adjacent industries.

Why Markets Need A Brain, Not A Bigger Calculator​

Markets don’t move logically. They behave like neurons firing spontaneously, unpredictably and chaotically. A rumor spikes volatility. A politician’s tweet swings currencies. A regulation drops, and compliance engines scramble.
Traditional compute, built for batch processing and deterministic math, wasn’t designed for this. Neuromorphic computing is. Neurons fire only when triggered, networks adapt continuously and power use falls dramatically. Instead of calculating everything, neuromorphic hardware calculates only what matters.
While no major financial institution has deployed full neuromorphic systems yet, early proofs from other domains hint at what’s possible.

Early Signals From Other Fields​

China’s Darwin Monkey Supercomputer simulates over 2 billion neurons and more than 100 trillion synapses on about 2,000 watts—the draw of a hair dryer. Imagine portfolio stress tests at that efficiency.




Innatera’s Pulsar Chip powers edge devices at roughly 1/500th the energy of GPUs, showing how point-of-sale terminals could one day detect fraud before data even leaves the device.
IBM’s TrueNorth chip demonstrates how event-driven architectures process streams efficiently. Replace pixels with trades, and you can envision compliance tools that flag rogue activity instantly.
These aren’t financial deployments, but they prove the model: Systems that react only to signal can outperform those that brute-force every calculation.

Lessons From Today’s Financial Trenches​

In my own work building fraud detection and compliance platforms, the bottleneck was rarely speed—it was explainability. Regulators demanded audit trails, but deep-learning models produced only opaque scores. Entire teams were tasked with building “explainability wrappers” around black-box models.
Neuromorphic approaches could change that. Spiking neural networks naturally leave event-driven traces. Each spike ties to a trigger, and that sequence becomes an audit log. Instead of reconstructing why a model flagged an anomaly, you simply follow the spikes. For regulators, that’s the difference between theory and trust.
This is still a vision, not a product you can buy today. But it shows why finance is a natural next step.

Where Neuromorphic Finance Could Land First​

When adoption begins, the first movers are likely to be:
• Fraud Detection: Traditional engines either choke on volume or consume enormous energy to catch marginal anomalies. Neuromorphic fraud engines could monitor streams continuously and react only when something looks off, enabling real-time detection at far lower cost.
• Trading Desks: High-frequency strategies still rely on brute-force scenario crunching, even though markets move on discrete events. Neuromorphic processors are designed for this unpredictability, acting more like a trader’s intuition than a calculator’s routine.
• Regulatory Compliance: Every rule change today triggers lengthy model-retraining cycles. Spiking models could adapt continuously, leaving spike trails that auditors can verify.

The Architect’s Playbook​

Firms considering this path should think evolution, not revolution:
• Start with edge pilots like payment terminals or ATMs equipped with neuromorphic fraud-detection chips that could prove value quickly once hardware is available.
• Adopt hybrid stacks that let CPUs, GPUs and neuromorphic chips each do what they do best.
• Shift to event-driven design architect systems to respond to signals rather than process everything in bulk.
• Embrace compliance by design where neuromorphic spike trails don’t just detect anomalies; they create natural audit logs regulators can trust.
The harder leap isn’t technical; it’s cultural. Many IT groups still think like accountants: Process everything, log everything, control everything. Neuromorphic finance will require a trader’s mindset: Act quickly, filter noise, adapt instantly.

The Road Ahead​

The last century of finance rewarded speed. The next will reward adaptability.
Banks that dismissed the cloud lagged for years. Traders who ignored electronic exchanges vanished. Neuromorphic finance is on the same trajectory. It's not yet here, but coming fast enough that pilots launched today could define tomorrow’s standards.
Because in the end, markets don’t reward the biggest machines—they reward the smartes

Neuromorphic Finance: When The Trading Floor Thinks Like A Brain

The next competitive edge won’t come from brute force. It will come from systems that think more like markets themselves

www.forbes.com

 

[Diogenese]

Sean in Australia

At 3:45 Sean refers to 2 generations of IP. m

While not 100% probative, I think this supports my hypothesis that TENNs has ousted PvdM's original SNN NPU circuit in generation 2, making way for the 128 MACs per node.

https://brainchip.com/wp-content/uploads/2025/04/Akida1-IP-Product-Brochure-V2.1-1.pdf

Self-contained neural processor
• Scalable fabric of 1-128 nodes • Each neural node supports 128MACs
• Configurable 50-130K embedded local SRAM
• DMA for all memory and model operations
• Multi-layer execution without host CPU
• Integrate with ARM, RISC-V with AXI bus interface

The SNN SoC chips that we have for sale are the original 1st generation PvdM design.

However, as I said, this is just an hypothesis at the moment. We will have to wait for clarification from a higher power.

 

[Cirat]

Well at least Sean is very positive. I liked this:- When it was put to Sean that Brainchip was sort of a story and not a business, Sean replied, "Watch us now, just watch us now." (4.23)
Well I say to that, "bring it on, just bring it on."

That is the most confident interview I've seen around the question of revenue.
It's like Sean wants to tell all of us he's made his 9M contract/s requirement by years end and ........... then maybe even more.

I'm feeling very confident in the company. It's been a long haul and 2026 may finally be the big turn around year with a great finish to 2025!!

GLTA, DYOR, This is not financial advice.

 

[Fullmoonfever]

Just on TENNS & Pleiades, including eye tracking model...the below updates were added on GitHub in Aug.


Aug 18

ktsiknos-brainchip
2.15.0-doc-1
5fa24a5

Upgrade to Quantizeml 0.18.0, Akida/CNN2SNN 2.15.0 and Akida models 1.9.0

Update QuantizeML to version 0.18.0​

New features​

• Introduced PleiadesLayer for SpatioTemporal TENNs on Keras
• Keras sanitizer will now bufferize Conv3D layers, same as ONNX sanitizer. As a result, SpatioTemporal TENNs from both frameworks will be bufferized and quantized at once.
• Dropped quantization patterns with MaxPooling and LUT activation since this is not supported in Akida
• Dropped all transformers features

Update Akida and CNN2SNN to version 2.15.0​

Aligned with FPGA-1696(2-nodes)/1695(6-nodes)​

New features and updates:​

• [cnn2snn] Updated QuantizeML requirement to 0.18.0
• [akida] Full support of look-up-table (LUT) activation in 2.0 FPGA (HRC and CNP are the only NPs supporting LUT). Known limitation: MaxPooling and LUT are not possible together.
• [akida] Extended support to 4-bit weights and 8-bit activation combos (was limited to 4-bit weights and activation)
• [akida] Depthwise convolution now supports global average pooling
• [akida] Added equality operator to akida.Component for easier mapping comparison
• [akida/cnn2snn] Dropped all transformers layers and features

Bug fixes​

• [akida] When a single FNP is requested in akida.create_device, it will no longer be both FNP and CNP to ensure proper minimum device computation
• [akida] Mapping will now properly reject skip branches not ending with a merge layer (CNP)

Update Akida models to 1.9.0​

New features and updates:​

• Updated OpenCV dependency to <4.12 because later versions are based on numpy 2
• Updated spatiotemporal_block with a temporal_first parameter that allows to invert spatial and temporal blocks order
• Optimized the spatiotemporal TENNs model for EyeTracking leading to 1.4x speed increase and 54% memory reduction in HW while maintaining accuracy
• Dropped all transformers features

Bug fixes​

• Fixed reset_buffers usage in TENNs training modules when evaluating an Akida model

Documentation update​

• Add mentions to LUT in user guides and APIs references
• Updated buffering section in Jester TENNs tutorial

 

[Bombersfan]

Well, if it wasn't obvious over the last few months, 20 cents is definitely where someone wants it to be.

Over 15 million traded so far and it has barely moved!

Absolutely ridiculous in my opinion

I agree, it’s bullshit. Somethings up, getting a speeding ticket on an 11% jump was disgraceful. Shorts have gone from 2%-6% in last 12 weeks. The pricks rarely get it wrong so it’s a bit concerning but with the avalanche of news lately it’s a dangerous game. Unless we announce a big deal in the very near future BRN have to make the capital call in the next month so maybe they are sweating on that. With 120m shorts currently open it would be fireworks if they could announce a “profitable overnight deal”!

 

[smoothsailing18]

At 3:45 Sean refers to 2 generations of IP. m

While not 100% probative, I think this supports my hypothesis that TENNs has ousted PvdM's original SNN NPU circuit in generation 2, making way for the 128 MACs per node.

https://brainchip.com/wp-content/uploads/2025/04/Akida1-IP-Product-Brochure-V2.1-1.pdf

Self-contained neural processor
• Scalable fabric of 1-128 nodes • Each neural node supports 128MACs
• Configurable 50-130K embedded local SRAM
• DMA for all memory and model operations
• Multi-layer execution without host CPU
• Integrate with ARM, RISC-V with AXI bus interface

The SNN SoC chips that we have for sale are the original 1st generation PvdM design.

However, as I said, this is just an hypothesis at the moment. We will have to wait for clarification from a higher power.

It's interesting Sean mentioned Radar but not LiDAR which I find intriguing as to why not

 

[Diogenese]

It's interesting Sean mentioned Radar but not LiDAR which I find intriguing as to why not

I think radar is near term, while lidar may be medium term - choose your own yardstick.

We know that Valeo's Scala 3 does not include Akida. That's not to say that Akida could not process Scala 3 signals. But, with automotive, there is a long lead time for proving the reliability and safety of new tech. The paint isn't dry on Akida generation 2.

The microDoppler radar (RTX/ISL) will need to be proven to work, but it will not have the same lead time as automotive.

The main development required for Akida to work on any signal type is the compiling of a model library/database characterizing the types of subject to be classified by Akida. There would be no circuit redesign required.

 

[Deleted member 3781]

Unbelievable ridiculous

 

[smoothsailing18]

I think radar is near term, while lidar may be medium term - choose your own yardstick.

We know that Valeo's Scala 3 does not include Akida. That's not to say that Akida could not process Scala 3 signals. But, with automotive, there is a long lead time for proving the reliability and safety of new tech. The paint isn't dry on Akida generation 2.

The microDoppler radar (RTX/ISL) will need to be proven to work, but it will not have the same lead time as automotive.

The main development required for Akida to work on any signal type is the compiling of a model library/database characterizing the types of subject to be classified by Akida. There would be no circuit redesign required.

Hi Diogenese .. In reference to your last paragraph I am thinking these models will at some point work well in chiplets . My mind keeps coming back to these being apart of the big picture for brainchip. anyway.

 

[bludybludblud]

@Bravo @Neuromorphia Yea in that case there shouldn't be any IP problem with Brainchip owning PLEIADES. And I wouldn't read too much into Rudy presenting this work in conferences (from an IP perspective at least), except for the unpleasant reminder that both brains behind the TENNs series of products are no longer with Brainchip. This would include TENNs eye, Pleiades, aTENNuate speech enhancement, Centaurus quantum speedup, etc.

Of course, it is totally possible whoever remaining at Brainchip is smart enough to continue this line of work. But may be hard to tell until there is a concrete patent filing or publication in top AI conferences for TENNs, which we have yet to seen without Olivier or Rudy. So does TENNs even have a future at Brainchip, or would the company take another 5 years pivoting to a new product line? "Could be, who knows?" as eloquently remarked by Dr. Tony Lewis ..

 

[smoothsailing18]

AI-First Devices: Can They Replace the iPhone? | Built In | sri charan itragunta sri charan itragunta

AI Phones: - T-Mobile and Deutsche Telekom revealed concept AI phones at events like Mobile World Congress. These phones eliminate traditional apps, using an AI assistant to perform all tasks via voice or text commands. The AI acts like a concierge, handling tasks such as booking flights...

www.linkedin.com

 

[IloveLamp]

"Neuromorphic computing inspired by the brain's really really remarkable energy efficiency could fundamentally transform how AI operates.

It presents a promising and exciting opportunity to support both large scale and edge AI, and edge AI I think is going to be something that that we're going to have to think about much more.

Its brain inspired design could help AI systems to become more adaptive, more energy efficient. And all of this is especially important as models become bigger and more complex and the questions we ask of computing and of AI become more complicated, more difficult, bigger data sets.

The center I know will unite leading researchers in neuroscience, advanced materials, algorithms and optical and electronic hardware to pioneer new low energy computing concept and devices."

 

[Guzzi62]

@Bravo @Neuromorphia Yea in that case there shouldn't be any IP problem with Brainchip owning PLEIADES. And I wouldn't read too much into Rudy presenting this work in conferences (from an IP perspective at least), except for the unpleasant reminder that both brains behind the TENNs series of products are no longer with Brainchip. This would include TENNs eye, Pleiades, aTENNuate speech enhancement, Centaurus quantum speedup, etc.

Of course, it is totally possible whoever remaining at Brainchip is smart enough to continue this line of work. But may be hard to tell until there is a concrete patent filing or publication in top AI conferences for TENNs, which we have yet to seen without Olivier or Rudy. So does TENNs even have a future at Brainchip, or would the company take another 5 years pivoting to a new product line? "Could be, who knows?" as eloquently remarked by Dr. Tony Lewis ..

I think people are jumping the gun here, or you might try to sow fear amongst investors? I am not sure what you are trying to achieve here?

TENNs are mentioned on BRN webpages as today (see below), so just because those 2 guys are not there anymore means it will have any impact.

Engineers are logging their work down for others in the company to see, so anyone with expertise in this field can carry on where they left, it's not like they have done a Nobel Prize worthy discovery, is it?

Those 2 guys are not gods, O made problems and got 5 cold toes in the ass. R wanted to work for N and left. And that's that, end of story!

Technology

Learn about BrainChip's innovative technology, featuring the Akida Neuromorphic Processor for efficient, low-power AI processing at the edge.

brainchip.com

 

[Pom down under]

@Bravo @Neuromorphia Yea in that case there shouldn't be any IP problem with Brainchip owning PLEIADES. And I wouldn't read too much into Rudy presenting this work in conferences (from an IP perspective at least), except for the unpleasant reminder that both brains behind the TENNs series of products are no longer with Brainchip. This would include TENNs eye, Pleiades, aTENNuate speech enhancement, Centaurus quantum speedup, etc.

Of course, it is totally possible whoever remaining at Brainchip is smart enough to continue this line of work. But may be hard to tell until there is a concrete patent filing or publication in top AI conferences for TENNs, which we have yet to seen without Olivier or Rudy. So does TENNs even have a future at Brainchip, or would the company take another 5 years pivoting to a new product line? "Could be, who knows?" as eloquently remarked by Dr. Tony Lewis ..