BRN Discussion Ongoing

[Space Cadet]

Hi Bravo,

This is a real sign of life.

Would the real Akida 1500 please stand up.

Apparently the Akida 1500 we had grown to know and love was just the reference version. Essentially the original version of 1500 was just a cut down version of Akida 1 without the ARM Cortex, so it looks like there has been some changes in the circuit design to necessitate a new tape-out.

The Akida 1 datasheet now "supports" 128 MACs per node, (4*4 MACS). The inclusion of MACs in the design has only been public since the advent of TENNs, but the Akida 1 datasheet does not mention TENNs capability. Does this mean that the original brilliant SNN NPU has been outshone by MACs?

1500 is cheaper than Akida 1 because it does not entail the ARM Cortex licence and because it has a smaller wafer real estate footprint (also because of the lack of the Cortex) meaning more chips with your wafer.

1500 is processor agnostic. It can be used with any MCU/CPU/GPU.

The press release refers to "Moving to volume production" as a strategic step, and this suggests to me that the tapeout is not just for a single customer.

The strong customer demand (military, areospace, medical, wearables) suggests that the 1500 IP is to be used either as a standalone chip/chiplet, or as part of a NSoC which does not include the ARN Cortex.

Frontgrade is proposing to make its own NSoC, GR801 including Akida 1 IP with RISC-V CPU, but it would presumably need the new volume production Akida 1500 IP, so the tapeout will support the GR801.

Onsor is another potential customer that we know about.

I would guess that the micro-Doppler radar (RTX/ISL) would require the higher precision of 8-bit or 16-bit), but there are many other applications for 1500.

This bit suggests the IP only business model was overambitious, and Sean's initial assessment of Akida 1 was off the mark:

"As the market matures, and partners become more deeply integrated into the ecosystem, it is expected that a portion of this demand will transition into an IP licensing model."

The timeline of Q3 2026 is a bit deflating.

Bascom Hunter 3U VPX? 1500 instead of 1000. Any chance.

SC

 

[Rach2512]

𝐅𝐫𝐨𝐦 𝐀𝐆𝐈 𝐭𝐨 𝐀𝐈 𝐏𝐨𝐫𝐧: 𝐓𝐡𝐞 $35 𝐓𝐫𝐢𝐥𝐥𝐢𝐨𝐧 𝐁𝐨𝐧𝐟𝐢𝐫𝐞 𝐨𝐟 𝐭𝐡𝐞 ‘𝐒𝐜𝐚𝐥𝐞’ 𝐃𝐞𝐥𝐮𝐬𝐢𝐨𝐧. '𝘚𝘤𝘢𝘭𝘦' 𝘸𝘢𝘴 𝘵𝘩𝘦 𝘮𝘢𝘯𝘵𝘳𝘢, 𝘵𝘩𝘦 𝘮𝘦𝘵𝘳𝘪𝘤, 𝘢𝘯𝘥 𝘯𝘰𝘸… | Srini Pagidyala | 34 comments

𝐅𝐫𝐨𝐦 𝐀𝐆𝐈 𝐭𝐨 𝐀𝐈 𝐏𝐨𝐫𝐧: 𝐓𝐡𝐞 $35 𝐓𝐫𝐢𝐥𝐥𝐢𝐨𝐧 𝐁𝐨𝐧𝐟𝐢𝐫𝐞 𝐨𝐟 𝐭𝐡𝐞 ‘𝐒𝐜𝐚𝐥𝐞’ 𝐃𝐞𝐥𝐮𝐬𝐢𝐨𝐧. '𝘚𝘤𝘢𝘭𝘦' 𝘸𝘢𝘴 𝘵𝘩𝘦 𝘮𝘢𝘯𝘵𝘳𝘢, 𝘵𝘩𝘦 𝘮𝘦𝘵𝘳𝘪𝘤, 𝘢𝘯𝘥 𝘯𝘰𝘸 𝘵𝘩𝘦 '𝘔𝘰𝘴𝘵 𝘌𝘹𝘱𝘦𝘯𝘴𝘪𝘷𝘦' 𝘮𝘪𝘴𝘵𝘢𝘬𝘦. We’re knee-deep in 𝐒𝐥𝐨𝐩𝐨𝐜𝐨𝐥𝐲𝐩𝐬𝐞, industrialized attention extraction at scale. 𝐁𝐮𝐛𝐛𝐥𝐞 𝐓𝐫𝐨𝐮𝐛𝐥𝐞 sold as “AI-driven productivity” is now brewing...

www.linkedin.com

 

[Space Cadet]

First quarter 2026 tap out I believe is the reference, in saying that we still have plenty of irons in the fire with Akida 1 and 2 that will be profitable very soon imo.

Regarding Onsor it would seem they aren't using the 1500 as the ducks don't line up with early 2026 release . So this medical application would be something else

TCS have been all over us for years and said they were working on health devices if I am remembering right.

SC

 

[Space Cadet]

This point has been made elsewhere ...

In late December, BrainChip signed a US$100,000 contract with U.S. defence contractor Bascom
Hunter, for the sale and support of AKD1500 chips for full scale evaluation of commercial products,
which will provide significant improvements over traditional CPU, GPU, and TPU processors and
are considered optimal for low Size, Weight, and Power (SWaP) Machine Learning (ML)
applications. Bascom Hunter works collaboratively with 3rd party technology vendors, such as BrainChip to
integrate their technology for Defense and Intelligence applications.

If the full scale evaluation is complete, perhaps we have been given the green light.
Fantastic news if this is the case.

Doesn't mean the IP negotiations with other entities will not progress ... in fact this may give any negotiations added impetuous in the knowledge that BasCom Hunter are moving forward with rolling out AKD1500 chips ( if that is indeed the case ).

Beat me to it

SC

 

[Space Cadet]

Now 1 question. Is 3rd qtr 2026 Jul - September qtr, or as we are on ASX, Jan - Mar 2026 financial year? Prefer the latter.

SC

 

[smoothsailing18]

Atley Loughridge demos virtual teleprompter with Spectacles at Lens Fest | Summer Wu posted on the topic | LinkedIn

Today at Lens Fest, Atley Loughridge did the most incredible live demo- streaming her view of a virtual teleprompter from Spectacles to the screen, and pinching her fingers to advance her slides, no clicker necessary. There's a lot of powerful technology behind this teleprompter experience- in...

www.linkedin.com

Any thoughts on these spectacles ?

They would be a great partner.

 

[Xray1]

I don't understand why with todays announcement and the contents contained therein, as to why some of these material facts such as the "Tapeout of Akida 1500" being imo "Price Sensitive" in nature were not at all disclosed by the Co in the most recent ASX Response letter ???

 

[smoothsailing18]

I don't understand why with todays announcement and the contents contained therein, as to why some of these material facts such as the "Tapeout of Akida 1500" being imo "Price Sensitive" in nature were not at all disclosed by the Co in the most recent ASX Response letter ???

I think price sensitive announcements will only relate to proven income.

 

[Baron Von Ricta]

Atley Loughridge demos virtual teleprompter with Spectacles at Lens Fest | Summer Wu posted on the topic | LinkedIn

Today at Lens Fest, Atley Loughridge did the most incredible live demo- streaming her view of a virtual teleprompter from Spectacles to the screen, and pinching her fingers to advance her slides, no clicker necessary. There's a lot of powerful technology behind this teleprompter experience- in...

www.linkedin.com

Any thoughts on these spectacles ?

Reminds me of the 70's, they didn't work then, maybe now they do.

 

[Xray1]

Just a historical reminder:
Brainchip Holdings Ltd (BRN) floated on the Australian Securities Exchange (ASX) on Wednesday, 9 November 2011.
BRN's current share price of ~$0.20 which is a current -$0.05c decrease in S/P or roughly a ~20% discount to its original listing offer price of $0.25 cents.

 

[Xray1]

I think price sensitive announcements will only relate to proven income.

I Disagree with your view, that Price sensitive announcements is related to a proven income ..........

 

[Diogenese]

Bascom Hunter 3U VPX? 1500 instead of 1000. Any chance.

SC

Hi SC,

That's an interesting thought. Removing ARM Cortex in 5 Akida 1s would remove the cost of the 5 ARM licences. The SNAP card already has a couple of ARM processors so one of them could be pressed into service for configuring the 1500s.

https://bascomhunter.com/deg/digita...c-processors/asic-solutions/3u-vpx-snap-card/

FPGA PROCESSOR (PRE/POST PROCESSING)

Zynq US+ Quad Core ARM A53 Zynq US+ Dual Core ARM R5 930,300 FPGA Logic Cells

Bascom Hunter’s SNAP Card (Spiking Neuromorphic Advanced Processor) is a high-performance 3U OpenVPX AI/ML processor built for rugged, mission-critical environments. The card is SOSA-aligned, HOST-compatible, and conduction cooled. It combines a Xilinx UltraScale+ RFSoC FPGA with five BrainChip AKD1000 spiking neuromorphic processors to achieve the best of both signal processing and neuromorphic computing – featuring a total of 6 million neurons and 50 billion synapses across the card. Unlike traditional machine learning accelerators such as GPUs or TPUs, neuromorphic processors are designed to mimic the biological efficiency of the human brain, allowing the SNAP Card to run multiple ML models in parallel at exceptionally low power — just 1W per model — without sacrificing speed or accuracy while the FPGA enhances input/output operations and performs any desired signal processing tasks. This combination in a rugged, interoperable, and military-hardened package makes Bascom Hunter’s SNAP Card the ideal solution for the concurrent and parallel processing of real-time, multi-modal, and multi sensor data on autonomous, unattended, denied, or otherwise battery constrained military systems.


FN: Gate Arrays, especially FPGAs, have come a long way from the original array of AND and OR (NAND & NOR) gates to now include processors along with about 1 million logic gates, and there are other advanced special purpose FPGAs.

 

[davidfitz]

Just a historical reminder:
Brainchip Holdings Ltd (BRN) floated on the Australian Securities Exchange (ASX) on Wednesday, 9 November 2011.
BRN's current share price of ~$0.20 which is a current -$0.05c decrease in S/P or roughly a ~20% discount to its original listing offer price of $0.25 cents.

I could be wrong but isn't this information relating to Aziana (AZK) which was initially a mining company somewhere in Africa. I believe that Brainchip came into the equation in 2015 and the ASX code was changed sometime in 2016 to BRN.

Still a good point as not many of us are lucky enough to still be in front on this one.

Sorry a bit bored at work today and thought it had been a while since I posted anything.

Back to the dark shadows for me.

 

[smoothsailing18]

I Disagree with your view, that Price sensitive announcements is related to a proven income ..........

I see future income from it 100% however that's unwritten when we receive that income and how much we receive ,so my guess is that's why it sits here and not on the ASX.

 

[Tezza]

Is this the butter up before the qr which will likely be followed by the cr.
Hope not

 

[JDelekto]

I know I keep harping on about this, but the new 1500 design "supports" 128 MACs per node. I don't know what exactly "supports" means, but if it means that the original Akida NPU has been replaced by MACs, I'd be saddened to see the demise of the brilliance and elegance of the original NPU design. Perhaps it was the move from 1-bit to 4-bits that was the writing on the wall, necessitating the inclusion of MACs to support the 4-bit configuration, and leading over time to the conclusion that, for multi-bit, MACs were better. If this were to have been the case, the economic realities of wafer real estate would have seen to the demise of the original NPU even if it were better for 1-bit applications. So once again, Progress 1, luddites 0.

Just a thought, but is it possible that the support for the 128 MACs per node is not specifically to replace their original NPU design, but to integrate with sensors that are not neuromorphic in nature, such as Prophesee's event-based vision cameras? This is just speculation on my part, but maybe it is to support the conversion of CNN to SNN in hardware without the need to capture and convert the models offline using their SDK tools. It would also have the direct benefit of allowing customers to use existing models without doing their own conversions.

 

[Boab]

Is this the butter up before the qr which will likely be followed by the cr.
Hope not

That's what I didn't want to say out loud. Let's hope not.

 

[smoothsailing18]

Is this the butter up before the qr which will likely be followed by the cr.
Hope not

It wouldn't be the first time.

 

[Diogenese]

Just a thought, but is it possible that the support for the 128 MACs per node is not specifically to replace their original NPU design, but to integrate with sensors that are not neuromorphic in nature, such as Prophesee's event-based vision cameras? This is just speculation on my part, but maybe it is to support the conversion of CNN to SNN in hardware without the need to capture and convert the models offline using their SDK tools. It would also have the direct benefit of allowing customers to use existing models without doing their own conversions.

As I said, I'm not certain what "support for 128 MACs per node" means, but there is no mention of an NPU per se.

I think it was Tony Lewis who was reported as saying they initially had trouble getting recurrence (RNN) to work with TENNs, affecting ML. I'm guessing that TENNs is happy on MACs, so initially they may have used MACs and the old NPU for ML together, but, once they got TENNs bedded down with recurrence, the old NPU was no longer needed. I still think the old NPU would have been best for 1-bit - it is such an elegant solution - but the cost of the extra wafer real estate would have ruled it out once MACs were shown to be better for multi-bit/TENNs.

CNN-to-SNN would also be affected by the 1-bit/multi-bit split. The actual CNN-to-SNN converter did not involve the NPU, so it could still be the same with TENNs MACs as it was with the NPU.

 

[HopalongPetrovski]

Is this the butter up before the qr which will likely be followed by the cr.
Hope not

It's the butter up to the institutional investors Sean hopes to interest in our wares on Wednesday night.

Semiconductor Australia 2025​

We’re pleased to announce that BrainChip CEO Sean Hehir will be presenting at Semiconductor Australia 2025 on October 22nd at ILUMINA in Sydney.
Semiconductor Australia is the nation’s premier deep-tech conference, bringing together industry leaders, policy makers, and the investment community. Last year’s inaugural event attracted over 1,000 participants.
Sean will discuss how BrainChip’s neuromorphic AI technology is advancing Australia’s semiconductor capabilities and reinforcing our position as a leader in edge AI innovation.
Join us at this premier event to connect with Australia’s deep-tech ecosystem and hear firsthand from our leadership team about BrainChip’s strategic direction in the global semiconductor market.