BRN Discussion Ongoing

[Bravo]

o

" The AKD1500 integrates seamlessly with x86, ARM, and RISC-V host microcontrollers and multiprocessor (MPUs) chips via PCIe or Quad SPI serial interfaces, supporting broad market adoption across a wide range of applications. The AKD1500 co-processor approach is ideal for a wide range of environments and industries, upgrading multi-processor SoCs within defence, industrial and enterprise settings, and upgrading embedded microcontrollers for AI solutions in healthcare, wearables, and consumer electronics without a complete system redesign."
My bold above.
That’s a compelling hybrid model: ARM for control + neuromorphic for inference.
Essentially we are just waiting for industry to demand the benefits of Neuromorphic and ARM is ready. At the CES26 the 'edge' was all the rage and it appears that we are now seeing the beginning to the Edge take off - finally.
No doubt ARM will integrate AKIDA into their Auto suitable chips - likely Gen 2?
The Quarterly was ok - shows we are making progress.
.

Hi Manny,

I agree the ARM + accelerator hybrid model makes sense, but I am NOT optimistic that ARM will integrate Akida into its own automotive chips. Because that’s not really how ARM operates.

ARM licenses CPU and NPU IP (like Ethos), but it doesn’t generally integrate third-party processors into its silicon. That integration is done by OEMs.

I'd say the more realistic path is for Akida to co-exist alongside ARM cores in OEM SoCs or as a host-attached accelerator. If that happens at scale, that would still be a big win but it would be different from an ARM-led integration.

Happy to be corrected, but this is consistent with how ARM’s business model and ecosystem usually work.

 

[Jacob Kiraz]

Just because it’s red today, doesn't mean it’s red for everyone…. You know what I mean mkey?

Its coming for you

 

[HopalongPetrovski]

There it is. Like clockwork.

 

[Jacob Kiraz]

Not so sure, with a current quarterly spend of about $4-5m and a bank balance of over 30m, that gives a runway of about 6 quarters. Considering revenue is starting to come in, I would assume this runway will actually be much longer and as soon as revenue hits $4-5m a quarter, we are profitable… DYOR. All guesswork here

Revenue is coming Mr Heir said so

 

[Jacob Kiraz]

I always laugh at the AGM 's the question always comes up when are we expecting dividends,
I feel what has come out of it at this stage the company wasted time going down the path of IP licenses,
Probably cost us alot of ground

 

[Guzzi62]

The 4c was all about AKD1500, what's going on with Pico and AKD2000?

Key highlight that caught my eyes:
Ongoing collaboration with multiple high-quality companies demonstrating strong
commercial applicability of BrainChip’s technology.

Yes Nex Novus is as mentioned a new one and not PR'ed?

Can't say that I am too impressed with the 4c, I have hope for an announcement prior to the 4c, supporting the statement: watch us now but no!

So we wait!

 

[Bravo]

So we now have confirmed orders from Parsons and Nex Novus totaling 11,200 chips.

Based on the previous announcement below, Akida 1500 pricing ranges from roughly $4 to $50 per chip, depending on volume.

What do people consider a “small” versus “large” volume order in this context, and where do you think a 10,00 unit order (Parsons) and a 1,200 unit order (Nex Novus) realistically sit on that pricing spectrum?

 

[itsol4605]

Great conversation with Pete Bernard at EDGE AI FOUNDATION about where the industry really is today. Whether in the data center or at the edge, progress is not about throwing more power at the… | Sean Hehir

Great conversation with Pete Bernard at EDGE AI FOUNDATION about where the industry really is today. Whether in the data center or at the edge, progress is not about throwing more power at the problem. It is about using the right architecture for the right job. At BrainChip, this is exactly...

www.linkedin.com

 

[Jacob Kiraz]

So we now have confirmed orders from Parsons and Nex Novus totaling 11,200 chips.

Based on the previous announcement below, Akida 1500 pricing ranges from roughly $4 to $50 per chip, depending on volume.

What do people consider a “small” versus “large” volume order in this context, and where do you think a 10,00 unit order and a 1,200 unit order realistically sit on that pricing spectrum?



View attachment 94629

I dont want to doubt but i wouldnt be surprised if the company did it as a discount with a hope a contract will be signed

 

[perceptron]

So we now have confirmed orders from Parsons and Nex Novus totaling 11,200 chips.

Based on the previous announcement below, Akida 1500 pricing ranges from roughly $4 to $50 per chip, depending on volume.

What do people consider a “small” versus “large” volume order in this context, and where do you think a 10,00 unit order and a 1,200 unit order realistically sit on that pricing spectrum?



View attachment 94629

Are you asking for the total cost price to produce those chips? Further, there is no explicit reference to customer price range.

 

[Bravo]

Are you asking for the total cost price to produce those chips? Further, there is no explicit reference to customer price range.

My post was really just a roundabout way of saying that we don’t actually know what pricing applies to these orders, because the announcement doesn’t define what BrainChip considers a small versus volume order.

Without that context, it’s hard to infer the revenue impact from unit numbers alone.

 

[Diogenese]

o

" The AKD1500 integrates seamlessly with x86, ARM, and RISC-V host microcontrollers and multiprocessor (MPUs) chips via PCIe or Quad SPI serial interfaces, supporting broad market adoption across a wide range of applications. The AKD1500 co-processor approach is ideal for a wide range of environments and industries, upgrading multi-processor SoCs within defence, industrial and enterprise settings, and upgrading embedded microcontrollers for AI solutions in healthcare, wearables, and consumer electronics without a complete system redesign."
My bold above.
That’s a compelling hybrid model: ARM for control + neuromorphic for inference.
Essentially we are just waiting for industry to demand the benefits of Neuromorphic and ARM is ready. At the CES26 the 'edge' was all the rage and it appears that we are now seeing the beginning to the Edge take off - finally.
No doubt ARM will integrate AKIDA into their Auto suitable chips - likely Gen 2?
The Quarterly was ok - shows we are making progress.
.

Hi Manny,

I think that Akida 1500 and Akida 2 will give customers a price option. Akida 2 will have a lot more transistors than 1500 because it is 8-bit compared with 4-bit. On a chips per wafer basis, this makes Akida 2 ~ 4 times as expensive to manufacture as a stand alone chip (chiplet?). Of course, where it (Akida) is included with a CPU in a SoC with an associated processor, as in the ARM case you foreshadowed, the difference will be substantially reduced. because the CPU will have a larger silicon footprint Similarly, Akida 3 will be more expensive to manufacture than Akida 2 due to the step up to 16/32 bits. The added transistors go to increased precision, a trade-off against power consumption.

1500 will be good enuf for a lot of applications, and, if someone is going to take a licence, they will be making millions of chips, which means that a smallish price difference per chip could be significant,

So I guess the licencee will have the choice depending on the level of precision and power consumption.

 

[perceptron]

My post was really just a roundabout way of saying that we don’t actually know what pricing applies to these orders, because the announcement doesn’t define what BrainChip considers a small versus volume order.

Without that context, it’s hard to infer the revenue impact from unit numbers alone.

Do you believe the price/chip range refers to the price to customers? If that is the case, by your logic the foundry is only charging the variable costs to produce the chips.

 

[Bravo]

Do you believe the price/chip range refers to the price to customers? If that is the case, by your logic the foundry is only charging the variable costs to produce the chips.

Not sure what you're asking but the announcement shows the variable cost per chip = $2.94 and the price for the customer = $4 - $50 a chip depending on volume.

My point was we don't know what revenue we will achieve because there is no definition of what volume is required for the order to reach a $4 as opposed to a $50 price tag, for example.

 

[manny100]

Hi Manny,

I agree the ARM + accelerator hybrid model makes sense, but I am NOT optimistic that ARM will integrate Akida into its own automotive chips. Because that’s not really how ARM operates.

ARM licenses CPU and NPU IP (like Ethos), but it doesn’t generally integrate third-party processors into its silicon. That integration is done by OEMs.

I'd say the more realistic path is for Akida to co-exist alongside ARM cores in OEM SoCs or as a host-attached accelerator. If that happens at scale, that would still be a big win but it would be different from an ARM-led integration.

Happy to be corrected, but this is consistent with how ARM’s business model and ecosystem usually work.

Hi Bravo, You are right i used the word 'integrate' by habit.
ARM does not integrate AKIDA. I should have said compatible as a co processor.
" BrainChip’s Akida IP is fully compatible with Arm’s product families. System-level pre-integration accelerates evaluation and development. BrainChip hardware acceleration supports real-time multi-sensor inference, and frees up the Arm CPU for improved system-level performance and unparalleled efficiency."
My bold above.
I note that ARM actually says that Brainchip improves ARM so that is a big plus for us.

https://armkeil.blob.core.windows.net/developer/Files/pdf/ai-ecosystem-catalogue/BrainChip_Arm-AI-Partner-Solution-Brief-Template-1.pdf

 

[Diogenese]

So we now have confirmed orders from Parsons and Nex Novus totaling 11,200 chips.

Based on the previous announcement below, Akida 1500 pricing ranges from roughly $4 to $50 per chip, depending on volume.

What do people consider a “small” versus “large” volume order in this context, and where do you think a 10,00 unit order and a 1,200 unit order realistically sit on that pricing spectrum?



View attachment 94629

Hi Bravo,

I guess the manufacturing cost of a chip is based on how many you can fit on a wafer and the number of wafers manufactured in one production run. Then there is the question of where the break even point is between buying the chips v taking an IP licence.

I don't know how many 1500s we get to a wafer, and this, of course, depends on the size chosen. Frontgrade has chosen 28 nm for radhardness, and this will be in a SoC with a CPU (RISC-V).

 

[perceptron]

Not sure what you're asking but the announcement shows the variable cost per chip = $2.94 and the price for the customer = $4 - $50 a chip depending on volume.

My point was we don't know what revenue we will achieve because there is no definition of what volume is required for the order to reach a $4 as opposed to a $50 price tag, for example.

There is no wording on that slide that says price to customers. Could it be price per chip to manufacture volume dependent? Find out what variable costs are.

 

[Bravo]

The 4c was all about AKD1500, what's going on with Pico and AKD2000?

Key highlight that caught my eyes:
Ongoing collaboration with multiple high-quality companies demonstrating strong
commercial applicability of BrainChip’s technology.

Yes Nex Novus is as mentioned a new one and not PR'ed?

Can't say that I am too impressed with the 4c, I have hope for an announcement prior to the 4c, supporting the statement: watch us now but no!

So we wait!

Hi @Guzzi6,

Agreed.

My worry is that AKD2000 is clearly the real volume play, while AKD1500 feels more like a stepping stone.

The company’s own numbers kind for AKD1500 of say that out loud - only 73,000 units in 2027 and another 450,000 by 2028. That not really breakout volume by chip standards.

I get the need to execute cleanly and not trip over ourselves. But I believe there’s also a risk in being too cautious. If we wait too long to pull the higher-volume AKD2000 forward, we risk burning through cash, facing further dilution, and potentially missing the window for the product with the highest volume potential.

These are my non-expert thoughts, for all they're worth.

 

[7für7]

Its coming for you

Relax dude… good luck and hope you didn’t bet your house into this stock! Igno

 

[manny100]

Hi @Guzzi6,

Agreed.

My worry is that AKD2000 is clearly the real volume play, while AKD1500 feels more like a stepping stone.

The company’s own numbers kind for AKD1500 of say that out loud - only 73,000 units in 2027 and another 450,000 by 2028. That not really breakout volume by chip standards.

I get the need to execute cleanly and not trip over ourselves. But I believe there’s also a risk in being too cautious. If we wait too long to pull the higher-volume AKD2000 forward, we risk burning through cash, facing further dilution, and potentially missing the window for the product with the highest volume potential.

These are my non-expert thoughts, for all they're worth.

Yep, i guess ARM will respond to their customer needs. We have 1500 available soon and Gen 2 planned to be taped out this Quarter