BRN Discussion Ongoing

[manny100]

Here here

Agree, i liken it to a meeting in a telephone box.

 

[Guzzi62]

The Shift to Custom Silicon: Why Companies Are Designing Their Own Chips​

May 20, 2025 — 03:38 am EDT

Written by Hedder->



To receive tech updates in your inbox, sign up to the newsletter ARPU.
Xiaomi, the Chinese smartphone giant, recently announced its self-developed mobile chipset, XRING 01, reportedly built on TSMC’s advanced 3nm process and aiming to rival performance levels seen in top-tier chips like Qualcomm’s Snapdragon 8 Gen 2. The move positions Xiaomi as only the fourth global smartphone brand, after Apple, Samsung, and Huawei, to develop its own core mobile processor. This isn’t just news for the smartphone market; it’s the latest example of a growing trend across the tech industry: companies increasingly designing their own custom-tailored silicon rather than relying solely on standardized components.

What is “Custom Silicon” and why are companies pursuing it?​

Custom silicon refers to semiconductor chips designed by a company specifically for its own internal use or products, rather than buying standard chips from a general-purpose supplier like Intel, AMD, or Qualcomm. This trend, once limited to a few tech giants like Apple, is now expanding rapidly. Companies are choosing to design their own chips for several key reasons:

• Performance and Efficiency: Custom chips can be optimized precisely for a company’s specific software, algorithms (especially for AI workloads), and product needs, often leading to significant gains in performance, power efficiency (performance per watt), and form factor integration compared to general-purpose chips.
• Cost Control: By designing their own chips, companies can potentially reduce per-unit costs over time by cutting out the profit margins of traditional chip vendors, especially for high-volume applications.
• Differentiation and Competitive Advantage: Unique chip capabilities can provide a distinct edge in product features or service delivery, making offerings stand out in crowded markets.
• Supply Chain Security and Control: Designing in-house provides greater control over the chip roadmap, supply chain dynamics, and reduces dependence on external suppliers, mitigating risks of shortages or geopolitical restrictions on specific components.

Who are the major players driving the custom silicon trend?​

The trend spans various parts of the tech landscape.

• Consumer Device Makers: Apple is a prime example, having successfully transitioned its Macs to Apple Silicon (M-series chips) leveraging its long-standing expertise in designing chips for iPhones (A-series chips) and integrating specialized “Neural Processors” for on-device AI. Samsung designs its own Exynos processors for some of its devices. Xiaomi is now explicitly joining this group with XRING 01, dedicating a 1,000-person team to the effort. Huawei also designs its own Kirin and Ascend chips for its devices and systems, though its ability to manufacture them is now significantly constrained by export controls.
• Cloud and Data Center Operators (Hyperscalers): Some of the largest consumers of chips, including Google (TPUs since 2016), Amazon (Trainium and Inferentia chips), Microsoft, and Meta, are investing billions in developing custom silicon for their vast data center infrastructure. This allows them to optimize chips specifically for the AI training and inference workloads that are driving massive data center expansion. Amazon is reportedly building out data centers featuring billions of dollars of its internal chips, including a cluster for Anthropic.
• Emerging Hardware Innovators: While not designing chips for their own products in the traditional sense, companies like Cerebras (wafer-scale chips) and Groq (TPU architecture) also represent a form of custom silicon designed specifically for AI workloads, offering alternative architectures to traditional GPUs and targeting performance/efficiency advantages.

How does this trend impact traditional chip suppliers?​

The rise of custom silicon directly challenges traditional chip suppliers, particularly those who primarily sell standardized components or designs. Fabless companies like Qualcomm, historically a dominant player in mobile SoCs, face competition from customers like Apple, Samsung, and now Xiaomi designing their own chips that might otherwise use Snapdragon processors. NVIDIA, while still dominant in the overall AI chip market, faces potential headwinds from hyperscalers replacing some of its high-margin GPUs with their own custom AI accelerators optimized for their specific data center needs. While this doesn’t mean the end for these companies, it forces them to adapt, focus on segments less targeted by custom designs, or offer their IP/services in new ways.

What role do foundries like TSMC and Arm play in enabling custom silicon?​

The success of the custom silicon trend is heavily reliant on the capabilities of third-party manufacturers, particularly advanced foundries and intellectual property (IP) providers. The “pure-play” foundry model, pioneered by TSMC, allows companies to focus entirely on chip design without the prohibitively high cost and complexity of building and operating their own fabrication plants. Companies like Apple, the hyperscalers, and now Xiaomi can leverage TSMC’s cutting-edge process nodes (like 3nm and the upcoming 2nm) to turn their complex designs into physical chips.

IP providers like Arm are also critical. Many custom chips, including Xiaomi’s XRING 01, are based on licensed core architectures (like Arm’s Cortex-X925 CPU and Immortalis-G925 GPU). This allows companies to build sophisticated custom chips without having to design every component from scratch, dramatically accelerating development. Thus, while custom silicon shifts power dynamics away from traditional chip vendors, it often strengthens the position of cutting-edge foundries like TSMC and key IP providers like Arm by increasing demand for their fundamental services.

Are there export control issues for companies designing chips in China?​

The Xiaomi development clarifies a key point regarding US export controls: current restrictions primarily target advanced chips and equipment critical for AI and military modernization, not generally consumer-grade chips. This is why a Chinese company like Xiaomi can reportedly design a high-end mobile SoC (XRING 01) and have it manufactured by TSMC on a 3nm process, even though TSMC is based in Taiwan and uses US technology in its fabs. While companies like Huawei have faced severe restrictions limiting their access to advanced manufacturing for any of their chip designs, including mobile ones, the XRING 01 case demonstrates that for other Chinese companies focused on the consumer market, access to global foundries for advanced nodes remains possible within the current regulatory framework. This illustrates the nuanced nature of the export controls, which are targeted at specific technologies and end-uses rather than a blanket ban on all advanced chip production for China.

What’s next for the custom silicon trend?​

The trend is likely to continue as AI permeates more products and services, and as companies seek competitive advantages through hardware optimization. More companies across various sectors — from automotive to industrial equipment — may explore designing specialized chips.

While building in-house design teams and navigating manufacturing relationships remains challenging, the proven benefits in performance, efficiency, and strategic control suggest that custom silicon will play an increasingly important role in the technology landscape. This will in turn challenge traditional business models and further solidify the critical position of advanced foundries.



The views and opinions expressed herein are the views and opinions of the author and do not necessarily reflect those of Nasdaq, Inc.

https://www.nasdaq.com/articles/shift-custom-silicon-why-companies-are-designing-their-own-chips

 

[Bravo]

Quite a few have asked the same question years ago.

How many companies can and will design their own chips, have them tapped out and then made?

Not many, only the big heavy players that so far haven't signed any IP deals with BRN have the resources for such an undertaking.

Most posts about Akida on LinkedIn are about AK1000 & 1500, nothing so far on AK2/TENNs.

I really really hope that some of the big chipmakers will sign a deal and then design/make a chip that incorporates AK2 in their own design, but hinting they might deploy AK2 themselves is likely due to slow uptake and there is a market for selling chips to smaller players, just like the 1500.

If they can't sign any IP deals before the next AGM, Sean's business plan has failed. He does seem very confident and upbeat at the last interviews we have seen, so let's hope It's happing very soon!

I can’t help feeling a bit disenchanted.

The fact that the company is now allocating SPP funds to continue Akida 2 development and prepare for potential silicon deployment suggests they’re no longer concerned about “competing with customers,” which was the original reason Sean gave for not taping it out themselves.

I realise this doesn’t necessarily mean there’s zero interest - the neuromorphic market is still nascent, and many OEMs prefer to see actual silicon before committing. But it does seem to imply that no customer has taken up a licence strong enough to move Akida 2 into production.

It makes me wonder whether this was a significant misstep, and whether the path to Akida 2 commercialisation has now become unnecessarily protracted as a result.

I genuinely hope you’re right and that there is a solid basis for Sean’s upbeat comments about expected revenue in his recent interview.

But as far as I can tell, revenue can only come from new licences being signed. There are no royalties yet (at least none that we’re aware of), and the first revenue from Akida 1500 won’t arrive until the initial batch of 73,000 chips hits the market in 2027, unless I'm either mistaken or missing something.

Don't mean to be a Debbie Downer, but my optimism appears to be on annual leave at the moment...

 

[jtardif999]

China have just developed a nuclear coin size battery that lasts for 50 years for power restraint products.

China's
Betavolt BV100 is a coin-sized nuclear battery that uses radioactive Nickel-63 to generate electricity for up to 50 years without recharging. It works by converting the energy from beta particles into an electrical current using diamond semiconductor layers. While the current version outputs 100 microwatts, future versions aim for higher power, potentially for devices like medical implants, aerospace systems, and sensors.

Imagine if children accidentally swallowed nuclear batteries … no thanks!

 

[gilti]

Time	Price $	Volume	Value $	Market	Condition
2:38:22 PM​	0.1725​	5​	0.8625​	ASX	CXXT
2:38:19 PM​	0.1725​	1​	0.1725​	ASX	CXXT
2:38:19 PM​	0.1725​	27​	4.6575​	ASX	CXXT
2:38:09 PM​	0.1725​	1​	0.1725​	ASX	CXXT
2:38:09 PM​	0.1725​	27​	4.6575​	ASX	CXXT
2:38:00 PM​	0.1725​	27​	4.6575​	ASX	CXXT
2:37:50 PM​	0.1725​	1​	0.1725​	ASX	CXXT
2:37:50 PM​	0.1725​	27​	4.6575​	ASX	CXXT
2:37:40 PM​	0.1725​	1​	0.1725​	ASX	CXXT
2:37:40 PM​	0.1725​	27​	4.6575​	ASX	CXXT
2:37:31 PM​	0.1725​	27​	4.6575​	ASX	CXXT
2:37:21 PM​	0.1725​	2​	0.345​	ASX	CXXT
2:37:21 PM​	0.1725​	27​	4.6575​	ASX	CXXT
2:37:07 PM​	0.1725​	1​	0.1725​	ASX	CXXT
2:37:07 PM​	0.1725​	27​	4.6575​	ASX	CXXT
2:37:07 PM​	0.1725​	5​	0.8625​	ASX	CXXT
2:37:02 PM​	0.1725​	1​	0.1725​	ASX	CXXT
2:37:02 PM​	0.1725​	27​	4.6575​	ASX	CXXT
2:36:52 PM​	0.1725​	28​	4.83​	ASX	CXXT
2:36:33 PM​	0.1725​	1​	0.1725​	ASX	CXXT
2:36:33 PM​	0.1725​	27​	4.6575​	ASX	CXXT
2:36:24 PM​	0.1725​	1​	0.1725​	ASX	CXXT
2:36:24 PM​	0.1725​	27​	4.6575​	ASX	CXXT
2:36:14 PM​	0.1725​	27​	4.6575​	ASX	CXXT
2:36:05 PM​	0.1725​	1​	0.1725​	ASX	CXXT
2:36:05 PM​	0.1725​	27​	4.6575​	ASX	CXXT
2:35:55 PM​	0.1725​	1​	0.1725​	ASX	CXXT
2:35:55 PM​	0.1725​	27​	4.6575​	ASX	CXXT
2:35:51 PM​	0.1725​	5​	0.8625​	ASX	CXXT
2:35:46 PM​	0.1725​	1​	0.1725​	ASX	CXXT
2:35:46 PM​	0.1725​	27​	4.6575​	ASX	CXXT
2:35:36 PM​	0.1725​	27​	4.6575​	ASX	CXXT
2:35:27 PM​	0.1725​	1​	0.1725​	ASX	CXXT
2:35:27 PM​	0.1725​	27​	4.6575​	ASX	CXXT
2:35:17 PM​	0.1725​	1​	0.1725​	ASX	CXXT
2:35:17 PM​	0.1725​	27​	4.6575​	ASX	CXXT
2:35:08 PM​	0.1725​	27​	4.6575​	ASX	CXXT
2:34:57 PM​	0.1725​	6​	1.035​	ASX	CXXT
2:34:57 PM​	0.1725​	28​	4.83​	ASX	CXXT
2:34:57 PM​	0.1725​	148​	25.53​	ASX	CXXT
2:34:57 PM​	0.17​	5388​	915.96​	ASX
2:34:44 PM​	0.1725​	36​	6.21​	ASX	CXXT
2:34:43 PM​	0.1725​	2​	0.345​	ASX	CXXT
2:34:43 PM​	0.1725​	27​	4.6575​	ASX	CXXT
2:34:43 PM​	0.1725​	5​	0.8625​	ASX	CXXT
2:34:38 PM​	0.1725​	1​	0.1725​	ASX	CXXT
2:34:38 PM​	0.1725​	27​	4.6575​	ASX	CXXT
2:34:29 PM​	0.1725​	1​	0.1725​	ASX	CXXT
2:34:29 PM​	0.1725​	27​	4.6575​	ASX	CXXT
2:34:19 PM​	0.1725​	27​	4.6575​	ASX	CXXT
2:34:04 PM​	0.1725​	1​	0.1725​	ASX	CXXT
2:34:04 PM​	0.1725​	27​	4.6575​	ASX	CXXT

manipulation, why hush your mouth
this is just "to provide liquidity"
say the same pricks that are manipulating the sp on a daily basis

 

[Space Cadet]

Dated August
Quantum Ventura, Penn State, University of Mitchigan

https://arxiv.org/html/2508.04610v1

 

[gilti]

its got even worse in the last 5 minutes
cockroaches is too nice a word.

Time	Price $	Volume	Value $	Market	Condition
2:46:41 PM​	0.1725​	27​	4.6575​	CXA
2:46:32 PM​	0.1725​	27​	4.6575​	CXA
2:46:21 PM​	0.1725​	27​	4.6575​	CXA
2:46:21 PM​	0.1725​	21​	3.6225​	CXA
2:46:21 PM​	0.1725​	27​	4.6575​	CXA
2:46:21 PM​	0.1725​	27​	4.6575​	CXA
2:46:21 PM​	0.1725​	27​	4.6575​	CXA
2:46:21 PM​	0.1725​	27​	4.6575​	CXA
2:46:21 PM​	0.1725​	27​	4.6575​	CXA
2:46:21 PM​	0.1725​	27​	4.6575​	CXA
2:46:21 PM​	0.1725​	27​	4.6575​	CXA
2:46:21 PM​	0.1725​	27​	4.6575​	CXA
2:46:21 PM​	0.1725​	27​	4.6575​	CXA
2:46:21 PM​	0.1725​	27​	4.6575​	CXA
2:46:21 PM​	0.1725​	7​	1.2075​	CXA
2:46:09 PM​	0.1725​	986​	170.085​	ASX	CXXT
2:46:07 PM​	0.175​	5​	0.875​	ASX
2:45:53 PM​	0.1725​	5069​	874.4025​	ASX	CXXT
2:45:19 PM​	0.1725​	986​	170.085​	ASX	CXXT
2:45:09 PM​	0.1725​	986​	170.085​	ASX	CXXT
2:45:01 PM​	0.175​	5​	0.875​	ASX
2:44:22 PM​	0.1725​	986​	170.085​	ASX	CXXT
2:44:22 PM​	0.1725​	986​	170.085​	ASX	CXXT
2:44:22 PM​	0.1725​	3893​	671.5425​	ASX	CXXT
2:44:18 PM​	0.1725​	23​	3.9675​	ASX	CXXT
2:44:18 PM​	0.1725​	3634​	626.865​	ASX	CXXT
2:44:18 PM​	0.1725​	7500​	1293.75​	ASX	CXXT
2:44:15 PM​	0.175​	9118​	1595.65​	CXA
2:44:15 PM​	0.175​	1361​	238.175​	CXA
2:44:13 PM​	0.1725​	1​	0.1725​	ASX	CXXT
2:44:13 PM​	0.1725​	26​	4.485​	ASX	CXXT
2:44:13 PM​	0.1725​	2​	0.345​	ASX	CXXT
2:44:13 PM​	0.1725​	27​	4.6575​	ASX	CXXT
2:44:13 PM​	0.1725​	544​	93.84​	ASX	CXXT
2:44:13 PM​	0.1725​	534​	92.115​	ASX	CXXT
2:44:13 PM​	0.1725​	45621​	7869.623​	ASX	CXXT
2:44:13 PM​	0.1725​	51000​	8797.5​	CXA
2:44:13 PM​	0.1725​	2845​	490.7625​	CXA
2:44:02 PM​	0.1725​	5​	0.8625​	ASX	CXXT
2:44:02 PM​	0.1725​	1​	0.1725​	ASX	CXXT
2:44:02 PM​	0.1725​	27​	4.6575​	ASX	CXXT
2:43:52 PM​	0.1725​	1​	0.1725​	ASX	CXXT
2:43:52 PM​	0.1725​	28​	4.83​	ASX	CXXT
2:43:43 PM​	0.1725​	5​	0.8625​	ASX	CXXT
2:43:43 PM​	0.1725​	27​	4.6575​	ASX	CXXT
2:43:42 PM​	0.1725​	155​	26.7375​	ASX	CXXT
2:43:42 PM​	0.17​	5628​	956.76​	ASX
2:43:33 PM​	0.1725​	27​	4.6575​	ASX	CXXT
2:43:24 PM​	0.1725​	1​	0.1725​	ASX	CXXT
2:43:24 PM​	0.1725​	27​	4.6575​	ASX	CXXT
2:43:17 PM​	0.1725​	3000​	517.5​	CXA
2:43:15 PM​	0.1725​	1​	0.1725​	ASX	CXXT
2:43:15 PM​	0.1725​	27​	4.6575​	ASX	CXXT
2:43:05 PM​	0.1725​	5​	0.8625​	ASX	CXXT
2:43:05 PM​	0.1725​	1​	0.1725​	ASX	CXXT
2:43:05 PM​	0.1725​	27​	4.6575​	ASX	CXXT
2:42:56 PM​	0.1725​	27​	4.6575​	ASX	CXXT
2:42:47 PM​	0.1725​	1​	0.1725​	ASX	CXXT
2:42:47 PM​	0.1725​	27​	4.6575​	ASX	CXXT
2:42:38 PM​	0.1725​	1​	0.1725​	ASX	CXXT
2:42:38 PM​	0.1725​	27​	4.6575​	ASX	CXXT
2:42:29 PM​	0.1725​	1​	0.1725​	ASX	CXXT
2:42:28 PM​	0.1725​	27​	4.6575​	ASX	CXXT
2:42:10 PM​	0.1725​	3​	0.5175​	ASX	CXXT
2:42:10 PM​	0.1725​	28​	4.83​	ASX	CXXT
2:42:00 PM​	0.1725​	27​	4.6575​	ASX	CXXT
2:42:00 PM​	0.1725​	5​	0.8625​	ASX	CXXT
2:41:56 PM​	0.1725​	44​	7.59​	CXA
2:41:56 PM​	0.1725​	59​	10.1775​	ASX	CXXT
2:41:56 PM​	0.1725​	58​	10.005​	ASX	CXXT
2:41:56 PM​	0.17​	2093​	355.81​	ASX
2:41:56 PM​	0.17​	2093​	355.81​	ASX
2:41:56 PM​	0.1725​	60000​	10350​	CXA
2:41:51 PM​	0.1725​	27​	4.6575​	ASX	CXXT
2:41:41 PM​	0.1725​	1​	0.1725​	ASX	CXXT
2:41:41 PM​	0.1725​	27​	4.6575​	ASX	CXXT
2:41:31 PM​	0.1725​	1​	0.1725​	ASX	CXXT
2:41:31 PM​	0.1725​	27​	4.6575​	ASX	CXXT
2:41:20 PM​	0.1725​	30​	5.175​	ASX	CXXT
2:41:19 PM​	0.1725​	27​	4.6575​	ASX	CXXT
2:41:19 PM​	0.17​	880​	149.6​	ASX
2:41:19 PM​	0.1725​	148​	25.53​	ASX	CXXT
2:41:19 PM​	0.17​	5380​	914.6​	ASX
2:41:12 PM​	0.1725​	1​	0.1725​	ASX	CXXT
2:41:12 PM​	0.1725​	27​	4.6575​	ASX	CXXT
2:41:02 PM​	0.1725​	1​	0.1725​	ASX	CXXT
2:41:02 PM​	0.1725​	27​	4.6575​	ASX	CXXT
2:41:02 PM​	0.1725​	5​	0.8625​	ASX	CXXT
2:40:52 PM​	0.1725​	1​	0.1725​	ASX	CXXT
2:40:52 PM​	0.1725​	27​	4.6575​	ASX	CXXT
2:40:43 PM​	0.1725​	24​	4.14​	ASX	CXXT
2:40:43 PM​	0.17​	880​	149.6​	ASX
2:40:42 PM​	0.1725​	1​	0.1725​	ASX	CXXT
2:40:42 PM​	0.1725​	27​	4.6575​	ASX	CXXT
2:40:32 PM​	0.1725​	1​	0.1725​	ASX	CXXT
2:40:32 PM​	0.1725​	27​	4.6575​	ASX	CXXT
2:40:23 PM​	0.1725​	28​	4.83​	ASX	CXXT
2:40:14 PM​	0.1725​	1​	0.1725​	ASX	CXXT
2:40:14 PM​	0.1725​	27​	4.6575​	ASX	CXXT
2:40:04 PM​	0.1725​	1​	0.1725​	ASX	CXXT
2:40:04 PM​	0.1725​	27​	4.6575​	ASX	CXXT
2:39:55 PM​	0.1725​	27​	4.6575​	ASX	CXXT
2:39:39 PM​	0.1725​	2​	0.345​	ASX	CXXT
2:39:39 PM​	0.1725​	27​	4.6575​	ASX	CXXT
2:39:39 PM​	0.1725​	5​	0.8625​	ASX	CXXT
2:39:36 PM​	0.1725​	1​	0.1725​	ASX	CXXT
2:39:36 PM​	0.1725​	27​	4.6575​	ASX	CXXT
2:39:17 PM​	0.1725​	1​	0.1725​	ASX	CXXT
2:39:17 PM​	0.1725​	27​	4.6575​	ASX	CXXT
2:39:07 PM​	0.1725​	27​	4.6575​	ASX	CXXT
2:38:58 PM​	0.1725​	1​	0.1725​	ASX	CXXT
2:38:58 PM​	0.1725​	27​	4.6575​	ASX	CXXT
2:38:48 PM​	0.1725​	1​	0.1725​	ASX	CXXT
2:38:48 PM​	0.1725​	27​	4.6575​	ASX	CXXT
2:38:38 PM​	0.1725​	28​	4.83​	ASX	CXXT
2:38:29 PM​	0.1725​	1​	0.1725​	ASX	CXXT
2:38:29 PM​	0.1725​	27​	4.6575​	ASX	CXXT

 

[gex]

well I topped up, Too cheap if you ask me. But I'm a nobody

 

[Terroni2105]

I can’t help feeling a bit disenchanted.

The fact that the company is now allocating SPP funds to continue Akida 2 development and prepare for potential silicon deployment suggests they’re no longer concerned about “competing with customers,” which was the original reason Sean gave for not taping it out themselves.

I realise this doesn’t necessarily mean there’s zero interest - the neuromorphic market is still nascent, and many OEMs prefer to see actual silicon before committing. But it does seem to imply that no customer has taken up a licence strong enough to move Akida 2 into production.

It makes me wonder whether this was a significant misstep, and whether the path to Akida 2 commercialisation has now become unnecessarily protracted as a result.

I genuinely hope you’re right and that there is a solid basis for Sean’s upbeat comments about expected revenue in his recent interview.

But as far as I can tell, revenue can only come from new licences being signed. There are no royalties yet (at least none that we’re aware of), and the first revenue from Akida 1500 won’t arrive until the initial batch of 73,000 chips hits the market in 2027, unless I'm either mistaken or missing something.

Don't mean to be a Debbie Downer, but my optimism appears to be on annual leave at the moment...

“It makes me wonder whether this was a significant misstep, and whether the path to Akida 2 commercialisation has now become unnecessarily protracted as a result.”

That is % what I believe has happened. Very unfortunate for us long term holders.
I also hold Sean primarily responsible, secondly the board.
Can’t help but feel deflated.
I continue to hold and continue to have faith in the technology, just need a management team up to the mark.

 

[Diogenese]

I can’t help feeling a bit disenchanted.

The fact that the company is now allocating SPP funds to continue Akida 2 development and prepare for potential silicon deployment suggests they’re no longer concerned about “competing with customers,” which was the original reason Sean gave for not taping it out themselves.

I realise this doesn’t necessarily mean there’s zero interest - the neuromorphic market is still nascent, and many OEMs prefer to see actual silicon before committing. But it does seem to imply that no customer has taken up a licence strong enough to move Akida 2 into production.

It makes me wonder whether this was a significant misstep, and whether the path to Akida 2 commercialisation has now become unnecessarily protracted as a result.

I genuinely hope you’re right and that there is a solid basis for Sean’s upbeat comments about expected revenue in his recent interview.

But as far as I can tell, revenue can only come from new licences being signed. There are no royalties yet (at least none that we’re aware of), and the first revenue from Akida 1500 won’t arrive until the initial batch of 73,000 chips hits the market in 2027, unless I'm either mistaken or missing something.

Don't mean to be a Debbie Downer, but my optimism appears to be on annual leave at the moment...

Yes. The delay in revenue is disheartening, but there is a lot of genuine progress ... Frontgrade, RTX, Onsor, ... the recent mention of Akida for cybersecurity, ... (I know: "Show me the money!")

Did I ever mention that I think cybersecurity is the greatest unmet need?

I mistakenly thought 1500 would have TENNs because the new datasheet mentions MACs, but @Guzzi62 pointed out that there is no mention of TENNs. So the real world is yet to experience the full capabilities of TENNs in Akida 2 SoC (ignoring the on-line FPGA which is a pale imitation) ... only 582 more sleeps ...

 

[Bravo]

Yes. The delay in revenue is disheartening, but there is a lot of genuine progress ... Frontgrade, RTX, Onsor, ... the recent mention of Akida for cybersecurity, ... (I know: "Show me the money!")

Did I ever mention that I think cybersecurity is the greatest unmet need?

I mistakenly thought 1500 would have TENNs because the new datasheet mentions MACs, but @Guzzi62 pointed out that there is no mention of TENNs. So the real world is yet to experience the full capabilities of TENNs in Akida 2 SoC (ignoring the on-line FPGA which is a pale imitation) ... only 582 more sleeps ...

582 more sleeps!

OMG! Someone pass me the Grape-Flavoured Emotion Support Fluid (aka wine) ASAP… preferably in a bottle the size of a rain water tank!

 

[Tothemoon24]

It’s based on the same event-based neuromorphic technology found on the earlier AKD1000, relying on spiking neural networks (SNN) to deliver real-time inference in a way that is much more efficient than traditional AI chips. What’s new here is that thanks to PCIe and SPI interfaces, the new ADK1500 co-processor can be paired to a wide range of hosts, ranging from Linux-capable application processors to resource-constrained microcontrollers with x86, Arm, or RISC-V architectures.



BrainChip AKD1500 key features and specifications:

• Akida Neuron Fabric
  • Clocked at 5 to 400 MHz
  • Delivering up to 800 effective GOPS at <1mW/GOPS
  • On-device learning capabilities to enable “secure application personalization, without the need for a Cloud connection or retraining”
• On-Chip Conversion Complex
• Memory/Storage
  • 1MB on-chip local memory
  • SPI D/Q/O memory expansion interface (can also be used for sensors)
• Host interfaces
  • PCIe Gen2 Endpoint Interface
  • SPI S/D/Q/O Peripheral Interface
• Power dissipation – 250 mW typical at 400 MHz
• Dimensions – 7×7 mm MFCTFBGA169 package, 0.5 mm pitch
• Process – GlobalFoundries 22 nm FD-SOI CMOS digital logic process

BrainChip Software development platforms and workflow
The new Edge AI co-processor still relies on BrainChip’s MetaTF software flow, leveraging popular AI frameworks such as TensorFlow/Keras and ONNX/Pytorch. Developers can also make use of MCU SDKs like Edge Impulse or DeGirum, and compile and optimize their chosen models for the AKD1500. There’s some limited documentation (datasheet) available through the developer website(free email registration required).

Potential applications for the AKD1500 include personalized learning Edge AI systems, Edge AI Vision Systems for ADAS/autonomous vehicles, tobots, drones, and video surveillance, industrial IoTfor monitoring, control, and predictive maintenance, as well as Smart Home devices such as Smart Speakers and other voice-controlled appliances.

BrainChip AKD1500 samples are available now, and volume production is scheduled for Q3 2026. Some early customers, namely Parsons, Bascom Hunter, and Onsor Technologies, have already designed AI-enabled sensing applications based on the AKD1500 co-processor for medical and defense-related applications.

Brainchip AKD1500 PCIe/SPI Edge AI co-processor to power battery-operated wearables and IoT devices - CNX Software

BrainChip AKD1500 is an ultra-efficient Edge AI co-processor delivering up to 800 GOPS while consuming just 300 mW of power, making it suitable for

www.cnx-software.com

 

[Jacob Kiraz]

Yes. The delay in revenue is disheartening, but there is a lot of genuine progress ... Frontgrade, RTX, Onsor, ... the recent mention of Akida for cybersecurity, ... (I know: "Show me the money!")

Did I ever mention that I think cybersecurity is the greatest unmet need?

I mistakenly thought 1500 would have TENNs because the new datasheet mentions MACs, but @Guzzi62 pointed out that there is no mention of TENNs. So the real world is yet to experience the full capabilities of TENNs in Akida 2 SoC (ignoring the on-line FPGA which is a pale imitation) ... only 582 more sleeps ...

Question is there any other company out there that can do what akida can do, if so is there a product out that does similar in the .market place today

 

[Diogenese]

“It makes me wonder whether this was a significant misstep, and whether the path to Akida 2 commercialisation has now become unnecessarily protracted as a result.”

That is % what I believe has happened. Very unfortunate for us long term holders.
I also hold Sean primarily responsible, secondly the board.
Can’t help but feel deflated.
I continue to hold and continue to have faith in the technology, just need a management team up to the mark.

Back when we went "IP-only", we took the high road (very few customers), while Synsense took the low road.

At the time (unbeknownst to SHs) TENNs was already in the pipeline, so, although we threw the newborn Akida 1000 out with the bath water, there was some rationale for the decision, albeit based partially on a mistaken undervaluing of Akida 1000. At the time the company was under extreme financial pressure, but still needed to develop the silicon for TENNs.

Nevertheless, the worth of Akida 1000/1500 shone through, and management belatedly (about 2.75 years belatedly) woke up to the market need for 1500. 1500 has the advantage of not requiring the ARM Cortex licence of 1000, freeing it up to be used with any other CPU/GPU including RISC-V for configuration.

There was some hubris born of ignorance about the difficulties of the IP-only model, but I think there was also some ethical concern about keeping faith with EAPs by not wanting to sell them a pup which would be superseded by TENNs. Indeed the EAPs may have been in on the TENNs secret.

So much for the spilt milk ...

We do have genuine customers for 1500.

WE have also expanded our product range to include software/model libraries.

So the short term forecast is overcast, but every cloud has an Akida lining. (Not meteorological advice).

 

[Diogenese]

Question is there any other company out there that can do what akida can do, if so is there a product out that does similar in the .market place today

Hi JK,

Our on-chip learning is a distinguishing feature.

Loihi 2 is still a lab curiosity.

Synsense needs different chips for different applications (audio v video).

There are a few analog proponents, but I believe analog remains fairly low-fi.

There are several other players, but I don't think there are any who can match TENNs.

 

[manny100]

Hi JK,

Our on-chip learning is a distinguishing feature.

Loihi 2 is still a lab curiosity.

Synsense needs different chips for different applications (audio v video).

There are a few analog proponents, but I believe analog remains fairly low-fi.

There are several other players, but I don't think there are any who can match TENNs.

Spot on Dio, there are countless inventions/new products/devices etc emanating from Space and defense and turning into mass adoption. We are fortunate to be just getting into both space and defense.
Our big advantage is adaptable, on chip learning.

 

[Frangipani]

#gsa2025 #shanghai #brainchip #edgeai #neuromorphicai | Jerry Kuo

Super excited to represent Brainchip at GSA International Summit 2025 in Shanghai! Awesome people, edge AI vibes, and epic views at the Bund 🌉⚡ Let’s build smarter Edge AI together! #GSA2025 #Shanghai #Brainchip #EdgeAI #NeuromorphicAI

www.linkedin.com

 

[TECH]

Brainchip AKD1500 PCIe/SPI Edge AI co-processor to power battery-operated wearables and IoT devices - CNX Software

BrainChip AKD1500 is an ultra-efficient Edge AI co-processor delivering up to 800 GOPS while consuming just 300 mW of power, making it suitable for

www.cnx-software.com

Nice article, reprint, but still will be picked up by interested parties worldwide, nothing beats getting your name out there,
awareness beats hiding in a lab with a bong hoping to make it big one day.........

 

[DK6161]

Managed to buy back your shorts then

Too funny, Pom. You are the only reason I keep coming back.

 

[smoothsailing18]

Back when we went "IP-only", we took the high road (very few customers), while Synsense took the low road.

At the time (unbeknownst to SHs) TENNs was already in the pipeline, so, although we threw the newborn Akida 1000 out with the bath water, there was some rationale for the decision, albeit based partially on a mistaken undervaluing of Akida 1000. At the time the company was under extreme financial pressure, but still needed to develop the silicon for TENNs.

Nevertheless, the worth of Akida 1000/1500 shone through, and management belatedly (about 2.75 years belatedly) woke up to the market need for 1500. 1500 has the advantage of not requiring the ARM Cortex licence of 1000, freeing it up to be used with any other CPU/GPU including RISC-V for configuration.

There was some hubris born of ignorance about the difficulties of the IP-only model, but I think there was also some ethical concern about keeping faith with EAPs by not wanting to sell them a pup which would be superseded by TENNs. Indeed the EAPs may have been in on the TENNs secret.

So much for the spilt milk ...

We do have genuine customers for 1500.

WE have also expanded our product range to include software/model libraries.

So the short term forecast is overcast, but every cloud has an Akida lining. (Not meteorological advice).

Csn i ask then, what is you estimate time to blue skys from this short term over cast run way.

One would expect clouds thinning out by Jan 2026 ?