BRN Discussion Ongoing

[Frangipani]

Nandan Nayampally will be one of the speakers at tinyML Asia 2023 in Seoul:

Thursday November 16, 2023
9:00 am to 9:15 am

Opening / Welcome​

9:15 am to 10:00 am

Keynote - TBA​

10:00 am to 10:20 am

Solutions and Applications​

Brainchip’s Specialized Akida Architecture for Low Power Edge Inference with HW Spatiotemporal Acceleration and On-chip Learning​

​

Nandan NAYAMPALLY, CMO, Brainchip
Abstract (English)

Brainchip’s Innovative Akida v2.0 Architecture​

• Event-based, neuromorphic processing
• Extremely efficient computation only processes and communicates on events resulting in minimal latency and power consumption. Hardware accelerates event domain convolutions, temporal convolutions, and vision transformers while consuming very low power for battery-operated devices.
• At-memory compute
• Low-latency and low-power computation via parallel execution of events propagating through a mesh-connected array of Neural Processing Units (NPUs) each with a dedicated SRAM memory.
• Quantized parameters and activations
• Supports 8, 4, 2, 1-bit parameters and activations on layer-by-layer basis for model implementations tailored for optimum accuracy versus size, power, and latency.
• Accelerates Traditional AI models
• The MetaTF software stack quantizes and converts models in Keras/TensorFlow and Pytorch/ONNX to Akida event-driven models.
• Fully-digital Implementation
• Portable across process technology with standard synthesis design flow.
• Plus Novel Architectural Features below …

Select Architectural Highlights​

• Temporal Event-based Neural Networks (TENNs)
• Radically reduces number of parameters and operations by an order of magnitude or more for spatiotemporal computations with 2D streaming input such as video and 1D streaming input such as raw audio. Easily trained with back propagation like a CNN and inferences like a RNN. Extracts 2D spatial and temporal 1D kernels for 3D input and extracts temporal 1D kernels for 1D input.
• Vision Transformers (ViTs)
• Multi-head attention included in hardware implementation of Transformer Encoder block with configurations from 2 to 12 nodes. A two-node configuration at 800 MHz executes TinyViT at 30 fps (224x224x3).
• On-device learning

Brainchip’s patented algorithm for homeostatic Spike Time Dependent Plasticity (STDP) enables on-device learning. AI applications can implement incremental learning so that models can adapt to changes in the field. On-device learning can add new classes to a model while leveraging features learned in its original, off-line training.

Model Development Methodologies​

• No-code Model Development and Deployment with Edge Impulse Users with limited AI expertise can develop and deploy optimized Akida models.
• Build-Your-Own-Model with Brainchip’s MetaTF Experienced AI users can build and optimize custom Akida models.

Results​

• Video Benchmarks
• Audio and Medical Vital Signs Benchmarks
• Image Benchmarks

 

[charles2]

Nandan Nayampally will be one of the speakers at tinyML Asia 2023 in Seoul:


View attachment 44221

Thursday November 16, 2023
9:00 am to 9:15 am

Opening / Welcome​

9:15 am to 10:00 am

Keynote - TBA​

10:00 am to 10:20 am

Solutions and Applications​

Brainchip’s Specialized Akida Architecture for Low Power Edge Inference with HW Spatiotemporal Acceleration and On-chip Learning​

​

Nandan NAYAMPALLY, CMO, Brainchip
Abstract (English)

Brainchip’s Innovative Akida v2.0 Architecture​

• Event-based, neuromorphic processing
• Extremely efficient computation only processes and communicates on events resulting in minimal latency and power consumption. Hardware accelerates event domain convolutions, temporal convolutions, and vision transformers while consuming very low power for battery-operated devices.
• At-memory compute
• Low-latency and low-power computation via parallel execution of events propagating through a mesh-connected array of Neural Processing Units (NPUs) each with a dedicated SRAM memory.
• Quantized parameters and activations
• Supports 8, 4, 2, 1-bit parameters and activations on layer-by-layer basis for model implementations tailored for optimum accuracy versus size, power, and latency.
• Accelerates Traditional AI models
• The MetaTF software stack quantizes and converts models in Keras/TensorFlow and Pytorch/ONNX to Akida event-driven models.
• Fully-digital Implementation
• Portable across process technology with standard synthesis design flow.
• Plus Novel Architectural Features below …

Select Architectural Highlights​

• Temporal Event-based Neural Networks (TENNs)
• Radically reduces number of parameters and operations by an order of magnitude or more for spatiotemporal computations with 2D streaming input such as video and 1D streaming input such as raw audio. Easily trained with back propagation like a CNN and inferences like a RNN. Extracts 2D spatial and temporal 1D kernels for 3D input and extracts temporal 1D kernels for 1D input.
• Vision Transformers (ViTs)
• Multi-head attention included in hardware implementation of Transformer Encoder block with configurations from 2 to 12 nodes. A two-node configuration at 800 MHz executes TinyViT at 30 fps (224x224x3).
• On-device learning

Brainchip’s patented algorithm for homeostatic Spike Time Dependent Plasticity (STDP) enables on-device learning. AI applications can implement incremental learning so that models can adapt to changes in the field. On-device learning can add new classes to a model while leveraging features learned in its original, off-line training.

Model Development Methodologies​

• No-code Model Development and Deployment with Edge Impulse Users with limited AI expertise can develop and deploy optimized Akida models.
• Build-Your-Own-Model with Brainchip’s MetaTF Experienced AI users can build and optimize custom Akida models.

Results​

• Video Benchmarks
• Audio and Medical Vital Signs Benchmarks
• Image Benchmarks

And this presumably after the ARM IPO.

Emphasis on this: (Availability of v2.0)

Brainchip’s Innovative Akida v2.0 Architecture​

 

[Frangipani]

According to Google, it looks as if the below update was uploaded to the Brainchip website two days ago.
Can’t be long now until Akida 2.0 will become generally available (plus there are only three weeks left in Q3/2023 ):

Overview — Akida Examples documentation

doc.brainchipinc.com

Upgrading to Akida 2.0 — Akida Examples documentation

doc.brainchipinc.com

 

[Rise from the ashes]

Whoever's chummy with brn through contact either by phone or email
Can you ask if they are willing to send these guys a reference chip.
I'm fully prepared to pay for the chip and delivery costs upfront.

 

[Cartagena]

I'm sorely tempted to apply for this position but I can't leave Australia because of my cat and my partner. Also I'm not a Software Engineer which may be problematic.

View attachment 44201
View attachment 44202 View attachment 44206

Machine Learning Software Engineer - Tools in Cambridge at Arm

Learn more about and apply for the Machine Learning Software Engineer - Tools job in Cambridge at Arm

careers.arm.com

I see Booze Allen Hamilton seem to think there is something on neuromorphic

Heather Via
Associate at Booz Allen Hamilton
1mo

In comparison to the brain, with artificial intelligence, there is no awareness, just computation. Machine learning is a subset of AI which uses algorithms and data sets to simulate human intelligence. In 2021, investment in AI reached $93.5 billion... and this trend will only continue to grow. Seek out opportunities to explore AI, machine learning and neuromorphic processing now - failing to do so inevitably means falling behind.


Booz Allen Hamilton
626,929 followers
1mo
There are three questions #AI leaders should consider when thinking about the future of AI ethics. Here they are: https://lnkd.in/egkJMUa2 xx

Watch this clip from Embedded World and talks about person detection using the M85 cortex. Renesas and ARM collaboration and I'm sure Brainchip is part of this

Instant-On AI Leverages Cortex-M85 MCU​

April 17, 2023

Renesas Electronics presented live demos of AI and machine-learning implementations on an MCU based on the Arm Cortex-M85 processor.
Alix Paultre
Related To: Electronic Design

Check out our embedded world 2023 coverage.

Renesas Electronics presented a live demonstration of AI and machine learning implementations on an MCU based on the Arm Cortex-M85 processor, showing how AI/ML applications can be addressed by the new Cortex-M85 core and Arm’s Helium technology.
One example of a demanding AI use case involved a people detection application developed along with Plumerai that identifies and tracks people in the camera frame in varying lighting and environmental conditions. Another was a motorcontrol predictive maintenance use-case with an AI-based unbalanced load detection application based on TensorFlow Lite for Microcontrollers with CMSIS-NN.
The compact and efficient tinyML models used show low-cost and lower-power AI solutions can serve a wide range of IoT implementations. Delivering over 6 CoreMark/MHz, the Arm Cortex-M85 offers Helium technology, Arm’s M-Profile Vector Extension, as part of the Armv8.1M architecture. The demos showcased the performance advantage in AI use cases, highlighting Cortex-M hallmarks such as deterministic operation, short interrupt response time, and low-power support.
Renesas is implementing the new Arm processor technology in its RA (Renesas Advanced) family of MCUs, along with a robust ecosystem of partners providing customers with comprehensive solutions. The Cortex-M85 core supports Arm TrustZone technology for protection of secure assets.
Combined with TrustZone, Renesas’ integrated cryptographic engine, immutable storage, key management, and tamper protection against DPA/SPA side-channel attacks will provide a comprehensive and integrated secure-element functionality. The Armv8-M architecture also brings Pointer Authentication/Branch Target Identification (PAC/BTI) security extension, delivering enhanced mitigation from software attack threats.

RA MCUs based on the Cortex-M85 core will be supported by Renesas’ Flexible Software Package (FSP), which provides all of the infrastructure software needed, including multiple RTOS, BSP, peripheral drivers, middleware, connectivity, networking, and security stacks as well as reference software. It enables customers to integrate their own legacy code and choice of RTOS with FSP, thus providing full flexibility in application development, easing the migration of existing designs to new RA-based devices.

Related links:​

• Renesas Electronics
• RA family of Arm Cortex-M-based MCUs
• More embedded world 2023 coverage

Link for video :

Instant-On AI Leverages Cortex-M85 MCU

Renesas Electronics presented live demos of AI and machine-learning implementations on an MCU based on the Arm Cortex-M85 processor.

www.electronicdesign.com

 

[IloveLamp]

Scalable Computing Infrastructure and Networking for AI and Edge

Computing infrastructure and networking from Arm enable scalable, secure AI and edge solutions with high performance and energy-efficient design.

www.arm.com

 

[Esq.111]

Whoever's chummy with brn through contact either by phone or email
Can you ask if they are willing to send these guys a reference chip.
I'm fully prepared to pay for the chip and delivery costs upfront.

Good Morning Rise from the ashes,

The computer , HAL , will not obey your request for departure.....

.

Appreciate your input.

Think this track sums it up pretty well.....



Regards,
Esq.

 

[Cartagena]

According to Google, it looks as if the below update was uploaded to the Brainchip website two days ago.
Can’t be long now until Akida 2.0 will become generally available (plus there are only three weeks left in Q3/2023 ):

Overview — Akida Examples documentation

doc.brainchipinc.com

View attachment 44223

Upgrading to Akida 2.0 — Akida Examples documentation

doc.brainchipinc.com

View attachment 44224

Hi Frangipani,

This is a use case where I think Akida will be used as an embedded sensor. This sensor detects VOC's and air quality. The applications are enormous.

From my knowledge Akida was being used in another application which detected VOC's I just can't remember which one?


Source: Renesas website

Main navigation​

1. Products arrow_right
2. Sensor Products arrow_right
3. Environmental Sensors arrow_right
4. Digital Gas Sensors arrow_right
5. ZMOD4410

ZMOD4410​

infoActiveSamples Available

Firmware Configurable Indoor Air Quality (IAQ) Sensor with Embedded Artificial Intelligence (AI)​

Order NowDownload Datasheet

Jump to Page Section:arrow_drop_down
Product TreeOpen product tree menu

Overview​

• Description
• Features
• Applications

Description​

The ZMOD4410 gas sensor module is an easily integrated firmware configurable architecture enabling multiple sensing solutions with a common hardware platform. Simply download one of the firmware libraries that best supports your application from the Renesas website and integrate it with your host microcontroller. The JEDEC JESD47 qualified device supports 10+ years of life, supporting your indoor air quality (IAQ) application designed for detecting total volatile organic compounds (TVOCs), estimating CO2, and monitoring indoor air quality in different smell-based use cases, including very humid and dusty applications with the possibility of water spray, condensation or immersion. The miniature 12-LGA assembly consists of a gas sense element and a CMOS signal conditioning IC. The sensor is pre-calibrated with a signal conditioner controlling the sensor temperatures and measures the MOx conductivity, which is a function of the gas concentration. The measurement results can be read via an I2C interface with the user’s microprocessor, which processes the data. The ZMOD4410 is ideal for applications such as air purifiers, IAQ monitors, kitchen hoods, HVAC equipment, building controls, thermostats, smart fans, and other smart devices.
expand_lessRead Less
, Enlarge image in dialog
, Enlarge image in dialog
, Enlarge image in dialog

​

 

[Jchandel]

Watch this clip from Embedded World and talks about person detection using the M85 cortex. Renesas and ARM collaboration and I'm sure Brainchip is part of this

Instant-On AI Leverages Cortex-M85 MCU​

April 17, 2023

Renesas Electronics presented live demos of AI and machine-learning implementations on an MCU based on the Arm Cortex-M85 processor.
Alix Paultre
Related To: Electronic Design

Check out our embedded world 2023 coverage.

Renesas Electronics presented a live demonstration of AI and machine learning implementations on an MCU based on the Arm Cortex-M85 processor, showing how AI/ML applications can be addressed by the new Cortex-M85 core and Arm’s Helium technology.
One example of a demanding AI use case involved a people detection application developed along with Plumerai that identifies and tracks people in the camera frame in varying lighting and environmental conditions. Another was a motorcontrol predictive maintenance use-case with an AI-based unbalanced load detection application based on TensorFlow Lite for Microcontrollers with CMSIS-NN.
The compact and efficient tinyML models used show low-cost and lower-power AI solutions can serve a wide range of IoT implementations. Delivering over 6 CoreMark/MHz, the Arm Cortex-M85 offers Helium technology, Arm’s M-Profile Vector Extension, as part of the Armv8.1M architecture. The demos showcased the performance advantage in AI use cases, highlighting Cortex-M hallmarks such as deterministic operation, short interrupt response time, and low-power support.
Renesas is implementing the new Arm processor technology in its RA (Renesas Advanced) family of MCUs, along with a robust ecosystem of partners providing customers with comprehensive solutions. The Cortex-M85 core supports Arm TrustZone technology for protection of secure assets.
Combined with TrustZone, Renesas’ integrated cryptographic engine, immutable storage, key management, and tamper protection against DPA/SPA side-channel attacks will provide a comprehensive and integrated secure-element functionality. The Armv8-M architecture also brings Pointer Authentication/Branch Target Identification (PAC/BTI) security extension, delivering enhanced mitigation from software attack threats.

RA MCUs based on the Cortex-M85 core will be supported by Renesas’ Flexible Software Package (FSP), which provides all of the infrastructure software needed, including multiple RTOS, BSP, peripheral drivers, middleware, connectivity, networking, and security stacks as well as reference software. It enables customers to integrate their own legacy code and choice of RTOS with FSP, thus providing full flexibility in application development, easing the migration of existing designs to new RA-based devices.

Related links:​

• Renesas Electronics
• RA family of Arm Cortex-M-based MCUs
• More embedded world 2023 coverage

Link for video :

Instant-On AI Leverages Cortex-M85 MCU

Renesas Electronics presented live demos of AI and machine-learning implementations on an MCU based on the Arm Cortex-M85 processor.

www.electronicdesign.com

Related post from Bravo in March this year-
Post in thread 'BRN Discussion Ongoing'
https://thestockexchange.com.au/threads/brn-discussion-ongoing.1/post-259483

 

[Foxdog]

Related post from Bravo in March this year-
Post in thread 'BRN Discussion Ongoing'
https://thestockexchange.com.au/threads/brn-discussion-ongoing.1/post-259483

Ah yes, and if you scroll down a bit from Bravo's post back in March a few are raving about 48c being an absolute bargain. Only 6 months later and look where we are. This has been a difficult year for BRN shareholders, so far.

 

[Cartagena]

Related post from Bravo in March this year-
Post in thread 'BRN Discussion Ongoing'
https://thestockexchange.com.au/threads/brn-discussion-ongoing.1/post-259483

Thanks Jchandel, we are in the right place at the right time. If we can leverage the power of Renesas and ARM, success is guaranteed.
Still cannot believe how we are at 30 cents with all the new potential products that can hit the market in the near future.
The following exerpt from Renesas regarding winning combinations does not mention Akida but it's obvious that it will be incorporated into many use cases including the smart doorbells, AI embedded air sensor, list goes on

These Winning Combinations accelerate time to market for automotive, industrial, infrastructure and IoT applications, enabling billions of connected, intelligent devices that enhance the way people work and live.

 

[Jchandel]

Related post from Bravo in March this year-
Post in thread 'BRN Discussion Ongoing'
https://thestockexchange.com.au/threads/brn-discussion-ongoing.1/post-259483

Actually the original find was made by @IloveLamp . This goes into a bit more detail.

Post in thread 'BRN Discussion Ongoing'
https://thestockexchange.com.au/threads/brn-discussion-ongoing.1/post-70441

 

[Frangipani]

Hi Frangipani,

This is a use case where I think Akida will be used as an embedded sensor. This sensor detects VOC's and air quality. The applications are enormous.

From my knowledge Akida was being used in another application which detected VOC's I just can't remember which one?


Source: Renesas website

Main navigation​

1. Products arrow_right
2. Sensor Products arrow_right
3. Environmental Sensors arrow_right
4. Digital Gas Sensors arrow_right
5. ZMOD4410

ZMOD4410​

infoActiveSamples Available

Firmware Configurable Indoor Air Quality (IAQ) Sensor with Embedded Artificial Intelligence (AI)​

Order NowDownload Datasheet

Jump to Page Section:arrow_drop_down
Product TreeOpen product tree menu

Overview​

• Description
• Features
• Applications

Description​

The ZMOD4410 gas sensor module is an easily integrated firmware configurable architecture enabling multiple sensing solutions with a common hardware platform. Simply download one of the firmware libraries that best supports your application from the Renesas website and integrate it with your host microcontroller. The JEDEC JESD47 qualified device supports 10+ years of life, supporting your indoor air quality (IAQ) application designed for detecting total volatile organic compounds (TVOCs), estimating CO2, and monitoring indoor air quality in different smell-based use cases, including very humid and dusty applications with the possibility of water spray, condensation or immersion. The miniature 12-LGA assembly consists of a gas sense element and a CMOS signal conditioning IC. The sensor is pre-calibrated with a signal conditioner controlling the sensor temperatures and measures the MOx conductivity, which is a function of the gas concentration. The measurement results can be read via an I2C interface with the user’s microprocessor, which processes the data. The ZMOD4410 is ideal for applications such as air purifiers, IAQ monitors, kitchen hoods, HVAC equipment, building controls, thermostats, smart fans, and other smart devices.
expand_lessRead Less
, Enlarge image in dialog
, Enlarge image in dialog
, Enlarge image in dialog

​

Hi Cartagena,

you were probably thinking of Prof. Hossam Haick’s NaNose (see the
Thread 'BrainChip + Nanose' https://thestockexchange.com.au/threads/brainchip-nanose.2346/)?

Besides that confirmed use of Akida (to this day, there is however no commercial availability of the NaNose breathaliser), there has also been some speculation here on TSE about Akida possibly being integrated into various gas sensors such as those by Bosch (eg
https://thestockexchange.com.au/threads/brn-discussion-ongoing.1/post-38166 or https://thestockexchange.com.au/threads/brn-discussion-ongoing.1/post-354750)
or into said Renesas gas sensor (https://thestockexchange.com.au/threads/brn-discussion-ongoing.1/post-273219 and
https://thestockexchange.com.au/threads/brn-discussion-ongoing.1/post-306526), which all happened to be “liked” by Rob Telson. Which may or may not mean anything.

In addition, I posted an update on NASA’s testing of VOC devices a couple of weeks ago
https://thestockexchange.com.au/threads/brn-nasa.4/post-306527 - but again no concrete mention of Akida.

 

[Rise from the ashes]

Good Morning Rise from the ashes,

The computer , HAL , will not obey your request for departure.....

.

Appreciate your input.

Think this track sums it up pretty well.....



Regards,
Esq.

Thank you good sir. Still pondering if I should be filled with the emotion of being appreciated/wanted or in fear of my life

 

[Xhosa12345]

Just stating my opinion of the content of his posting.. I'm waiting on zee to delete my account
Why heck can't people just be nice here

Yep love it how ppl say things online they wouldnt have cohunas to say to ppls faces...

Are we finally going to see a pullback of the shorts? Hopefully!

Go the pies, but we were lucky...

and yes i know i need to charge my phone....

Happy weekend all!

 

[Diogenese]

Hi Frangipani,

Akida 2 has some great performance benchmarks:

"Multi-head attention included in hardware implementation of Transformer Encoder block with configurations from 2 to 12 nodes. A two-node configuration at 800 MHz executes TinyViT at 30 fps (224x224x3)."

I thought ViT was only available on the Akida 2P (8 to 128 nodes), but running 2 nodes (Akida 2E) at 800 MHz, it can process 30 fps on TinyViT. Akida 1 normally runs at 300MHz.

TinyViT is a recent Microsoft developed small image model which is compressed from large models, and may be important in future AGI developments:

https://arxiv.org/pdf/2207.10666.pdf (20220721)

Nandan Nayampally will be one of the speakers at tinyML Asia 2023 in Seoul:


View attachment 44221

Thursday November 16, 2023
9:00 am to 9:15 am

Opening / Welcome​

9:15 am to 10:00 am

Keynote - TBA​

10:00 am to 10:20 am

Solutions and Applications​

Brainchip’s Specialized Akida Architecture for Low Power Edge Inference with HW Spatiotemporal Acceleration and On-chip Learning​

​

Nandan NAYAMPALLY, CMO, Brainchip
Abstract (English)

Brainchip’s Innovative Akida v2.0 Architecture​

• Event-based, neuromorphic processing
• Extremely efficient computation only processes and communicates on events resulting in minimal latency and power consumption. Hardware accelerates event domain convolutions, temporal convolutions, and vision transformers while consuming very low power for battery-operated devices.
• At-memory compute
• Low-latency and low-power computation via parallel execution of events propagating through a mesh-connected array of Neural Processing Units (NPUs) each with a dedicated SRAM memory.
• Quantized parameters and activations
• Supports 8, 4, 2, 1-bit parameters and activations on layer-by-layer basis for model implementations tailored for optimum accuracy versus size, power, and latency.
• Accelerates Traditional AI models
• The MetaTF software stack quantizes and converts models in Keras/TensorFlow and Pytorch/ONNX to Akida event-driven models.
• Fully-digital Implementation
• Portable across process technology with standard synthesis design flow.
• Plus Novel Architectural Features below …

Select Architectural Highlights​

• Temporal Event-based Neural Networks (TENNs)
• Radically reduces number of parameters and operations by an order of magnitude or more for spatiotemporal computations with 2D streaming input such as video and 1D streaming input such as raw audio. Easily trained with back propagation like a CNN and inferences like a RNN. Extracts 2D spatial and temporal 1D kernels for 3D input and extracts temporal 1D kernels for 1D input.
• Vision Transformers (ViTs)
• Multi-head attention included in hardware implementation of Transformer Encoder block with configurations from 2 to 12 nodes. A two-node configuration at 800 MHz executes TinyViT at 30 fps (224x224x3).
• On-device learning

Brainchip’s patented algorithm for homeostatic Spike Time Dependent Plasticity (STDP) enables on-device learning. AI applications can implement incremental learning so that models can adapt to changes in the field. On-device learning can add new classes to a model while leveraging features learned in its original, off-line training.

Model Development Methodologies​

• No-code Model Development and Deployment with Edge Impulse Users with limited AI expertise can develop and deploy optimized Akida models.
• Build-Your-Own-Model with Brainchip’s MetaTF Experienced AI users can build and optimize custom Akida models.

Results​

• Video Benchmarks
• Audio and Medical Vital Signs Benchmarks
• Image Benchmarks

 

[Fastback6666]

Thank you good sir. Still pondering if I should be filled with the emotion of being appreciated/wanted or in fear of my life

Good to see you are still around mate. I would not take anything too personally on any forum especially when a share price of the stock in discussion is being hammered down.

Some people/investors go into that mass fear and anger mode and just start taking it out in individual’s and the company. Some may well be warranted against the company but in the end the person just needs to deal with their anger personally or take up their issues with the company or sell out.

Ideallly no one wants to be invested on stock that has fallen so much. But hey that is the real reality for any company that is pre-royalty revenue, in the end everyone has to know that there needs to be ongoing sustaining revenue shown to maintain a share price. Until we see this quantified in an announcement (partnership with money attached) or see it in the balance sheet then the SP will just move up and down a lot, not a lot we can do apart with raising issues with the company and push for a more comprehensive shareholder communications plan around prospects and revenue.

We have to see the current times as an opportunity if you still believe in what Brainchip have and the extensive partners they are working with. Time in partner development now has been extensive of many many years so you would expect something has to give up some revenue soon in royalties.

While the bulk revenue may not have been as early as hoped for Akida 1000 it seems like it still will happen and more Brainchip products on the market and some driven by specific customer/partner wants and needs is a great sign and magnifies possibilities for revenue.

Partners are clearly still there and interest is still growing on BRN tech and product pipeline, so just remember the 30 cent opportunity. If people here were buying in the past with less partners and less products at 50 cents or 80 cents or higher then….food for thought.

 

[wilzy123]

View attachment 44197

*announces exit then everyday since"

 

[Fullmoonfever]

Wider impact of supply chains and products coming to market.

Can obviously create knock on effect to take up to an extent.

Though, BRN (& Megachips / Renesas) need to ensure they do have some additional deals stitched up in the background given the comment on accelerators....including neuromorphic.

Lead times for generative AI systems extending into 2024

Lead times for systems for generative AI and AI workloads are lengthy to the point where infrastructure ordered today may not be installed until Spring 2024. That's the takeaway from a bevy of earnings conference calls from infrastructure providers. Dell Technologies reported...

www.constellationr.com

SEP 5, 2023

Lead times for generative AI systems extending into 2024​

By Larry Dignan

Lead times for systems for generative AI and AI workloads are lengthy to the point where infrastructure ordered today may not be installed until Spring 2024.
That's the takeaway from a bevy of earnings conference calls from infrastructure providers. Dell

Technologies reported better-than-expected second quarter results and noted that its PowerEdge XE9680 GPU-enabled server is the fastest ramping product it has launched.

Dell's problem: Orders are off the charts, but parts--namely GPUs--are hard to come by even though the company has a strong partnership with Nvidia. Dell Chief Operating Officer Jeff Clarke said the company has about $2 billion of XE9680 orders in backlog with a higher sales pipeline.

Yes, generative AI investment is taking off. But the revolution--and the enterprise use cases on premises and locally--will be delayed due to supply and demand imbalances. Clarke later put a number on the backlog issue (emphasis mine):

"Maybe the easiest measure to determine where we are with supply is demand is way ahead of supply. If you order a product today, it's a 39-week lead time, which would be delivered the last week of May of next year. So, we are certainly asking for more parts working to get more parts. It's what we do. I'm not the allocator, I'm the allocatee. We’re advocating our position on our demand. Again, we are winning business signaled by the $2 billion in backlog today with a pipeline that's significantly bigger I was in the discussion yesterday with two different customers about AI, the day before about AI. It is constantly something that's coming into our business that we're fielding the opportunities. From different cloud providers to folks building AI as a service to enterprises now beginning to do proof of concept and trying to figure out how they do exactly what I just said earlier, use their data on-premises to actually drive AI to improve their business."

A few notable takeaways from Clarke's comments:

• Nvidia has big investments in its supply chain and is moving GPUs at a rapid clip. Hyperscale cloud providers are gobbling up systems.
• Enterprises are looking to train models on-premises too. That migration rhymes with how enterprises are re-evaluating public cloud workloads.
• Supply will improve as Nvidia rivals come online with GPUs and accelerators.
• AI workloads will reshape data center demand.

To that end, Clarke added that supply will catch up.

"We're tracking at least 30 different accelerator chips that are in the pipeline in development that are coming. So, there are many people that see the opportunity. Some of these new technologies are fairly exciting from neuromorphic type of processors to types of accelerators there's a series of new technologies and quite frankly, new algorithms that we think open up the marketplace and we'll obviously be watching that and driving that across our businesses and helping customers."

AMD is the likely beneficiary with its fourth quarter ramp of next-generation GPUs.

This GPU supply issue is also affecting other infrastructure players.
HPE CEO Antonio Neri said on the company's latest earnings call:

"We start now shipping some of those orders, those wins, but it's a long way to go. And remember that there's two components related to that. Number one is availability of supply, which obviously in the AI space is constrained. Number two is the fact that when you deploy these deals, you have to install it and then drive acceptances, which means elongated times for revenue recognition. And then maybe in a specific win or two, there are other conditions related to the contractual agreements."

Neri added that he's pleased with the quality of AI deals the company is making. HPE, like Dell, is building a lineup that aims for a portfolio of gear for AI training to tuning to inferencing for enterprises looking at domain-specific models.

Broadcom CEO Hock Tan said AI systems take time and supply issues are going to impact a bevy of infrastructure players.

"These products for generative AI take long lead times," said Tan. "We're trying to supply, like everybody else wants to have, within lead times. You have constraints. And we're trying to work through the constraints, but it's a lot of constraints. And you'll never change as long as demand, orders flow in shorter than the lead time needed for production."

According to Intel CEO Pat Gelsinger, who spoke at an investor conference last week, competitors will start grabbing some of that GPU demand. Intel's entry in the GPU and AI accelerator race is its Gaudi lineup. Gaudi is a line of AI accelerators built off the acquisition of Habana in 2019.

Gelsinger said Nvidia "has a great leadership position," but "there's a bit of false economy there." He said there's huge demand, high prices and supply chain constraints for Nvidia GPUs but that won't last forever. "Lots of people are showing up, including us, to compete," said Gelsinger. "We've seen a rapid expansion of our Gaudi pipeline. We're building our supply chains to get much larger for our footprint there as we start competing as well as others will."

Intel's argument is that GPUs and CPUs will be used for AI models and enterprises will follow pricing vs returns. "There is a bit of euphoria so overall I expect to see more moderation," he said. "We're going to be competing more for the GPU and accelerator, but we also see workloads driving energy that will create opportunities for our CPU offerings as well."

 

[Terroni2105]

not sure if this was posted already? Anyway it’s good for a second look.

Sunil is a senior guy at Tata Elxsi