Is AI the best news for semiconductor industry since microprocessors?

Note: This blog was published via Tractica in 2017.

Microprocessors form the basis of any computer system. Whether the device is used for IoT, consumer or cloud, the microprocessor chip is there that acts as a brain. In fact, Semiconductor industry owes its success to microprocessor when Intel introduced 4004 back in the 70s. That later evolved into 808X series leading to widespread adaption of chips.

Even today, microprocessors get more coverage than other types of chipsets in the media. Microprocessors are still the single most expensive semiconductor component in BOM on many devices and server processors could fetch as high as $1000 for a single chip.

So why are microprocessors able to command higher premium? One can use the analogy of human anatomy. With advances in science, we are able to fix heart, limbs and other parts of the body. But there aren’t any fixes just yet for brain. Microprocessors by and large has acted as the powerful brain of computer systems that only select few are able to manufacture and advance the technology.  There should not be any surprise then that they command higher margins and premium. Even today, Intel’s margins are close to 60%.

As of 2017, we are at a cusp of AI revolution. The need for faster hardware for AI applications has been identified in academia for long time and industry has just started jumping in by creating new architectures. CPU, GPU, FPGA and ASIC companies are heavily investing into the area and new products are being introduced at a rapid pace.

These AI chipsets will essentially act as a brain for all these systems, and will provide intelligence in conjunction with the microprocessor. So previously, if thinking was done by processor, then vast majority will be done by AI chipsets in the future. This means that deep learning chipsets will be able to command similar, if not more, premium pricing. This would mean a  great news for semiconductor industry which has suffered from low growth rate due to eroding ASP.

Here’s another potential direction that AI chipsets could take semiconductor industry– it will lead into newer business models for semiconductor companies. For instance, semiconductor industry could sell boxes or cloud services. Nvidia is already offering DGX1 box optimized for deep learning for $129,000 – far above an ASP of a chip. Nervana, an ASICs company for deep learning chipsets (now acquired by Intel), planned on offering a cloud service using their chipset. In the future, the cloud could be enhanced to include intelligent compute, storage and network systems. Such system could adapt itself based on the demand and nature of the application and be scale up or down as needed while providing maximum performance at any given time.

This could also lead to a few other interesting evolutions of service based business models. For instance, we could have ‘Intelligence’ as a service business model. In this model, companies would periodically update the neural networks on their customer’s devices making sure that the devices are always up to the task with latest data. Services such as ‘vision as a service’ or ‘emotions as service’ could also emerge in which devices could add these capabilities as and when needed.

Is AI the best news for semiconductor industry since microprocessors?

Microprocessors form the basis of any computer system. Whether the device is used for IoT, consumer or cloud, the microprocessor chip is there that acts as a brain. In fact, Semiconductor industry owes its success to microprocessor when Intel introduced 4004 back in the 70s. That later evolved into 808X series leading to widespread adaption of chips.

Even today, microprocessors get more coverage than other types of chipsets in the media. Microprocessors are still the single most expensive semiconductor component in BOM on many devices and server processors could fetch as high as $1000 for a single chip.

So why are microprocessors able to command higher premium? One can use the analogy of human anatomy. With advances in science, we are able to fix heart, limbs and other parts of the body. But there aren’t any fixes just yet for brain. Microprocessors by and large has acted as the powerful brain of computer systems that only select few are able to manufacture and advance the technology.  There should not be any surprise then that they command higher margins and premium. Even today, Intel’s margins are close to 60%.

As of 2017, we are at a cusp of AI revolution. The need for faster hardware for AI applications has been identified in academia for long time and industry has just started jumping in by creating new architectures. CPU, GPU, FPGA and ASIC companies are heavily investing into the area and new products are being introduced at a rapid pace.

These AI chipsets will essentially act as a brain for all these systems, and will provide intelligence in conjunction with the microprocessor. So previously, if thinking was done by processor, then vast majority will be done by AI chipsets in the future. This means that deep learning chipsets will be able to command similar, if not more, premium pricing. This would mean a  great news for semiconductor industry which has suffered from low growth rate due to eroding ASP.

Here’s another potential direction that AI chipsets could take semiconductor industry– it will lead into newer business models for semiconductor companies. For instance, semiconductor industry could sell boxes or cloud services. Nvidia is already offering DGX1 box optimized for deep learning for $129,000 – far above an ASP of a chip. Nervana, an ASICs company for deep learning chipsets (now acquired by Intel), planned on offering a cloud service using their chipset. In the future, the cloud could be enhanced to include intelligent compute, storage and network systems. Such system could adapt itself based on the demand and nature of the application and be scale up or down as needed while providing maximum performance at any given time.

This could also lead to a few other interesting evolutions of service based business models. For instance, we could have ‘Intelligence’ as a service business model. In this model, companies would periodically update the neural networks on their customer’s devices making sure that the devices are always up to the task with latest data. Services such as ‘vision as a service’ or ‘emotions as service’ could also emerge in which devices could add these capabilities as and when needed.

Higher premiums and new business models is one of many ways AI could impact semiconductor industry. The AI revolution is just beginning and we are going to see many innovations in the coming days. So buckle your seatbelt. The plane is just about to take off.