28th February - Valencia.

According to Deloitte Global the world’s leading semiconductor companies are predicted to spend US$300 million on internal and third-party AI tools for designing chips in 2023, and that number will grow by 20% annually for the next four years to surpass US$500 million in 2026.

AI chips refers to a new generation of microprocessors which are specifically designed to process artificial intelligence tasks faster, using less power. AI processors are embedded with logic gates and highly parallel calculation systems that are suited to typical AI tasks such as image processing, machine vision, machine learning, deep learning and artificial neural networks. 

With AI applications gaining traction in the industrial, retail, health care, military, research, and consumer sectors, demand for specialised sensors, integrated circuits, improved memory, and enhanced processors is increasing. And this demand is changing the semiconductor supply chain by directly impacting design and manufacturing decisions.

New AI chip design creating new opportunities

Semiconductor architectural improvements are needed to address data use in AI-integrated circuits. Running large AI algorithms is already harmful for the environment, and it will only get worse as the algorithms grow longer. One solution to counter this problem is neuromorphic computing, which takes inspiration from biological brains to create energy-efficient designs.

Specialised AI Neuromorphic chips, called NeuRRAM  include 3 million memory cells and thousands of neurons built into its hardware that have been developed to run ever increasing algorithms. The NeuRRAM chip can store more information from massive AI algorithms in the same amount of chip space. As a result, the new chip performs as well as digital computers on complex AI tasks like image and speech recognition, and is claimed to be up to 1,000 times more energy efficient. This is opening up the possibility for tiny chips to run increasingly complicated algorithms within small devices previously unsuitable for AI and open up new opportunities like smart watches and phones.

But it’s not all about transistor count and the actual number of components, it’s about how those components are used, and how a complete neural network accelerator solution for the AI chip can be a landscape for the future.

Improvements in semiconductor design for AI will be less about improving overall performance and more about speeding the movement of data in and out of memory with increased power and more efficient memory systems.

New AI Design Tools coming to fruition

Deloitte calculates AI design tools are enabling chipmakers to push the boundaries of Moore’s law, save time and money, reduce resources and even drag older chip designs into the modern era. At the same time, these tools can increase supply chain security and help mitigate the next chip shortage.

With AI applications gaining traction in the industrial, retail, health care, military, research, and consumer sectors, demand for specialised sensors, integrated circuits, improved memory, and enhanced processors is increasing. And this demand is changing the semiconductor supply chain by directly impacting design and manufacturing decisions.

Semiconductor architectural improvements are needed to address data use in AI-integrated circuits. Improvements in semiconductor design for AI will be less about improving overall performance and more about speeding the movement of data in and out of memory with increased power and more efficient memory systems.

Optimisation of PPA

AI is making great strides also in Electronic Design Automation tools (EDA) and in design efficiency for next-gen chip tech. Major AI player Synopsys, recently produced its 100th AI-designed chip tape-out that places bare semiconductor chips like integrated circuits onto a flexible circuit board (FPC). Top industry players like STMicroelectronics, SK Hynix and Microsoft are supporting the company’s DSO.ai (Design Space Optimization) place and route technology.

Traditionally engineers design the circuits, logic, memory structures, etc., then these blocks need to be mapped to a silicon mask set that ultimately can be patterned on a wafer for chip production. That mapping and connecting of various circuits has historically been very labour-intensive, taking a team of engineers weeks or months to achieve optimal performance, silicon area and power efficiency, or “PPA” — Performance, Power and Area optimisation. The Synopsis DSO.ai with reinforced machine (RE) learning capability can now deliver better results in far less time with far fewer design engineering human resources required. The company is also looking to utilise AI in other areas of chip design as well, like verification and even logic circuit design optimisation.

Advanced AI tools can test human designs by finding placement errors that increase power consumption, discover errors that impede performance, or even those that use space inefficiently. These tools learn from prior iterations to improve PPA and what really makes advanced AI stand out is that it can do it in hours with a single design engineer compared to weeks or months with an engineering team.

"AI's ability to explore broader design spaces is accelerating customers’ relentless drive towards better PPA and higher productivity with fewer engineering resources," noted Shankar Krishnamoorthy, GM for the EDA Group at Synopsys. “Designers are seeing significant gains from optimised designs delivering better results and faster time-to-market. “

Some examples of the impact of AI can be found in MIT’s AI tool which developed circuit designs that were 2.3 times more energy-efficient than human-designed circuits or MediaTek using AI tools to trim a key processor component’s size by 5% and reduce power consumption by 6%.

These tools are being used in the real world across many chip designs that are likely to be worth billions of dollars annually. Though they won’t replace human designers, their complementary strengths in speed and cost-effectiveness give chipmakers much stronger design capabilities.

AI is having a great impact on the productivity and supply chain for semiconductor development and will require the best talented engineers to guide the next AI developments and applications. It requires experts to discover the right resources amongst a dwindling global supply. Experts such as CIS have placed the right people in the right projects since the embryonic stages of AI chip development so make sure your next project is securely covered call CIS on +34 963 943 500 or email us on info@cis-ee.com.