AI Predictions 2022: Applications & Trends in AI Chip Design

Stelios Diamantidis

Jan 05, 2022 / 5 min read

Artificial intelligence (AI) applications have become more integrated into the fabric of business success than many would have imagined even a few years ago. In the last year alone, AI innovation witnessed many big waves, from the advancements in edge AI and computer vision to data center modernization and specialized AI chips, to AI designing those very chips. These milestones of progress are unlocking exciting opportunities for the industry.

This AI revolution is also driving the need for an entirely new generation of AI systems-on-chip (SoCs). By 2026, the global market value for AI chipsets is expected to grow to more than $70 billion from $8 billion in 2019, as the surge in IoT devices with machine-learning capabilities and the development of smart cities propels the market. Investors are already pushing record money into AI startups, with the third quarter of 2021 witnessing a new global funding record of $17.9 billion, reflecting the broad requirement for an AI-everywhere world.

A new year brings new goals, opportunities, and challenges. With the semiconductor industry working hard to address supply chain constraints, the global chip shortage, and the economic effects of COVID-19 while also the exploring the possibilities of a “metaverse,” the growing desire to incorporate intelligence into the underlying silicon has made every organization want a fair share of the AI pie.

So, what does the next year have in store for AI? Read on for five key predictions on what we expect will fuel AI’s next act in 2022.

AI (Artificial Intelligence) concept

1. The Impact of AI in Chip Design Will Continue to Grow

With more workloads requiring advanced levels of AI processing to power intelligent functions, the need for dedicated chips that are energy-efficient and perform computations at higher speeds will make robust AI chip designs paramount. A new breed of design tools that learn from iterations and leverage data in chip design environments offer a leap forward in productivity and cost efficiency. The AI disruption will open new opportunities not only for the semiconductor leaders but also for companies that typically had smaller teams or more limited financial resources and weren’t previously seen as powerhouses of chip design to deliver exciting silicon solutions ― leveling the playing field, in a sense, and creating a symmetry of companies leveraging AI in chip design across the global economy.

Designing the AI hardware of the future will require a revolution in chip design technology. In 2021, companies that dared to invest in the data center market witnessed significant gains and demonstrated impressive technical capabilities that increased the demand for dedicated AI chips, raising AI funds at an unprecedented pace. With GPUs continuing to be the dominant architecture for training in the data center market, we expect this growth to continue and see hyperscalers opting for next-generation AI-assisted design systems to massively scale exploration design workflows while automating less consequential decisions. Companies will start pivoting to the cloud for their chip design needs ― a victory for additional capacity, faster turnaround times, and high-quality application-optimized design.

2. AIoT Will Drive Tremendous Opportunity in the Era of Smart Everything

With more devices connecting to the cloud, the internet of things (IoT) continues to provide enormous benefits to industries driving real-world applications at scale. A relatively new acronym that combines the powers of both AI and IoT, AIoT promises a more intelligent, smart, connected network of devices that can process and compute large data volumes previously not possible using traditional processing methods.

With the opportunities for technologies that are at the edge of IoT — like augmented intelligence and development of the metaverse — larger companies will refocus their strategies around where it makes the most sense to process real-time data and invest in AI innovations that will capitalize on the growth of AIoT devices.

3. Three Key Applications Will Push the AI Envelope

There will be three clear market winners that will prioritize using AI more to build better chips: high-performance computing (HPC), autonomous devices, and healthcare.

The HPC market has and will continue to drive major investments in AI chips, resulting in the need for dedicated chips for the data center that can perform computations for AI workloads that now consist of over a trillion nodes. At the edge, we are seeing more companies expand the design of AI chips for the automotive industry as well as for a variety of autonomous machines ― from industrial machines to autonomous robots and aerial machines like drones. Thanks to big data and the desire for everything to be connected, design teams will require silicon-proven IP solutions for complex SoCs across a variety of applications, an area that will continue to grow in 2022 amidst the strained supply chain.

For a world continuing to adapt to the COVID-19 pandemic, the incorporation of AI in healthcare and medicine promises a lot of opportunities, especially in diagnosis and medical research. Although the computation requirements may not be as extreme as that of a data center, the unique requirements around data protection, security, and real-time analytics require a localized environment that is safety-critical to evaluate and perform analysis in situ. From today’s AI accelerators to tomorrow’s cognitive systems, these three markets will witness a growing interest from both companies and investors, driving the growth of AI in chip design and the seamless integration of AI in devices.

4. More Non-Traditional Companies Will Double Down on Designing Chips

If there’s anything that 2021 has taught us, it’s that the rapid evolution of AI’s potential to reshape the chip design landscape has made almost every technology company think about joining the chip-design wagon. Apple’s recent launch of the impressive M1 Max that was developed in-house showcases the sheer innovation that can be realized to integrate several powerful computational components and deliver some of the industry’s most capable chips for desktop devices and beyond. The speed with which non-traditional semiconductor companies can scale up development efforts of their own custom ASICs (application-specific integrated circuits) incentivizes companies to look carefully into the competitive advantages of bringing silicon development in-house, especially against the backdrop of rapid growth in the key markets they serve.

These incentives include maximum data control and reduced latency between speed, insights, decisions, and outcomes. While building a world-class chip design team becomes an important way of creating and securing intellectual property, it will become increasingly difficult for companies to attract and retain a skilled workforce, as development rapidly expands to newer markets.

5. Establishing Trust in the Entire System Stack Will Become Critical

Whether it’s the AI-enabled system powering an autonomous car or making a financial transaction, or the AI tools making chip design decisions — all these parameters will require us to trust that these decisions will lead to better outcomes and productivity, rather than causing major specification defects, program delays, or financial implications for the customer. This will make companies prioritize different levels of trust within the hardware infrastructure that lies beneath to create secure channels for remote device management, service deployment, and lifecycle management (thus, ensuring that the entire system stack is trustworthy for the end-customer and not just the software).

As AI becomes more prevalent in computing applications, so too will the need for advanced trust and security at all levels of the system, especially at the design and integration stages. Up until now, AI hardware was not seen as critical in comparison to software. But with trust chains becoming especially important in the current environment of supply chain issues, companies will need a trusted chain throughout the workflow.

Ultimately, all these predictions will be driven by the need for faster computations, more intelligence at the edge, processing larger data volumes effectively, and automating more functions in the products we use. As AI permeates the enterprise, bold new hardware architectures and well-defined AI strategies will become a core enabler of innovation and seamless integration of AI into software systems. At Synopsys, we are committed to making technology smarter and more secure, from silicon to software, and to continue  investing in accelerating the growth of disruptive AI-driven design solutions in the years to come.

Continue Reading