Semiconductors Accelerate The Artificial Intelligence Revolution

Deploying AI in the semiconductor industry to create a virtuous cycle of innovation.


By Ajit Manocha, Pushkar Apte, and Melissa Grupen-Shemansky

Five years ago, SEMI published an article predicting that Artificial Intelligence (AI) would change everything – and it has! AI has moved out of labs and taken popular imagination by storm. It is now a hot topic of discussion everywhere, from family dining tables and corporate boardrooms to corridors of government.

While it may seem like the height of a hype cycle for AI, we believe that that the technology has only just begun to take hold. Imagine if we could predict the precise path of a hurricane or a tornado early enough to move people out of harm’s way. Imagine if we could detect and prevent wildfires and avoid the havoc they wreak on the planet. Imagine if we could stop the next pandemic in its tracks so it does not paralyze the world for three years and kill millions of people like COVID-19 has. We think AI could help realize such a future, and more.

While everyone is watching and marveling at the AI like they might a supercharged racecar, what is often overlooked is the semiconductor engine under the hood. While the concept of AI is half a century old, development of the technology has accelerated rapidly over the past decade, mainly due to amazing advances in semiconductor chips. Why? AI algorithms function by crunching huge amounts of data and being trained to recognize patterns – for example, an image or a sequence of words. But AI is only as good as the data it is fed. It depends critically on the quantity and quality of data, and how quickly this data can be processed and analyzed.

Driven by Moore’s Law and the fierce pace of innovation, semiconductor chips have enabled an unprecedented amount of data to be sensed, stored, communicated and processed at incredible speeds. This souped-up engine powers the AI racecar to even higher speeds.

Many types of semiconductor chips come together to enable AI. Internet of Things (IoT) devices, which now number more than 15 billion, sense data from the physical world. IoT devices are usually the source of large data sets fed into AI algorithms, especially in industrial applications. High-bandwidth communication chips then transmit this data either wirelessly through 5G/6G networks or through high-capacity fiber optic cables, with bandwidths up to 100 Gbps. Solid-state memory devices store this data, squeezing gigabits of data into a single chip or package. Power management chips distribute and manage efficient power delivery to billions of devices while optimizing power consumption. Last but not least, processor chips run the AI algorithms by analyzing this data at speeds up to 100 billion instructions per second!

This progress is often taken for granted but it is worth a moment of reflection to appreciate the innumerable semiconductor innovations that have made this possible.

Keeping up this blistering pace of innovation has become increasingly challenging as Moore’s Law has slowed and R&D has become very expensive. The semiconductor industry spends ~17% of revenue on R&D – perhaps more than any other industry – or a staggering $100 billion a year. We believe that the path forward is smart innovation. Can we use AI to create a virtuous cycle? Can we harness AI to accelerate semiconductor technology development to build better chips that can run more powerful AI algorithms?

To help accomplish this, SEMI launched its Smart Data-AI initiative in 2018. One element of this effort is a proof-of-concept (POC) project being executed at Cornell University. Funded by the Army Research Laboratory, the project has demonstrated an AI model that predicts the precise patterned features on the chip – the tinier and more precise the features, the higher the chip performance. In parallel, SEMI formed the Smart Data-AI industry advisory council (IAC) to advance the state of the art and to guide the POC effort. We are now about to launch the next phase of the project with the ambitious goal of creating a virtual innovation environment with digital twins of processes to help drive semiconductor advances.

To our knowledge, this is the first such POC project involving multiple companies across the entire ecosystem together with academia and government. Bringing together such diverse entities is extremely challenging, but this is SEMI’s core strength: to develop such platforms to help advance the industry with SEMI’s mantra of ‘Connect, Collaborate and Innovate.’

The economic stakes are high. Since AI is now ubiquitous, projections show it adding several trillions of dollars across industries, with McKinsey & Company analysis projecting $4.4 trillion in value for generative AI alone. The global semiconductor industry itself is on a path to double in revenue over about the next seven years to $1 trillion.

The market is beginning to appreciate the contribution of semiconductor companies. For example, Nvidia, the maker of graphic processing units (GPUs) used extensively to run AI algorithms, became the first chip company to be valued at over $1 trillion with revenue of $27 billion, making for an impressive price-to-earnings ratio of 90! In parallel, the entrepreneurial push for pursuing creative new hardware solutions continues to grow rapidly with hundreds of well-funded start-ups now in the arena.

However, continuing growth on this innovative path will not happen on autopilot and will require proactive action. SEMI is driving precisely such action in keeping with its pivotal role as the global electronics industry association representing 3,000+ companies worldwide across the entire value chain. Since publishing the prediction five years ago, SEMI has walked the walk – starting with the launch of four industrywide Smart Initiatives for Data-AI, Manufacturing, Mobility and MedTech in 2018. Following are some multi-faceted activities that SEMI is driving for Data-AI:

  • Forming an industry community, the Smart Data-AI IAC that spans the full semiconductor ecosystem and focuses on advancing the state-of-the-art for Data-AI.
  • Advancing knowledge-sharing on Data-AI through numerous technical conferences, workshops and global SEMICONs.
  • Driving proof-of-concept projects that bring together industry, academia and government researchers in meaningful ways.
  • Ensuring access to talent through a comprehensive workforce development program to grow the talent pipeline, with a special focus on underserved communities to encourage diversity, equity and inclusion.
  • Training the existing workforce through SEMI University, which offers AI-focused courses to help semiconductor professionals to come up to speed on AI techniques and introductory semiconductor courses to help Data-AI professionals understand technology basics.
  • Driving sustainability through several R&D projects in a $25 million portfolio managed by SEMI Technology Communities to research low-power technologies that could reduce energy consumption significantly in the future. AI algorithms are usually power-hungry due to the immense amounts of data they must process, and techniques to mitigate this energy consumption are crucial for the sustainable growth of AI capabilities.
  • Overcoming potential roadblocks to innovation by addressing major industry concerns – for example, ways to share data securely across company and national borders. We are developing methodologies and data standards, building on our experience minting 1,000+ standards that SEMI has already implemented.
  • Sharing best practices both within and across industries. The Smart Data-AI IAC has been working together for three years, sharing what’s effective and what isn’t in implementing AI projects, and identifying opportunities for improvement. We are also exploring synergies with other industries such as healthcare through the Smart MedTech initiative. For example, AI-enhanced semiconductor R&D and manufacturing techniques may help healthcare industries, while their leadership in data management and cybersecurity protocols may benefit the semiconductor industry.
  • Providing thought-leadership and technology stewardship on the future of computing by uniting the brightest minds at Smart Data-AI initiative events and webinars as well as sessions at SEMI expositions and conferences.
  • Anticipating disruptions such as quantum computing, which could be a game changer but is currently confined to niche applications. What if quantum computers that require complex refrigeration and occupy entire rooms today could fit in the palm of our hands some day? It is worth recalling that mainframe computers from the last century that occupied a room now sit in our hands as cellphones, running 10,000 times faster!

The AI racecar has had an exhilarating journey for the past decade, from pattern recognition to today’s generative AI tools such as GPT4 and Bard. We believe the best is yet to come for AI, but much work is still needed to realize the full potential of AI as a powerful and transformative tool. Safe and effective use of AI could support a path to a future for everyone with better health, prosperity and quality of life, and SEMI is working diligently to realize this future responsibly and equitably.

We must collaborate to learn how to drive the AI racecar safely – and perhaps turbocharge it in the future with quantum technology. We invite you to join us on this exciting journey – get updates on the Smart Data-AI initiative or contact Puskar Apte of SEMI to get involved.

Pushkar Apte is a strategic technology advisor and leader of the SEMI Smart Data-AI Initiative.

Melissa Grupen-Shemansky is CTO at SEMI.

Leave a Reply

(Note: This name will be displayed publicly)