EDA
System
IP
SolvNetPlus
Training & Education
Amit Jain, Senior Director at Samsung Semiconductor India, brings nearly 23 years of expertise in digital design and SoC management, with a strong focus on leveraging AI for innovation in semiconductor solutions.
The Era of AI Is Now
AI is not just a buzzword—it’s fundamentally reshaping our world, especially as it migrates from data centers into edge devices, enabling smarter, context-aware decision-making everywhere. This shift toward edge AI also poses new challenges and opportunities for our industry.
Is AI Over-Hyped?
Let’s begin with a look at the Gartner Hype Cycle for 2025. Technologies like artificial general intelligence (AGI) and generative AI (GenAI) have rapidly progressed from hype to deployment much faster than their predecessors. Investment in AI, particularly in silicon and device innovation, now outpaces non-AI investments by five times. Surveys show that 80% of organizations already use AI, and 70% employ GenAI. In fact, GenAI is now woven into our daily experiences. Applications range from text-to-video, text-to-image, and text-to-text generation to intelligent customer service bots, automated finance operations, healthcare diagnostics, and marketing content creation. AI is here to stay, permeating industries far beyond tech. This is all evidence that AI is not just the next big thing, but it is THE BIG THING.
How AI has Changed Chip Design
While AI dates back to the 1950s, its progress was stalled for decades during the so-called “AI Winter,” primarily due to a lack of computing power. The turning point came when advances in chip design enabled robust data storage and faster computation, allowing AI to flourish.
Today, chip design and AI share a symbiotic relationship. Complex system-on-chip (SoC) architectures and innovative packaging have enabled AI, and now AI is accelerating chip design itself. AI impacts RTL design, design verification, and implementation—optimizing workflows, improving quality, and driving efficiency. AI automates routine tasks, boosting productivity and accelerating time-to-market through machine learning (ML) and generative models.
AI is transforming every stage of chip design, from initial development to final implementation, by streamlining critical processes and enhancing overall efficiency:
- RTL Design: AI automates IP creation, SoC integration, and quality checks, leveraging databases and machine learning engines to abstract and accelerate the design process.
- Verification: AI-driven automation speeds up test bench generation, debug, and regression, making verification more efficient and thorough.
- Physical Design: Machine learning optimizes performance, power, and area (PPA), enabling engineers to manage increasingly complex designs with greater productivity.
Knowledge Assist and the Rise of the Copilot
AI also has an impact on the training and onboarding engineering talent. Workforce ramp-up is resource-intensive. AI-powered knowledge assistants—soon evolving from copilots to even autopilots—help query documentation, solve problems, and provide instant, context-aware guidance. As these tools become more personalized, prompt engineering (the skill of interacting efficiently with AI models) will emerge as a vital expertise for VLSI designers.
Edge AI: Bringing Intelligence to Source
AI-powered assistants are helping engineers learn faster and solve problems on the job by giving them quick, personalized answers. In a similar way, AI is moving closer to where data is actually created—right on our everyday devices. This is known as Edge AI.
Edge AI means that artificial intelligence runs directly on devices like smartphones, smartwatches, factory sensors, and cars, instead of relying on distant data centers. By processing information on the spot, these devices can respond more quickly and efficiently. This matters because:
- Data Volume: In 2024, 175 zettabytes of data will be generated, mostly on edge devices like smartwatches, smartphones, factories, and autonomous vehicles.
- Bandwidth & Efficiency: Transferring all this data to data centers is impractical due to bandwidth and energy constraints. Processing data locally is more efficient and secure.
- Latency & Context: Real-time decisions—like detecting a heart anomaly on a smartwatch—require low latency and context-aware processing.
- Security: Distributed processing enhances data security by avoiding centralized vulnerabilities.
Challenges and Opportunities for the Chip Industry
For edge AI to truly take off, engineering teams must develop new solutions in several key areas. Edge devices often have limited energy resources, so keeping power consumption low is essential. At the same time, these devices need enough performance to handle advanced AI models, making high computational power another important requirement. Real-time responses are critical for many applications, which means minimizing delays is non-negotiable. Additionally, any solution must be affordable and scalable to ensure broad adoption.
The industry is now moving toward custom ASICs designed specifically for edge applications. New connectivity technologies, such as 6G and near-cloud computing, will also play a major role in making edge devices even more capable and efficient.
What’s Next? Agentic AI and Autonomous Systems
Edge AI marks the evolution from the Internet of Things (IoT) to the “Intelligence of Things” – devices that don’t just sense but also analyze and act. The next paradigm shift is Agentic AI—goal-driven AI agents that can sense, think, and act independently. In home automation, for instance, multiple AI agents can collaborate to respond to emergencies, orchestrated by a master agent. This will lead to truly autonomous, productive, and sustainable systems.
Conclusion: Embracing the Future
AI is transforming the VLSI industry and the world at large. Understanding and mastering these technologies—and skills like prompt engineering—will be crucial for future innovators. Edge AI is where intelligence meets the physical world, promising profound social impact and new opportunities for the chip industry. The journey ahead is challenging, but also incredibly exciting and it’s happening now!
Continue Reading
3D Chips: Socionext Achieves Two Successful Tape-Outs in Just Seven Months
