Insights

AI chips driving the next semiconductor supercycle: Strategic analysis and industry outlook

Swasti Pujari, Practice Head - VLSI Backend, UST

The semiconductor industry is undergoing a structural shift. While familiar cycles have long shaped the market, recent data marks a departure from historical trends. In 2024, global chip sales hit $627.6 billion—a 19.1% jump from the previous year. But this growth wasn’t evenly spread. Companies focused on AI chips experienced valuation surges of approximately 93%, while traditional sectors, including analog, automotive, and smartphone chips, declined sharply. This divide signals that a new semiconductor supercycle has arrived and AI chips are shaping it.

DIVIDER

The AI revenue disparity

Despite accounting for less than 0.2% of total wafer volume in 2024, AI chips generated roughly 20% of industry revenue—a notable increase in silicon value density. NVIDIA dominated the AI-driven semiconductor market, with Q3 2025 data center revenue surging to $30.8 billion, a 112% year-over-year increase. The memory sector followed suit, with DRAM revenue climbing 82.6%, primarily driven by the high bandwidth requirements of AI chips.

DIVIDER

Is a semiconductor supercycle on the horizon?

Throughout most of 2024, the semiconductor industry was split. AI-focused companies, such as Marvell, Broadcom, and ARM, saw significant gains, while legacy segments struggled. The broader semiconductor industry index dropped even as leading AI-driven semiconductor market firms hit record highs. This points to a fundamental shift in how the industry operates.

DIVIDER

High Bandwidth Memory and the development of AI

High Bandwidth Memory (HBM) has moved from a niche component to a core piece of AI infrastructure. Sales are expected to grow from $15.2 billion in 2024 to $32.6 billion by 2026, with a compound annual growth rate of 21.7% through 2028. HBM still costs about five times more than standard server DRAM, and supply depends on advances in packaging. TSMC plans to double its CoWoS wafer output to 660,000 units by 2025; however, even then, supply is likely to remain tight.

DIVIDER

AI chips market forecast 2025–2030

Chips built on advanced nodes (under 7nm) are projected to grow at a 12% annual rate in 2025, with utilization rates exceeding 90%. However, production remains capital-intensive and supply-constrained. The U.S. CHIPS Act is now shaping domestic strategy, with Intel investing $8.5 billion in facilities across four states. Still, long lead times mean pricing and availability will stay uneven through 2030.

DIVIDER

How AI demand is fueling semiconductor growth

The rise of AI workloads and hyperscale data centers is pushing the industry in a new direction. These applications need specialized chips with high-speed interconnects, advanced packaging, and extreme memory bandwidth. This isn’t just evolution—it’s a fundamental change in computing architecture. Legacy nodes and general-purpose CPUs are struggling to keep up. AI chips are now at the center of the industry's future.

DIVIDER

How geopolitics affect the future of the semiconductor industry

U.S. restrictions on high-end GPUs to China hit hard. NVIDIA logged $4.5 billion in related losses in Q1 2025 and expects further hits. In retaliation, China blocked the exports of key chip materials, including gallium, germanium, and antimony, creating structural risks to supply chains. Companies that localize production or diversify suppliers are better prepared to manage these challenges.

DIVIDER

Advancing trends in AI chip development

New design priorities are emerging: power efficiency, memory access, and interconnect speed. DeepSeek’s efficient AI model release—even with lower compute power—triggered a $600 billion drop in NVIDIA’s market cap in one day. The message? Performance alone isn’t enough. Future AI chip innovation is focused on efficiency and cost control as much as raw compute.

DIVIDER

Challenges in AI chip-driven supercycle

Even with strong revenue growth, the AI-driven semiconductor market boom faces real challenges. Packaging lead times are long. Sub-5nm lithography equipment is scarce. There’s a talent shortage in AI hardware design. And economic uncertainty complicates demand forecasting.

DIVIDER

Future outlook: AI semiconductors

Gartner expects global chip revenue to grow 11.8% in 2025 and 11.2% in 2026, hitting $733 billion by then. IDC predicts AI chip sales will increase over 15% annually through 2028. However, sectors such as automotive and consumer electronics are likely to underperform due to oversupply and weak demand.

For a broader perspective on how these shifts play out across industries, see UST’s Thinking Ahead Report 2025.

DIVIDER

What is ahead in AI chip development?

Key questions remain: Will inference shift to edge devices or stay in data centers? Will custom AI chips replace GPUs in key sectors? Could advances in sparsity, compression, or analog computing reshape the field again? The direction is set, but the road is fluid. These questions will define the next decade of AI chip progress.

DIVIDER

Why AI is driving explosive chip demand

AI chips’ workloads are both compute-heavy and memory-hungry. Each new generation of models requires significantly more silicon. Inference—which once played second fiddle to training—now drives a large share of data center investment. Every major cloud player is either developing or buying AI chips. This confirms that demand is here to stay.

DIVIDER

Custom chips versus general-purpose: Which one drives future innovation?

Custom silicon is taking the lead in AI. It offers better performance, efficiency, and cost for specific tasks. Google, Amazon, and Microsoft all have in-house accelerators, and Meta is expanding its own chip initiatives. The competition between custom and general-purpose AI chips is now one of the most important fronts in semiconductor supercycle innovation.

Why this matters

The AI chip boom isn’t just another market upcycle—it’s a structural shift redefining the semiconductor industry. From high-bandwidth memory to custom silicon, every new leap in AI architecture pushes the limits of performance, efficiency, and supply chains. Companies that anticipate these shifts, invest in the right technologies, and adapt to changing market dynamics will shape the next decade of computing. Those who don’t risk being left behind as the supercycle accelerates.

DIVIDER

Be part of the transformation

The AI-driven semiconductor era is here, and the opportunities are unprecedented. Whether it’s scaling advanced node production, innovating in chip design, or navigating the challenges of geopolitics and supply, the next move matters. Let’s explore how your organization can lead in this new landscape through strategic partnerships, forward-looking investments, and innovation that keeps pace with AI’s explosive growth.

Step into the semiconductor supercycle. Discover how AI chips are redefining innovation, disrupting supply chains, and creating trillion-dollar opportunities—and learn how your business can lead the next wave.

DIVIDER

Resources

https://www.ust.com/en/silicon-engineering/pre-silicon-engineering

https://www.ust.com/en/silicon-engineering

https://www.ust.com/content/dam/ust/documents/ust-thinking-ahead-report-2025.pdf