The global semiconductor industry may be entering a fundamentally different growth cycle as artificial intelligence drives unprecedented demand for computing, memory, advanced packaging and manufacturing capacity.
Ajit Manocha, President and CEO of SEMI, believes the industry could be moving beyond its traditional boom-and-bust pattern into a longer period of structural expansion. Speaking as the semiconductor sector accelerates investment around artificial intelligence, Manocha has highlighted the possibility that global semiconductor revenue could approach US$2 trillion by 2030 and potentially reach US$3 trillion by 2035.
The scale of those numbers illustrates how dramatically the semiconductor industry’s growth drivers are changing. AI is no longer influencing only leading-edge processors. Its impact is spreading throughout the semiconductor supply chain, from memory and wafer fabrication to packaging, equipment, materials, photonics and data-center infrastructure.
The semiconductor industry has historically been highly cyclical. Periods of strong demand encouraged manufacturers to expand capacity, frequently resulting in oversupply, falling prices and another downturn.
AI could alter that pattern.
The enormous computing requirements associated with training and running increasingly sophisticated AI models are creating demand across multiple semiconductor technologies simultaneously. GPUs and other AI accelerators receive much of the attention, but the infrastructure behind those processors requires considerably more.
Modern AI systems need high-performance memory, advanced logic, high-speed networking, sophisticated packaging, power-management devices and increasingly complex data-center infrastructure.
As AI moves from primarily model training toward much broader inference deployment, semiconductor demand could expand across an even wider range of applications.
This is one reason Manocha has described the current semiconductor expansion as potentially different from previous industry cycles. Rather than depending on a single consumer electronics category, the AI infrastructure build-out is generating investment across several layers of the semiconductor ecosystem.
The long-term projections illustrate the potential magnitude of the transformation.
| Period | Potential Semiconductor Industry Revenue | Major Growth Drivers |
|---|---|---|
| 2025 | Industry surpasses US$1 trillion | AI accelerators, memory, data centers, advanced nodes |
| 2030 | Approaching US$2 trillion | AI inference, HBM, advanced packaging, silicon photonics, infrastructure |
| 2035 | Potentially US$3 trillion | Continued AI expansion plus emerging technologies including quantum computing |
Reaching those levels would represent an extraordinary expansion of the semiconductor market and would require corresponding growth in manufacturing infrastructure throughout the supply chain.
One of the clearest examples of AI’s impact can be seen in memory.
AI accelerators require enormous memory bandwidth to move data quickly between processors and memory. This has made High Bandwidth Memory (HBM) one of the semiconductor industry’s most strategically important technologies.
Demand for HBM, DDR5 and other advanced memory technologies is encouraging memory manufacturers to accelerate both capacity expansion and technology migration.
SEMI projects worldwide 300mm fab equipment investment in the memory sector to exceed US$50 billion for the first time in 2026, reaching approximately US$52 billion. Investment is projected to increase further to approximately US$57 billion in 2027.
| Memory Investment | 2026 Forecast | YoY Growth |
|---|---|---|
| Total 300mm Memory Equipment | US$52 billion | +29% |
| DRAM Equipment | US$37 billion | +29% |
| 3D NAND Equipment | US$14 billion | +28% |
However, additional investment does not immediately translate into additional usable capacity. Semiconductor fabs take years to plan, construct, equip and qualify. Advanced memory production also involves increasingly complicated process technology, while HBM introduces additional manufacturing and packaging requirements.
This creates a fundamental challenge: AI demand can increase considerably faster than the semiconductor industry can build qualified capacity.
AI is also changing the strategic importance of semiconductor packaging.
For decades, performance improvements were primarily associated with transistor scaling. While leading-edge process technology remains critical, AI systems increasingly depend on integrating multiple semiconductor devices into a single high-performance package.
Technologies including 2.5D integration, 3D integration, chiplets, HBM integration, advanced fan-out and other heterogeneous integration approaches allow designers to combine compute, memory and specialized functions more efficiently.
Advanced packaging is therefore becoming part of the semiconductor performance roadmap rather than simply the final manufacturing step after wafer fabrication.
This shift creates opportunities for OSAT companies, foundries, substrate manufacturers, equipment suppliers, materials companies, IP providers and engineering firms supporting advanced package development.
Moving data is becoming almost as important as processing it.
Large AI clusters may contain enormous numbers of processors communicating continuously with memory, storage and other accelerators. Traditional electrical interconnect technologies increasingly face challenges involving bandwidth, latency and power consumption as these systems scale.
Silicon photonics and optical interconnect technologies could therefore play an increasingly important role in AI infrastructure.
The combination of advanced logic, HBM, sophisticated packaging and high-speed optical communication is creating an entirely new semiconductor technology stack around AI computing.
The AI semiconductor opportunity is therefore considerably broader than GPU manufacturing.
| Semiconductor Segment | AI-Driven Requirement |
|---|---|
| Logic / Foundry | AI accelerators, CPUs, GPUs and custom ASICs |
| Memory | HBM, DDR5 and high-capacity storage |
| Advanced Packaging | 2.5D/3D integration, chiplets and HBM integration |
| Equipment | New fabs, capacity expansion and advanced process technologies |
| Materials | Advanced wafers, substrates, chemicals and packaging materials |
| Photonics | High-bandwidth optical data transmission |
| Test & Inspection | Testing increasingly complex multi-die systems |
| Automation | Higher fab productivity and manufacturing efficiency |
This breadth is particularly important for smaller semiconductor companies and technology suppliers. Companies do not necessarily need to compete directly in AI processors to participate in the AI semiconductor expansion.
A supplier providing packaging IP, test technology, specialized materials, design services, manufacturing equipment, interface IP, thermal management or high-speed connectivity may participate indirectly in the same growth cycle.
The industry must also physically manufacture everything required by this expansion.
SEMI’s latest outlook points to record levels of semiconductor manufacturing investment. Global 300mm front-end equipment spending is expected to reach approximately US$142 billion in 2026, with further investment growth expected through the remainder of the decade.
This creates opportunities, but it also creates potential constraints.
Building a semiconductor fab requires more than purchasing manufacturing equipment. Companies need suitable land, electricity, water, chemicals, materials, infrastructure, process technology and thousands of trained employees.
Consequently, access to manufacturing capacity itself is increasingly becoming a competitive factor.
Another constraint is people.
The semiconductor expansion requires engineers and technicians across IC design, process engineering, equipment, packaging, materials, testing, automation and manufacturing operations.
This is particularly relevant across Asia-Pacific, where semiconductor manufacturing and technology investment continues to expand.
The industry’s growth could create opportunities not only for experienced semiconductor engineers but also for graduates and professionals moving from adjacent engineering fields.
Electrical engineering, mechanical engineering, materials science, chemistry, software, automation, data science and AI increasingly intersect within modern semiconductor manufacturing.
The competition for qualified semiconductor talent may therefore become almost as strategically important as the competition for manufacturing capacity.
Perhaps the most interesting aspect of Manocha’s outlook is that semiconductor growth may not end when today’s AI infrastructure expansion eventually matures.
AI could remain one of the industry’s major growth engines throughout this decade. Beyond that, quantum computing and other emerging technologies could potentially generate another semiconductor investment cycle.
This raises the possibility that the semiconductor industry is entering a period characterized by successive technology waves rather than one isolated AI boom.
AI training is already driving demand for accelerators and HBM. AI inference could significantly broaden semiconductor deployment. Advanced packaging and silicon photonics may become increasingly important as systems scale. Quantum computing and technologies that have yet to reach commercial maturity could eventually create additional demand.
Calling the current expansion a permanent supercycle would still be premature. Semiconductors remain exposed to inventory corrections, capital-spending cycles, geopolitical risks, supply-chain disruptions and changes in end-market demand.
But the underlying structure of semiconductor demand is clearly changing.
AI is simultaneously increasing requirements for computing, memory, networking, packaging, manufacturing equipment and infrastructure. That makes the current expansion broader than many previous semiconductor growth cycles.
If the semiconductor industry approaches US$2 trillion around 2030 and continues toward US$3 trillion by 2035, the largest opportunities may not belong exclusively to the companies producing the most advanced AI processors.
They will extend across the ecosystem supporting them.
Foundries, memory manufacturers, OSATs, equipment suppliers, materials companies, IP vendors, design-service providers, test companies and specialist engineering firms could all participate in what may become one of the largest expansions the semiconductor industry has ever experienced.