Oyon Advisory

Back to essays

March 29, 2026

Semiconductor Industry and AI Investment Boom: State in 2026 and Outlook

In 2026, the semiconductor industry sits at the center of the AI investment boom. Market growth is being driven not only by end-user demand, but also by a historic wave of data-center, compute, and infrastructure spending whose long-term return still remains an open question.

Content: Yili

Revision, table and figure creation: ChatGPT

Share on LinkedIn

1. How Large Is the Global Semiconductor Market?

Global semiconductor sales reached $610-680 billion in 2024, representing around 20% year-on-year growth. However, according to McKinsey, this likely understates the true size of the market because conventional sales-based estimates do not fully capture the value created through in-house chip design, captive semiconductor development, and fabless business models. To address this gap, McKinsey incorporates end-product COGS and gross margin into its methodology and estimates that the global semiconductor market reached approximately $775 billion in 2024. Based on recent estimates from McKinsey and Deloitte, the market could reach $1.0-1.8 trillion by 2030, implying a CAGR of roughly 13% from 2024 to 2030.

The computing and data storage vertical accounted for nearly half of the semiconductor market in 2024 and is projected to remain the fastest-growing segment through 2030, with an estimated 15% CAGR. Wireless and automotive are expected to remain the second- and third-largest verticals, respectively. In growth terms, wired infrastructure and automotive are also projected to outperform the broader industry, with expected CAGRs of around 15% and 13%, both above the overall market average.

2. Investment Versus Applications

Generative AI has triggered strong demand for high-density computing, GPUs, networking, and power infrastructure. As a result, the recent investment boom has become the primary growth driver for the computing and data storage vertical, while application demand has not yet fully caught up.

Table 1. Hyperscaler data-center capex, 2020-2026.

2.1 Capex Outlook for Data Centers

In 2025, global operational data-center capacity reached roughly 100 GW. Of this, around 50-80% was used for non-AI workloads, while 20-50% was dedicated to AI workloads, depending on the source and methodology. By 2030, total capacity is expected to exceed 200 GW. At that point, AI workloads could account for roughly 50-70% of capacity, while non-AI workloads could represent 30-50%. Given the significant uncertainty around AI use cases, adoption speed, and technology improvements, it is not surprising that different research sources show meaningful gaps in their projections.

Investment is flowing into three main areas: data-center infrastructure, IT equipment, and power. Across the value chain, five major investor groups can be identified: builders, energizers, technology developers and designers, operators, and AI architects. Among these, technology developers and designers are expected to spend the most, with around $3.1 trillion projected for the development of GPUs, CPUs, memory, servers, and rack hardware.

2.2 AI Applications

Since the launch of ChatGPT in 2022, AI applications, including AI assistants and various forms of AI agents, have emerged rapidly across the market. These applications have already demonstrated strong usefulness in knowledge work, education, research, and analysis. Even so, there remains a fundamental question: how large will the AI application market ultimately become?

AI can clearly improve productivity, and there has already been substantial news coverage of companies reducing headcount as they adopt AI. This leads to an obvious question: if there is still limited solid evidence that AI is creating significant additional social value, does it make sense to invest trillions of dollars into AI?

What, then, will the long-term use case for AI be? One way to think about AI is not as a single product category, but as a foundational layer for humanity, similar to electricity, the internet, or a common language such as English. AI is a powerful enabler: it gives people access to capabilities that previously required either many years of learning or substantial financial resources. That said, AI does not create value on its own. Its value is realized only through human use. The more knowledge, judgment, and domain expertise an individual has, the more effectively that individual can create value with AI.

For individuals and small companies, especially those whose work is heavily computer-based, AI can create substantial value by reducing the cost of generating the same amount of output or revenue. For large corporations and more complex businesses, adoption may take a different form, with AI being deployed primarily through enterprise tools that enhance employee productivity rather than fully replace existing workflows.

This suggests that AI application development may ultimately unfold at two levels: first, individual users building or adapting tools for their own personal and professional needs; and second, companies, organizations, and teams developing AI applications tailored to their own specific use cases. If that happens, it could create significant opportunities for AI tool developers, integrators, and customization specialists.

As individual productivity rises sharply, the traditional time-based compensation model in white-collar work may increasingly become a constraint on both individuals and organizations. In an AI-enabled economy, outcome-based compensation may make more sense. Over time, this could help free labor from rigid time structures and potentially stimulate broader consumption.

The key question therefore remains: how large will the AI application market ultimately become?

To be continued.