AI chip boom drives fabless revenue up 73%

The global fabless semiconductor sector delivered exceptional growth in the second quarter of 2026, fueled by the rapid expansion of artificial intelligence infrastructure and applications. According to TrendForce, the combined revenue of the world’s ten largest fabless IC design companies surged 73% year on year to more than US$141.45 billion.

Demand is expanding well beyond graphics processing units. Investment in AI data centers is also driving sales of CPUs, application-specific integrated circuits, networking chips, optical interconnect products and power-management solutions. As AI workloads become more complex, chipmakers with broad computing and connectivity portfolios are emerging as major beneficiaries.

NVIDIA maintained its commanding lead, generating nearly US$91.16 billion in revenue during the quarter—a 99% increase from the previous year. The company accounted for approximately 64% of the top ten companies’ combined revenue, reinforcing its dominant position in the global AI accelerator market.

Demand remained strong among hyperscale cloud service providers, emerging NeoCloud operators and sovereign AI projects. Enterprise customers also contributed significantly as more businesses invested in AI computing infrastructure to support training, inference and industry-specific applications.

NVIDIA is simultaneously extending its reach into the custom silicon market through its NVLink Fusion ecosystem. The initiative enables custom processors to connect more effectively with NVIDIA’s computing infrastructure. Its investments in MediaTek and Marvell could further strengthen interoperability across custom XPU platforms while creating opportunities in data centers and future automotive AI applications.

Broadcom retained second place after its revenue jumped 112% year on year to more than US$18.89 billion. The company benefited from the initial shipments of a custom ASIC developed for OpenAI, the rollout of Google’s TPU 8i and robust demand for networking products used in large-scale XPU deployments.

Broadcom’s custom chip business is supported by six major customers, including frontier AI model developers such as Anthropic and OpenAI. Their continued investment in proprietary processors is expected to sustain demand for Broadcom’s ASIC design capabilities and advanced networking technologies.

One of the quarter’s most significant developments was AMD’s rise to third place. The company generated more than US$11.53 billion in revenue as the emergence of agentic AI systems increased the strategic importance of CPUs alongside GPUs and other accelerators.

Agentic AI applications must process instructions, manage workflows and coordinate multiple tasks, creating heavier demands on data center CPUs. This shift helped AMD’s data center CPU revenue reach a record high for the fifth consecutive quarter. The company also continued to gain market share from Intel in the x86 server segment.

Qualcomm slipped to fourth place as its smartphone chip business faced continuing pressure and its share of modem chips in the latest iPhone declined. Nevertheless, the company’s automotive division achieved double-digit year-on-year growth for the 23rd consecutive quarter, helping offset part of the weakness in mobile devices. Qualcomm is also returning to the data center market as it seeks to diversify its semiconductor portfolio.

MediaTek ranked fifth with revenue of approximately US$4.81 billion. Although it was also affected by softer smartphone demand, the company raised its outlook for AI ASICs and expects its data center revenue to exceed US$2 billion in 2026. MediaTek has estimated that the serviceable market for ASICs could reach US$80 billion by 2028, underscoring the scale of the opportunity created by custom AI processors.

The need to move enormous volumes of data between processors is also creating opportunities for connectivity specialists. Marvell placed sixth after revenue reached US$2.63 billion, supported by strong demand for high-speed interconnect products. The company remains a leading supplier of PAM4 digital signal processors, with continued demand for 800G solutions and a rapid ramp-up of next-generation 1.6T products.

Marvell has also secured an agreement with Google in the custom silicon market. Although the partnership is not expected to make a significant revenue contribution until fiscal 2029, it provides the company with another pathway into the fast-growing AI accelerator ecosystem.

Consumer-focused chip designers are likewise searching for new growth drivers. Realtek ranked seventh with revenue of nearly US$1.19 billion. With the PC market expected to weaken in the second half of 2026, the company plans to use its existing intellectual property to expand into server products, including 10G Ethernet solutions and control-plane switches.

Monolithic Power Systems placed eighth with revenue of US$981 million, supported primarily by demand for AI-related power-management solutions. As data centers deploy increasingly powerful processors, managing energy consumption, heat and power delivery has become a critical part of AI infrastructure development.

Novatek ranked ninth with revenue of US$907 million. System-on-chip products now generate more than half of its revenue, while the company continues investing in machine-vision SoCs for edge AI applications.

OmniVision completed the top ten with semiconductor design revenue of US$838 million. Higher memory prices weighed on its smartphone and automotive image sensor businesses, but emerging applications such as action cameras recorded strong growth. The company is also expanding its analog IC portfolio with products for optical communications.

The second-quarter results illustrate how AI is transforming the competitive landscape for semiconductor design companies. NVIDIA remains the dominant force, but the expansion of agentic AI, custom silicon and high-speed data center infrastructure is distributing growth across a much wider range of technologies.

From CPUs and ASICs to optical interconnects and power-management ICs, the AI boom is creating new opportunities throughout the semiconductor value chain. Chipmakers capable of supporting the complete computing ecosystem—not only processing power—are likely to play an increasingly important role as global AI investment continues to accelerate.