ON Semiconductor targets $213B AI power market amid strategic shift

Everyone’s been talking about compute and memory as the chokepoints holding back AI. Onsemi thinks they’re looking at the wrong bottleneck.

At its Investor Day on September 16, the chipmaker laid out a thesis that power density, not processing speed or storage, will be the constraint that defines the next era of AI infrastructure. And the company is positioning itself to be the one solving it, targeting a total addressable market of $213 billion by 2030 for its core power semiconductor business.

The numbers behind the bet

Onsemi’s AI data center revenue sits at roughly $500 million in 2026. By 2030, the company expects that figure to exceed $2.5 billion, representing a compound annual growth rate north of 50%.

The broader revenue picture is equally ambitious. Onsemi is aiming for approximately $11 billion in total revenue by 2030, implying a 12-14% CAGR across the business. That growth spans automotive electrification, industrial automation, AI data centers, and what the company calls “physical AI” applications.

The Synaptics acquisition, which onsemi announced as forthcoming, would push the total addressable market even higher, from $213 billion to over $243 billion.

Power density as the next constraint

CEO Hassane El-Khoury framed power density as the successor problem to compute and memory limitations. AI systems are anticipated to evolve from approximately 60 kW to 200 kW and possibly reach 1 MW in the future, making the electricity required to run these systems the binding constraint.

To that end, onsemi unveiled two new technology platforms. The GaNEXUS portfolio focuses on gallium nitride (GaN) solutions, a material that enables power conversion at higher efficiencies than traditional silicon. The Embedded Power Platform, or EPP, uses wafer-level integration to pack more power capability into smaller form factors.

Hyperscaler demand stays strong

El-Khoury addressed what might be the biggest bear case head-on: the possibility that AI investment is cooling. His answer was blunt. He dismissed near-term concerns about an AI slowdown, pointing to ongoing commitments from hyperscaler customers as evidence that demand remains intact.

What this means for the competitive landscape

Onsemi occupies an interesting position in the semiconductor food chain. It doesn’t compete directly with Nvidia for GPU dominance or with AMD for data center processors. Instead, it’s playing in the power management layer, the infrastructure beneath the infrastructure.

The company’s move into GaN technology adds another dimension to its positioning, since gallium nitride fabs require specialized processes that can’t simply be repurposed from existing silicon lines. Onsemi has also previously acquired GTAT for silicon carbide (SiC) technology, further deepening its advanced materials capabilities.

The stock trades on NASDAQ under the ticker ON. A 50%-plus CAGR in AI data center revenue is the kind of number that either validates a thesis spectacularly or becomes a cautionary tale about Investor Day optimism.