No chips, no AI: Semiconductors account for 95% of AI server rack value
The SIA warned that supply chain vulnerabilities could hamper.
America’s AI ambitions Semiconductors represent more than 95% of the content value of a leading artificial intelligence server rack, according to a new report from the Semiconductor Industry Association (SIA), underscoring the degree to which the global AI boom rests on a foundation of silicon.
Powering AI: The Semiconductor Ecosystem at the Foundation of Data Centers, produced by the SIA in partnership with Deloitte, offers an inside-out examination of AI infrastructure by conducting a virtual teardown of a state-of-the-art AI data server rack   — the foundational unit of centralised AI infrastructure.
The findings reveal the extraordinary chip density packed into today’s AI hardware. A single server rack contains more than 4,500 packaged chips, comprising approximately 20,000 individual dies — the unique integrated circuits etched from silicon wafers.
Semiconductors also account for more than half of the total capital expenditure required to build and operate an AI data centre. The report makes clear that AI is not driven by any single class of chip, but depends on the full spectrum of semiconductor technologies working in concert.
Advanced logic chips — including AI accelerators, application-specific integrated circuits, field-programmable gate arrays, and central processing units — handle the heavy computational lifting of AI training and inference.
Memory technologies, including high-bandwidth memory and various forms of dynamic and static random-access memory, provide the data throughput those processors demand.
Analogue and foundational chips, amongst them power management devices, transceivers, and sensors, underpin the entire system. The commercial stakes are considerable.
According to the report, government and industry are expected to invest more than four trillion dollars in new data centre infrastructure globally through 2028, with up to $2.8 trillion of that directed specifically at semiconductors.
Annual semiconductor revenues from AI data centres alone could exceed $1.2 trillion by 2028 — nearly ten times the figure recorded just four years earlier, surpassing total global semiconductor sales from 2025 across all end uses by more than 50%. 
The AI data centre market is projected to grow at a compound annual growth rate of 88.8% between 2022 and 2028.
Although much of the initial surge was fuelled by the rapid commercial adoption of generative AI, the report projects sustained momentum, with a 56.3% compound annual growth rate expected from 2025 to 2028.
The SIA’s report describes a self-reinforcing dynamic at the heart of the industry: advances in AI accelerate demand for more powerful and efficient chips, whilst improvements in semiconductor technology in turn enable more capable AI systems.
This feedback loop, the report argues, makes the health of the broader semiconductor supply chain a matter of strategic importance rather than mere industrial policy.
“The future of AI will be determined by the full array of semiconductor technologies that power it,” said John Neuffer, SIA president and chief executive. “To ensure global AI adoption is built on American chips, we need government policies that promote access to global markets and strengthen US competitiveness.” 
The warning carries particular resonance at a moment when geopolitical pressures, export controls, and manufacturing concentration have placed semiconductor supply chains under heightened scrutiny on both sides of the Atlantic and Pacific.
