Governments around the world are pouring tens of billions into semiconductor manufacturing in the name of chip sovereignty. From the U.S. to Europe to Asia, the assumption is clear: Control the fabs, and you control the future of computing.
But as AI reshapes the semiconductor landscape, that assumption no longer captures where value is created. Strategic advantage is shifting toward a broader—and more complex—stack: AI accelerators, high-performance CPUs, advanced packaging, memory bandwidth, and the software and data that tie them together.
The result is a growing disconnect between how policymakers define strategy and how the industry actually builds and deploys AI. In practice, industry is moving from a chip-centric model to a system-centric one.
From chips to systems
According to Mark Papermaster, CTO of AMD, the shift has happened faster than most expected. “The way people think about strategic leadership in AI has really changed as compared to just a year ago,” he told EE Times. “AI used to be thought of as associated with a GPU. But over the last year, it’s become clear that you need extensive computation across a much broader system.”
The system is no longer defined by any single class of chips. Instead, it spans CPUs, GPUs, specialized accelerators, memory, storage, and networking—working together to support increasingly complex workloads. The biggest change is the rise of what Papermaster describes as agentic workflows—AI-driven workflows that not only generate outputs but orchestrate entire sequences of tasks across enterprise software, databases, and applications.
