Tech BusinessChips & Semiconductors
Lam Research executive weighs physics against data in AI chip design
A Lam Research executive argues that chipmakers need both physics-based simulation and data-driven methods as AI pushes designs towards angstrom-scale dimensions.

Key takeaways
- Semiconductor Engineering reports that AI is increasing the number of decisions in chip design and manufacturing.
- David Fried of Lam Research discusses simulation speed and process variability at angstrom-level dimensions.
- The article argues that physics-based and data-driven methods are both needed.
- The discussion also covers memory types and packaging options for chip systems.
In this article
The debate at the heart of AI chip design
Chip engineers are facing a growing question as AI workloads expand: should key decisions rest on data or on physics, or on a mix of the two? Semiconductor Engineering reports that the issue now touches everything from atomistic simulation to materials chemistry and the structural stability of 3D architectures.
The publication frames the discussion around David Fried, who holds an AI leadership role and a corporate vice president title at Lam Research, a supplier of semiconductor manufacturing equipment. He leads Semiverse Solutions, a Lam business unit, according to the report. The article presents his views on how simulations can be accelerated and how process variability can be handled at very small dimensions.
Why angstrom-level accuracy is hard
Chips are now built at dimensions measured in angstroms, a unit far smaller than a nanometre. At that scale, small variations in a manufacturing process can change how a device behaves, so models must account for variability rather than assume ideal conditions.
Semiconductor Engineering says the conversation covers ways to speed up simulation and how to tune systems for different types of memory and for packaging options. These are practical concerns for companies designing chips that combine several components in one package.
The pressure from AI

The report says AI is pushing the chip industry harder and faster than before. That pressure creates more decisions across design and manufacturing, and each one carries a cost in time and money.
The surrounding coverage on the same site lists research into high-bandwidth memory, AI accelerator security and advanced packaging. That range shows how many parts of the supply chain are being reshaped by AI demand.
What it means for the industry
The article does not settle the physics-versus-data question. Instead, it presents the case that both approaches are needed when features are measured in angstroms. That position matters for equipment makers, chip designers and the research groups working on the next generation of processors.
The source is an interview-based feature rather than a peer-reviewed study, so its claims reflect the views of the executive quoted. Readers who want the full discussion can find it at Semiconductor Engineering.
Sources
- Physics Vs. DataSemiconductor Engineering
Spotted a mistake? See our corrections policy. Read how News by OnDevice works in our editorial policy.



