The Silicon Brief - 16 August 2026
Today's digest: Lam Research commits $3 B to R&D lab expansion, Michigan and Imec launch Common Earth to cut rare-earth dependence, AI chiplet test complexity deepens, CPO test standardization stalls, and the industry reframes competition around systems rather than silicon.

Today's edition covers Lam Research's $3 billion bet on faster process development, a University of Michigan-Imec project to engineer rare earths and forever chemicals out of the supply chain, the growing test burden created by AI chiplet architectures, the standardization gap blocking co-packaged optics at scale, and a broader industry argument that competitive advantage has moved from the transistor to the system.
Lam Research commits more than $3 billion over five years to expand its global R&D lab network, with the multi-site build-out designed to increase experiment capacity by more than 50% across facilities in the United States, Asia, and Europe[1]. The company's integrated 24/7 lab network already runs more than one million experiments annually and has shortened process development by up to 2.5 times in recent customer engagements[1]. The investment is a direct response to AI-era demand for chips with new materials, architectures, and nanoscale features - and signals that equipment suppliers now see R&D velocity, not just tool capability, as the differentiator that keeps them ahead of chipmaker roadmaps[1].
The University of Michigan and Imec have launched a project called Common Earth, aimed at replacing rare earth elements and forever chemicals in the semiconductor supply chain[3]. Researchers Valeria Bertacco and John Heron are leading the effort at Michigan, working with Imec's Jim Foresi to find material substitutes that sidestep both geopolitical concentration risk and environmental liability[3]. Many chip components are location-constrained and can easily be throttled by geopolitical factors, and chipmaking is also tied up with social and environmental concerns[3] - making Common Earth one of the more structurally significant research programs to emerge this cycle.
AI chiplet architectures are forcing a fundamental rethink of where and how test is inserted in the production flow, according to analysis published this week by Semiconductor Engineering[4]. AI accelerators are no longer constrained primarily by transistor scaling, but by integration strategy and package-level interconnect density, with compute die, HBM stacks, and I/O die assembled on silicon interposers in configurations that invalidate many assumptions built into conventional test flows[4]. The practical consequence is that known-good-die qualification, interposer verification, and system-level test must now be treated as distinct, non-interchangeable steps - a cost and cycle-time burden the industry has not yet fully priced in[4].
Co-packaged optics test will not scale to production volumes without agreed standards, Semiconductor Engineering reported this week, citing the absence of shared definitions for data rates, failure rate specifications, test limits, and guard-bands[5]. OSATs are beginning to define production test cell architectures, but the lack of a common framework means each customer engagement currently requires bespoke test equipment designed around a specific functionality and performance envelope[5]. Until chipmakers, OSATs, and test-equipment vendors converge on a standard, CPO yield learning will remain siloed and slow.
Semiconductor Engineering's "From Silicon to Systems" analysis argues that competitive advantage in the chip industry has structurally shifted away from transistor scaling toward systems, ecosystems, and integration[2]. For decades the industry was driven by making transistors smaller, faster, and cheaper - but today competition is increasingly happening at the level of systems and ecosystems[2]. The piece frames this not as a temporary plateau but as a permanent reorientation: winners will be defined by how well they integrate silicon, software, and supply chain rather than by node leadership alone[2]. For chip professionals, the implication is that roadmap conversations that once started at the transistor now have to start at the workload.
Written by Electronics Insider's automated desk from the sources above and published automatically. How we work.
Related
Design & EDAETH Zürich and NYU show JEDEC test loops miss worst-case HBM2 power by 23.6%
Researchers at ETH Zürich and NYU tested 36 real HBM2 chips and found JEDEC measurement loops understate peak power by 23.6%, with read ops and data-bus toggling as the main drivers.
15 Oct 2026
Design & EDAFlowAccel's Joel Jorgensen argues chip teams must design their workflows before AI can help them hit tape-out
Joel Jorgensen of FlowAccel argues that AI tools alone won't save a tape-out schedule - teams must first design the cross-team workflows that let those tools compound into real gains.
15 Oct 2026
SemiconductorsETH Zürich and NYU find JEDEC test loops underestimate HBM2 power by up to 35.4%
Researchers at ETH Zürich and NYU tested 36 real HBM2 chips and found JEDEC-specified current loops can underestimate peak power by up to 35.4%, with implications for AI accelerator thermal design.
15 Oct 2026