You are currently viewing SemiWiki as a guest which gives you limited access to the site. To view blog comments and experience other SemiWiki features you must be a registered member. Registration is fast, simple, and absolutely free so please,
join our community today!
Why bandwidth alone cannot determine the performance of the next generation of AI accelerators
High-bandwidth memory is entering another phase of rapid development. HBM4 increases interface width and available bandwidth, while custom HBM architectures create opportunities to optimize how memory connects to AI accelerators.… Read More
NVIDIA’s NVHBM could change more than how memory connects to an AI accelerator. It could also change how the semiconductor industry divides responsibility for designing, manufacturing, and qualifying high-bandwidth memory.
According to October 18 reporting from Seoul Economic Daily, NVIDIA has asked Samsung, SK … Read More
Why integration boundaries, assembly strategies, and manufacturing yield must be engineered together
The Shift from Optical Performance to Manufacturing Scale
Co-packaged optics (CPO) is moving beyond laboratory demonstrations toward commercial deployment in AI infrastructure. As switch bandwidth continues to increase,… Read More
The value of glass will be determined not by its material properties alone, but by whether it can expand the package architecture window and be realized repeatedly at manufacturing scale.
AI infrastructure is pushing advanced packaging into a regime where the substrate itself increasingly determines what can be built.
As packages… Read More
By Dr. Moh Kolbehdari
Putting roughly 25 million pixels into a wearable device weighing about six ounces sounds, at first, like a display problem.
It is not.
The displays must fit behind an optical system capable of turning those pixels into a usable visual field. Compute must generate and move the required data. Power must support… Read More
Before EDA becomes AI-native, focused agents have a window to prove measurable engineering value inside existing design flows.
AI is entering semiconductor engineering from several directions at once.
Some companies are building broad agentic design environments. Others are attacking specific bottlenecks: timing closure,… Read More
Agentic AI is beginning to move semiconductor engineering beyond copilots and isolated task automation.
From Agentrys’ public material and recent discussion, its architecture points toward a broader engineering flow:
Specification → RTL → verification → synthesis / physical design → PPA optimization → physical verification… Read More
Co-packaged optics is usually described as a proximity problem.
Move the optical engine closer to the switch ASIC or XPU. Shorten the electrical path. Reduce SerDes power. Increase bandwidth density.
That direction is real.
Optical functionality is moving progressively closer to compute.
But proximity does not eliminate … Read More
Product realization depends not only on reaching the next engineering state, but on preserving the relationships between states from intent through qualification, release, and lifecycle learning.
Semiconductor development is usually described as a sequence of activities: architecture, design, fabrication, packaging,… Read More
AI can compress the path to tape-out. The next challenge is compressing physical realization.
Something important is beginning to happen in semiconductor development: the design cycle itself is becoming compressible.
AI-assisted engineering can accelerate architecture exploration, RTL development, verification, physical… Read More
Physics Over Probability: How Agentic AI Solves Electronic Systems Verification