Four Architectural Opportunities for LLM Inference Hardware (Google)


A new technical paper titled "Challenges and Research Directions for Large Language Model Inference Hardware" was published by Google. Abstract "Large Language Model (LLM) inference is hard. The autoregressive Decode phase of the underlying Transformer model makes LLM inference fundamentally different from training. Exacerbated by recent AI trends, the primary challenges are memory and in... » read more

Annual Global IC Fabs And Facilities Report


Semiconductor companies announced a significant number of facilities in 2025 as global onshoring efforts continued across manufacturing, materials, packaging, design, and R&D. Investments came from both industry and government sources. Organizations worked together to solve current technology challenges, including soaring demand for AI chips and advanced memory, as well as complex applic... » read more

Scalable End-To-End Test Solutions For Today’s Complex SoCs


By Srikanth Venkat Raman and Sri Ganta Today’s highly complex and large system on chip (SoC) devices and systems present many challenges to be addressed from manufacturing tests to the field while meeting stringent requirements for test costs, test quality, yield, debug, and turn-around-times. Scalable and efficient end-to-end test solutions that scale to large and complex SoC design cores... » read more

Chip Industry Week in Review


Major Deals: Taiwan-based UMC is exploring possible collaboration with Polar Semiconductor for high-volume production of 8-inch wafers at Polar’s expanded Minnesota fab, a move that could provide domestic manufacturing capacity for automotive, data center, consumer, aerospace, and defense customers. Marvell will acquire Celestial AI for $3.25B, adding photonic fabric technology for o... » read more

Chip Industry Week In Review


Breaking news: Nvidia and Synopsys announced a multi-faceted, multi-year deal that includes everything from digital twins to CUDA programming, engineering, and marketing collaboration, and Nvidia's $2B purchase of Synopsys stock. [Updated 12/1] Memory news: Micron is building a $9.6B HBM facility in the city of Higashi-Hiroshima Japan, reports Nikkei. China's ChangXin Memory Technol... » read more

Chip Industry Week In Review


China's Hefei Lumiverse Technology reportedly has developed a desktop-sized High Harmonic Generation light source that generates wavelengths as small as 1nm. One customer already has used it to produce 14nm chips, which was the original target node for EUV, according to one report. As a point of comparison, TSMC and Samsung didn't start using EUV until the 7nm node, relying instead on immersion... » read more

Chip Industry Week in Review


Samsung reportedly is hiking memory chip prices by 30% to 60% due to high demand from AI data centers and constrained supplies. Those shortages are causing ripples elsewhere. SMIC, China's largest foundry, said its customers are holding back orders for other types of semiconductor due to concerns about memory supplies. Meanwhile, interest in photonics and power semiconductors is picking up, ... » read more

Small Vs. Large Language Models


The proliferation of edge AI will require fundamental changes in language models and chip architectures to make inferencing and learning outside of AI data centers a viable option. The initial goal for small language models (SLMs) — roughly 10 billion parameters or less, compared to more than a trillion parameters in the biggest LLMs — was to leverage them exclusively for inferencing. In... » read more

Chip Industry Week in Review


Retaliations and countermoves leading up to planned trade talks between the U.S. and China led experts to wonder, 'Who's winning?' New activity on this front: China issued questionnaires to some U.S. semiconductor firms as part of an anti-dumping probe, demanding detailed data on sales, profit margins, logistics costs and Chinese customer names for analog chips. The probe appears aimed at ... » read more

AI Bubble Or Boom?


Are we in an AI bubble? Parallels are being drawn to the dot.com boom/bust of 1999-2000. In the dot.com bust, many high-tech companies valuations soared up 10X, then deflated. The peak P/E ratio for the Nasdaq Composite was 200! Remember Webvan? It went public November 1999 with an $8 billion valuation, then filed for bankruptcy 19 months later. It was much speculation without profits or gro... » read more

← Older posts Newer posts →