The article discusses the principles of LiDAR technology, including the differences between 905nm and 1550nm wavelengths, as well as Time of Flight (ToF) and Frequency Modulated Continuous Wave (FMCW) methods. It also highlights the current market dynamics surrounding these technologies.
The article discusses how a new storage tier can enhance long context inference for AI agents while significantly reducing cache storage costs. It also highlights the specifics of the storage system, potential market leaders, and associated risks.
The article argues that the era of traditional software development, akin to PDFs, is being replaced by more flexible and efficient methods, represented by markdown. It suggests that memory hierarchies serve as a better analogy for the evolution of software, indicating a shift in value creation within the industry.
The article presents a unique perspective on the water consumption of datacenters by comparing it to the equivalent number of burgers produced. This approach shifts the focus from traditional metrics like tokens per watt or dollar to a more relatable measure of resource usage.
The article discusses Credo's potential to expand its market reach beyond Automotive Electronics Components (AECs). It raises questions about the company's ability to diversify its offerings in the semiconductor landscape.
Intel's product marketing at CES has shown improvement in understanding market realities, although their AI marketing efforts fell short again. The PTL 18A process technology has been positively highlighted.
The article discusses advancements in semiconductor technology showcased at IEDM 2025, including interconnects beyond copper, the introduction of 1,000 CFETs, and next-generation NAND from SK Hynix. It also highlights the potential of 2D materials in future applications.
The article provides an update on SRAM bandwidth calculations, highlighting the evolving requirements for inference tasks in the context of high bandwidth memory (HBM). It emphasizes the importance of accurate bandwidth metrics for optimizing SRAM performance in modern applications.
The article discusses the limitations of SRAM in AI inference, arguing it is not the ultimate solution many expect. It also emphasizes the continued relevance of HBM in the semiconductor landscape.
The article discusses the transition of advanced AI models from cloud environments to local supercomputers, drawing parallels to the evolution of computing from minicomputers to mainframes. This shift highlights the growing capabilities and relevance of local computing power in handling complex AI tasks.
The article discusses the evolving dynamics of wafer demand and node economics in the semiconductor industry, particularly as artificial intelligence (AI) influences the foundry landscape. It highlights the partnership between Apple and TSMC as a pivotal factor in shaping modern semiconductor manufacturing.
The article suggests a delay in the release of the 2026 AI & Semiconductor Outlook, indicating a shift in the usual schedule. It hints at a different approach or perspective being taken this year.
The article discusses advancements in worker automation and the integration of reinforcement learning (RL) in various scientific applications. It highlights the potential of RL as a service and the implications of multi-agent architectures in fields like biology and lab procurement.
The article provides coverage of the International Electron Devices Meeting (IEDM), a key conference focused on advancements in semiconductor devices. It highlights the significance of the event as a platform for showcasing cutting-edge research and innovations in the industry.
The article discusses the challenges associated with lithography in semiconductor manufacturing, focusing on key players like ASML and xLight. It highlights the economic implications of substrate choices in the lithography process.
Co-Packaged Optics (CPO) technology is finally reaching the market with deployment-ready products emerging in 2025. This development is set to significantly enhance datacenter connectivity.
The article discusses the shift towards onsite power generation solutions in response to the energy crisis, highlighting various technologies such as turbines, reciprocating engines, and fuel cells. It emphasizes the total cost of ownership (TCO) of these systems compared to traditional electric grid reliance.
Nvidia is expanding its product lineup by introducing a hyperspeed SKU, which is expected to enhance its competitive edge in the GPU market. This addition aims to cater to the growing demand for high-performance computing applications.
The article indicates that there will be significant price changes and new per-post options introduced in 2026. These changes may affect how semiconductor companies engage with their audience and market their products.
The article suggests that the complexities of the semiconductor landscape may be overstated, implying a simpler underlying reality. It challenges the notion that strategic maneuvers in the industry are as intricate as they are often portrayed.