The technology stories enterprise IT leaders actually read.
Volume 341, Issue 3 · 10 articles from 5 issues
Samsung is developing a fingernail-sized storage chip to replace M.2 SSDs and boost on-device AI.
Samsung is developing next-generation storage significantly smaller than conventional M.2 solid-state drives, roughly the size of a fingernail.
The miniaturization targets two main benefits: extending battery life in portable devices and speeding up on-device AI inference. Shrinking the storage component would free internal space and cut power consumption.
The move reflects Samsung's strategic pivot toward supporting local AI processing rather than relying on cloud-based solutions.
IBM packs 100 billion transistors into a chip using a new nanostack design likened to a skyscraper.
IBM has announced a record-breaking semiconductor achievement, fitting 100 billion transistors into an unprecedentedly compact footprint.
The company used an innovative nanostack design that the article compares to a 100-storey skyscraper packed with highly efficient processing power.
This miniaturization milestone surpasses earlier predictions about the limits of chip density.
The breakthrough demonstrates continued progress in semiconductor engineering, with significant implications for future computing performance and efficiency.
GPUs, TPUs and NPUs each serve distinct roles, optimized for workloads from training through edge inference.
The article explains how three specialized processors power contemporary machine learning.
GPUs provide flexibility for training large models with broad framework support, while Google's TPUs optimize tensor-heavy operations within its cloud ecosystem.
NPUs enable efficient on-device inference on smartphones and laptops while preserving battery life. Rather than competing, these chips work together in production systems, with GPUs or TPUs handling training and NPUs deploying models to end users.
Software maturity and framework integration often matter as much as raw specifications when choosing accelerators.
The semiconductor industry must scale silicon photonics production to keep pace with AI data center demand.
The semiconductor industry must accelerate silicon photonics production to keep pace with demand for tomorrow's data centers. The technology works, the physics is proven, and the manufacturing pathway exists; what is missing is widespread industry alignment and investment to produce silicon photonics at the pace AI infrastructure will require.
Silicon photonics are already integral to commercial production of communications transceivers in data centers, converting electrical signals into optical ones, transmitting them as infrared light pulses through microscopic waveguides etched into silicon, and converting them back at the destination.
Coordinated investment must span chipmakers, substrate suppliers, packaging specialists, and hyperscalers.
RISC-V's modular design is becoming key to custom silicon for physical AI in robotics and autonomous systems.
Physical AI requires processors that can sense, reason, and act simultaneously in dynamic environments, unlike traditional data center AI.
RISC-V's modular architecture lets chip designers add custom instructions optimized for AI inference, sensor integration, and machine vision.
Software-defined methods and virtual prototyping are accelerating the path from algorithm to fabricated silicon while reducing validation timelines.
The architecture's extensibility supports integrating CPUs, accelerators, and specialized processors into unified platforms while preserving code compatibility.
An expanding collaborative ecosystem is driving rapid innovation and reducing reliance on proprietary processor designs.
Quantum advances enable practical business applications in sensors, security, ML, and simulations as 52% of companies invest $5M+ annually.
Quantum computing is transitioning from experimental stage to commercial viability with applications across sensors, security, machine learning, and materials simulation.
Key use cases include quantum sensors for measuring time and gravity, quantum key distribution for tamper-resistant encryption, quantum ML for accelerated AI training, and quantum simulations for pharmaceutical and materials research.
The emerging quantum internet promises to connect isolated quantum computers into distributed networks using photons through existing fiber infrastructure. Despite high operational costs and current limitations, quantum breakthroughs could revolutionize business environments across industries including healthcare, finance, pharmaceuticals, and energy.
Quantum technology's future success depends on establishing trust with users and stakeholders.
The article explores how quantum computing's advancement hinges not just on technical progress but on building confidence among potential users.
Trust is a critical factor in quantum technology adoption, requiring transparency about both capabilities and limitations. As the industry develops quantum solutions, establishing credibility with enterprises and consumers becomes essential for widespread implementation and market growth.
SK Hynix and Sandisk are developing vertically stacked flash to relieve the KV cache bottleneck in AI inference.
SK Hynix showcased high bandwidth flash, a new vertically stacked flash storage technology developed with Sandisk, at the FMS 2026 conference in San Jose.
Although still about a year from production, backers say HBF could help alleviate one of the biggest challenges in AI inference: the memory wall.
The key value cache has emerged as a major obstacle to rolling out AI services, since it stores pre-computed answers to previous queries and functions as short-term memory during inference.
Larger KV caches let AI service providers serve more and longer contexts, making cheaper high-bandwidth capacity a direct lever on inference economics.
USB4 reaches 80Gbps, enabling fast transfers and high-resolution displays over a single connection.
USB4 supports three signaling tiers at 20, 40 and 80 Gbps for rapid data transfer, and can deliver up to 240W of charging power.
Every USB4 device uses a USB-C port, but not all USB-C ports support USB4 speeds.
Users must verify specifications for laptops, cables and accessories to get full performance across data, charging and display functions.
Cloud PCs offer a subscription alternative to hardware refresh cycles, with connectivity as the tradeoff.
As hybrid work reshapes endpoint management, organizations must choose between traditional PCs, Cloud PCs, DaaS and VDI.
Cloud PCs provide faster provisioning, simplified management and support for distributed workforces through subscription models, but require reliable network connectivity. Traditional hardware remains practical for power users and offline work.
The ideal strategy often combines both, balancing cost, flexibility and enterprise requirements.
More from Technology · All IT section newsletters · This week