Nvidia is working with Japan’s newly formed Noetra consortium to build a 140-megawatt AI factory that will serve as the compute backbone for the government’s FRONTia program, which focuses on robotics, industrial automation, and digital twins. The facility will run on Nvidia’s DSX data center platform and use Vera Rubin NVL72 racks, each combining 72 Rubin GPUs and 36 Vera CPUs.

In total, the project calls for 27,500 Rubin GPUs and 13,750 Vera CPUs connected through Spectrum-X Ethernet networking. The system is designed to handle the training of trillion-parameter models as capacity ramps up.

Nvidia is positioning the Rubin AI factory as a full-stack platform encompassing chips, networking, systems, and software for training and deploying open multimodal foundation models.  — Read More

TSMC achieved the impossible. Then, Huawei, IBM, and OpenAI each announced they no longer need it. Something fundamental broke in the chip industry last week, and it changes everything from your next.

The Rule That Ran the World for 50 Years Is Dying.

For half a century, the logic was almost offensively simple. Shrink the transistor, get twice the performance. No strategy needed. Physics handled it. The rule is called Moore’s Law — the empirical observation that transistor counts would roughly double every two years — and it’s what took computers from filling an entire room to fitting in your pocket.

But we are in mid-2026. And three major announcements in the space of about a week — from Huawei, IBM, and OpenAI — are all saying the same thing, from three different directions: the era of shrinking transistors as the primary lever of progress is closing. — Read More

A new chip architecture from IBM can integrate nearly 100 billion transistors on a chip the size of a human fingernail—nearly twice the transistor density of the company’s previous generation of chip technology. The resulting improvement in chip compute performance and energy efficiency comes from what IBM describes as the “world’s first sub-1 nanometer chip technology” for AI data centers. — Read More

#nvidia

OpenAI and Broadcom on Wednesday unveiled their debut custom chip, called Jalapeño, marking the ChatGPT maker’s first entry into artificial intelligence silicon.

The chips will be made by Broadcom and used by OpenAI for inference, the compute-intensive process of serving its AI models to users in ChatGPT and other applications. — Read More

Developers building for AR glasses and wearable devices face an infrastructure gap. The hardware is ready, but creating AI experiences requires integrating live camera and microphone streams, multimodal AI models, enterprise data, tool use, deployment infrastructure, and device-specific runtimes.

NVIDIA XR AI is designed to address this challenge by providing a reusable foundation for connecting extended reality (XR) devices to GPU-accelerated AI services running in the cloud, data center, workstation, or edge. — Read More

Nvidia has announced a new chip for PCs as it moves into the consumer market for devices integrated with AI technology.

“This reinvention of the computer is as big of a deal as the reinvention of the phone into what we now know as the smartphone,” Nvidia’s chief executive Jensen Huang said as he unveiled the RTX Spark chip. — Read More

For fifty years, the semiconductor industry ran on one rule: make the transistor smaller, and everything gets faster. You did not need a strategy. Physics did the work. Every new generation of chip delivered 30 to 50% more transistors on the same surface area, and the entire digital economy rode that escalator upward. A single, consistent dynamic fuels everything from your phone and laptop to artificial intelligence.

… That rule is dying. And how the two companies that matter most in semiconductor manufacturing respond to its death will reshape the chip industry for the next decade. — Read More

Today, Arm is announcing the Arm AGI CPU, a new class of production-ready silicon built on the Arm Neoverse platform and designed to power the next generation of AI infrastructure.

For the first time in our more than 35-year history, Arm is delivering its own silicon products – extending the Arm Neoverse platform beyond IP and Arm Compute Subsystems (CSS) to give customers greater choice in how they deploy Arm compute – from building custom silicon to integrating platform-level solutions or deploying Arm-designed processors. It reflects both the rapid evolution of AI infrastructure and growing demand from the ecosystem for production-ready Arm platforms that can be deployed at pace and scale. — Read More

The Nemotron 3 family of open models — in Nano, Super and Ultra sizes — introduces the most efficient family of open models with leading accuracy for building agentic AI applications.

Nemotron 3 Nano delivers 4x higher throughput than Nemotron 2 Nano and delivers the most tokens per second for multi-agent systems at scale through a breakthrough hybrid mixture-of-experts architecture.

Nemotron achieves superior accuracy from advanced reinforcement learning techniques with concurrent multi-environment post-training at scale.

NVIDIA is the first to release a collection of state-of-the-art open models, training datasets and reinforcement learning environments and libraries for building highly accurate, efficient, specialized AI agents. — Read More