Hardware News.
The chips, GPUs and data-center hardware underneath the AI boom, explained without the spec-sheet jargon. We cover new silicon from Nvidia, AMD, Intel, Apple and the custom-ASIC challengers, and why each launch matters for cost, performance and who wins.
Hardware
OpenAI's First Chip, Jalapeno, Targets Cheaper Inference
OpenAI and Broadcom unveiled Jalapeno on June 24, 2026, OpenAI's first custom chip: a TSMC 3nm inference-only ASIC with eight HBM stacks, taped out in nine months, that Broadcom claims runs LLMs about 50% cheaper than GPUs.
Hardware
Nvidia's RTX Spark Superchip Bets Windows Goes Agentic
Nvidia's RTX Spark is a single superchip pairing a 20-core Arm CPU, a Blackwell GPU and 128GB of unified memory to deliver about 1 petaflop of AI compute, built to run agents locally instead of in the cloud.
Hardware
AMD Brings FSR 4.1 to RX 7000 Cards, Weeks Ahead of Plan
AMD shipped FSR 4.1 upscaling to Radeon RX 7000 GPUs in the Adrenalin 26.6.2 driver on June 22, 2026, earlier than its July promise, using an INT8 model to reach RDNA 3 and light up 300+ games.
Hardware
HBM4 Enters Production for NVIDIA's Vera Rubin
SK hynix, Samsung, and Micron are all in mass production of HBM4 for NVIDIA's Vera Rubin GPUs, doubling the memory interface to 2,048 bits and pushing past 2TB/s per stack, with first systems shipping in Q3 2026.
Hardware
The Memory Supercycle Is Here, and AI Priced Everyone Out of RAM
DRAM contract prices are set to jump 50 to 55 percent this quarter and HBM is sold out for all of 2026 as AI data centers consume memory capacity. SK Hynix plans to double output, but consumers face pricier phones and PCs now.
Hardware
Nvidia's Vera Rubin NVL72 Enters Production, and AI Scale Climbs
Nvidia confirmed its next-generation Vera Rubin NVL72 system enters production ramp in Q3 2026, packing 72 Rubin GPUs and roughly 20.7 TB of HBM4 memory per rack. It is built to keep AI training scaling as models outgrow today's hardware.
Hardware
Intel Bets on Memory, Not Muscle, With the 480GB Crescent Island AI GPU
Intel unveiled Crescent Island, a data center GPU with up to 480GB of memory built for agentic AI. In a market obsessed with raw compute, Intel is quietly attacking a different bottleneck.
Hardware
Nvidia Is Reviving the RTX 3060 in 2026. That Tells You Everything About the GPU Market.
A five-year-old graphics card is coming back this month because the factories that should be building new ones are busy printing AI chips. The consumer GPU market has effectively frozen.
Hardware
Nvidia's RTX Spark Superchip Is a Land Grab for the Entire PC
At Computex, Nvidia unveiled an Arm-based superchip that fuses a 20-core CPU, a Blackwell GPU, and 128GB of unified memory into one package. It is a direct shot at AMD, Intel, and Qualcomm.
Hardware
Why ARM Chips Are Eating the Laptop Market
For decades the laptop meant x86. A wave of ARM-based machines flipped the script by leading with the one spec buyers actually feel: battery life.
Hardware
What '3nm' Actually Means on a Chip
Chipmakers love to brag about nanometers, and every new number sounds like progress. The truth behind the marketing is more interesting than the label.
Hardware
RISC-V Is the Open Standard Chipmakers Can't Ignore
Most chips speak an instruction set someone else owns and licenses. RISC-V is the radical idea that the basic language of a processor should be free.
Hardware
Why Your Next GPU Costs More Than Your Last
The graphics card used to be where enthusiasts splurged. Now it's where an entire industry's priorities collide, and gamers pay the difference.
Hardware
The Quiet Engineering Behind a Good Mechanical Keyboard
A keyboard seems like the most solved object in computing. Get close to one and you find a surprising amount of deliberate engineering under each key.
Hardware
Why Solid-State Drives Quietly Beat Hard Drives
One of the biggest leaps in everyday computing wasn't a faster processor. It was replacing a spinning disk with a chip, and most people barely noticed why it mattered.
Hardware
The Heat Problem: Why Chips Can't Just Get Faster
There's a simple reason your processor doesn't just run twice as fast every year anymore. Physics turned the dial down, and the whole industry had to adapt.
Hardware
What an NPU Is, and Why Your Laptop Suddenly Has One
New phones and laptops advertise a component most buyers have never heard of: the NPU. It reflects a real shift in what our devices are being asked to do.
Hardware
Why AI Data Centers Are Straining the Power Grid
The AI boom is, underneath, an electricity story. Training and running these models needs enormous power, and the grid was not built for the demand surge.
Hardware
HBM: The Memory That Makes Modern AI Possible
The chips that train AI get all the glory. The unsung hero sitting right next to them is a stack of memory most people have never heard of.
Hardware
Why Displays Moved From LCD to OLED
The screens on our best phones and TVs went through a quiet revolution in how they make light. The difference explains why a good OLED looks so striking.
Hardware
Apple Raises Mac and iPad Prices as Memory Costs Bite
Apple is pushing up MacBook and iPad prices, blaming soaring memory costs, a sign the AI-driven squeeze on DRAM is reaching consumers.
Hardware
Oxide's Rack-Scale Computer Is a Bet Against the Cloud Status Quo
Oxide ships a whole rack designed as one coherent computer, hardware, firmware, and control plane co-designed. An interactive 3D tour shows the pitch.