AI Hardware
54 articles
Mukesh Ambani's Jio Opens Cloud PC Service to All of India, Turning Old Computers Into AI-Ready Machines
Reliance Jio expanded JioPC beyond its broadband base, streaming up to 8 vCPUs and 16GB of RAM to aging computers from about $11 for two months.
OpenAI Is Buying Mac minis and Mac Studios by the Tens of Thousands to Train AI Agents
The Information reports OpenAI bought tens of thousands of Mac minis and Mac Studios to train computer-use agents, turning Apple silicon into AI infrastructure.
Apple Caught Off Guard by Enterprise AI Demand as Mac Studio and Mac Mini Sell Out
Enterprise AI demand caught Apple off guard, The Information reports, leaving Mac Studio and Mac mini builds out of stock for months amid the local AI boom.
Nvidia Bets $3.5 Billion on MediaTek as NVLink Fusion Opens the Door to Custom AI Chips
Nvidia invested $3.5 billion in MediaTek convertible bonds as the chipmaker adopts NVLink Fusion to bring custom XPUs into rack-scale AI factories.
US Robot and Drone Barriers Won't Fix the Scale Gap With China, Analysts Warn
Chinese firms built 86 percent of the world's new humanoid robots in early 2026. Analysts say US tariffs and bans will not erase that scale advantage.
Hugging Face Launches MicroDuck, a $399 Open-Source Robot Duck You Can Train Yourself
Hugging Face's MicroDuck is a $399 open-source robot duck that learns new tricks through reinforcement learning — and it ships before Christmas 2026.
Meta's MTIA 400 Chip Splits Its Brain Between AI Training and Ad Serving
Meta detailed its MTIA 400 accelerator at Hot Chips 2026: a 3nm LLM training chip built on Broadcom IP that also powers the ads that pay the bills.
OpenAI's Jalapeño Chip Beats Nvidia Blackwell in First Inference Benchmarks
OpenAI says its Broadcom-built Jalapeño chip delivers more tokens per user and per kilowatt than Nvidia Blackwell in first InferenceX benchmarks.
Apple Launches M6 and M5 Ultra Chips in New Mac Mini and Mac Studio Built for On-Device AI
Apple's first 2nm M6 chip and quad-die M5 Ultra debut in new Mac mini and Mac Studio, with up to 512GB unified memory for running giant AI models locally.
Intel Opens Hot Chips 2026 With Crescent Island: AI Inference GPU With Up to 480GB Memory
Intel detailed its Crescent Island data center GPU at Hot Chips 2026 - a 350W PCIe card built for agentic AI inference with up to 480GB of LPDDR5X.
Xiaomi Unveils Xring O3, O100 and D100 Chips Plus AI Cube Running 120B Models Locally
Xiaomi launched three self-developed Xring chips for phones, AI and cars — and an AI Cube mini PC that runs 120B-parameter models locally.
Nvidia's Groq 3 LPX Enters Full Production With 3,400 Tokens per Second for Agentic AI
Nvidia's Groq 3 LPX inference accelerator is in full production, posting a record 3,400 tokens per second at 100K context on the Vera Rubin platform.
Taiwan Indicts Nine Including Nvidia Manager Over Smuggled AI Servers Bound for China
Taiwan prosecutors indicted nine people, including an Nvidia manager and two Supermicro ex-employees, for smuggling 74 Blackwell AI servers into China.
Broadcom's Mega AI Debt Deal Could Reach $100 Billion to Fund Chips for Anthropic
Broadcom is in talks to raise up to $100 billion in debt for AI chips that Anthropic will lease, in what would be the largest AI financing deal ever.
Starcloud Raises $250 Million at $2.3 Billion Valuation to Build AI Data Centers in Orbit
Nvidia joins Starcloud's $250 million round at a $2.3 billion valuation as the startup races SpaceX to put AI data centers in orbit, with a launch in January.
Nvidia Customers Face Server Price Hikes Above 15% as Memory Costs Soar
Bloomberg reports Nvidia AI server prices will rise over 15% as memory costs climb, hitting Vera Rubin and Grace Blackwell systems due early next year.
Anthropic Hires Google TPU Architect Amir Salek to Accelerate Its Custom Silicon Push
Anthropic has hired ex-Google chip chief Amir Salek, who delivered seven TPU generations, as it builds an in-house semiconductor initiative.
Micron Commits $10 Billion to Boise Research Labs to Chase the Next Generation of AI Memory
Micron Research Labs will spend $10 billion over the next decade in Boise on post-DRAM memory, advanced packaging and manufacturing tech for the AI era.
Google Taps Marvell for Custom AI Chips in $12.2 Billion Warrant Deal
Google has signed Marvell to help build its custom TPU AI chips, granting a warrant worth up to $12.2 billion. Marvell stock jumped while Broadcom fell 5%.
Cerebras CS-4 Launches With WSE-3 Turbo: Rack-Scale AI Inference at 30x GPU Speed
Cerebras unveils CS-4 with WSE-3 Turbo wafers and Nexus rack architecture, claiming 30x faster inference than GPUs as it challenges Nvidia's dominance.
Google Reportedly Taps AMD for a CPU-Heavy Next-Gen TPU Built for Reinforcement Learning
A SemiAnalysis note says Google may pair AMD CPU technology with its TPU accelerators in a new design aimed at reinforcement learning and agentic AI.
Nvidia Cuts OpenAI Data Center Guarantee From $250 Billion to Under $120 Billion
Nvidia has cut its planned guarantee for OpenAI's Ohio data center from $250 billion to under $120 billion after investor concerns, WSJ and Reuters report.
Cerebras Powers OpenAI's New Ultrafast Mode: GPT-5.6 Sol Hits 750 Tokens Per Second
Cerebras and OpenAI unveil Ultrafast Mode for GPT-5.6 Sol, delivering 750 output tokens per second with no quality loss.
Cerebras and OpenAI Unveil GPT-5.6 Sol Ultrafast: 750 Tokens Per Second With No Quality Loss
Cerebras-powered Ultrafast Mode delivers up to 750 output tokens per second for OpenAI's GPT-5.6 Sol, running 11x faster than Claude Fable 5 with comparable accuracy.
TSMC Posts Record July Revenue, Up 44.7% as AI Chip Demand Surges
Taiwan Semiconductor's July revenue hit NT$467.58 billion on strong demand for 2-nanometer chips, with the company forecasting 40% growth for 2026 driven by AI applications.
Nvidia Taps Wall Street Giants for $500 Billion AI Infrastructure Funding Push
Apollo, Blackstone, BlackRock, Brookfield, Goldman Sachs, and KKR are in talks with Nvidia on a $500 billion AI infrastructure funding package as Nvidia shares fell on the news.
Microsoft to Unveil Next-Generation Maia 300 AI Chip as Soon as September
Microsoft is preparing to publicly unveil its new Maia 300 AI chip this fall, potentially in September, as it works to cut its heavy reliance on Nvidia's expensive GPUs.
OpenAI's Smart Speaker Will Cost Up to $400 and Use Moving Parts to Feel Alive
OpenAI's upcoming ChatGPT smart speaker, designed with Jony Ive's LoveFrom, could cost $300-$400 and feature moving parts that animate when it responds.
2027 Memory Capacity Is Reportedly Sold Out as AI Demand Triggers a New 'RAMageddon'
Memory capacity for 2027 is reportedly fully sold out as AI demand for HBM and DRAM soars. SK Hynix plans a $38 billion fab expansion, but relief is years away.
Lumilens Emerges From Stealth With $900 Million to Rewire AI Data Centers With Light
Optical interconnect startup Lumilens launched with over $900 million in funding and a $5.5 billion valuation, aiming to replace copper wiring between GPUs in AI data centers.
AMD Acquires Taalas, the Startup That Hardwires AI Models Directly Into Silicon
AMD has acquired Taalas, a Toronto startup that etches AI model weights into silicon for massive inference speedups, weeks after its Cerebras deal.
Lumilens Emerges From Stealth With $900 Million to Wire AI Data Centers With Light
Lumilens raised over $900 million at a $5.5 billion valuation to replace copper with optical interconnects and solve AI's GPU connectivity bottleneck.
SpaceX and Tesla Commit $16.8 Billion to Build Terafab Chip Plant in Texas
Elon Musk's SpaceX and Tesla confirmed a $16.8 billion initial investment for the Terafab semiconductor plant in Grimes County, Texas, near College Station.
Anthropic Builds an In-House Chip Design Team to Power Claude With Custom Silicon
Anthropic confirmed it is hiring a custom silicon team to co-design AI chips for Claude, joining OpenAI and Google in the race to control AI hardware.
Musk Commits SpaceX to Nvidia GPUs 'Exclusively,' Sending Shares Higher and AMD Lower
Elon Musk said SpaceX will use Nvidia chips exclusively going forward, with an optimized Vera Rubin NVL72 system slated to launch into space next year.
AMD's Q2 Revenue Surges as Data Center Sales Double, but Shares Slip on Lofty Expectations
AMD reported 50% revenue growth in Q2 2026 as data center sales more than doubled on Instinct GPU demand, yet shares fell on sky-high Wall Street expectations.
SK hynix and SanDisk Unveil First High Bandwidth Flash Standard Targeting AI Memory Bottleneck
SK hynix and SanDisk have published the first open High Bandwidth Flash (HBF) specification, promising up to 3 TB/s bandwidth to ease AI inference memory costs.
Nvidia's Jensen Huang Forecasts $7.9 Trillion Semiconductor Industry Fueled by Agentic AI
Nvidia CEO Jensen Huang says the global semiconductor industry could swell to $7.9 trillion as agentic AI drives unprecedented demand for computing hardware.
DeepSeek V4 Flash Runs on a Single AMD MI300X GPU, No Nvidia Required
A developer got DeepSeek's 304-billion-parameter V4 Flash model running on a single AMD MI300X at 168 tokens per second, proving AMD can serve frontier models without Nvidia.
Lenovo Googlebook Leak Reveals Gemini PCs With a Dedicated AI Key and a New Operating System
Leaked press images show Lenovo's first Googlebook laptop and 2-in-1 tablet with a dedicated Gemini key, light strips, and Gemini AI baked into a new OS replacing ChromeOS.
AI Chip Startups OLIX and DeepX Raise Billions as the Global Race to Challenge Nvidia Intensifies
London's OLIX raised $312M at a $3.3B valuation and Korea's DeepX hit $2.2B, as startups worldwide bet they can outbuild Nvidia on AI inference hardware.
AMD Releases Instella-MoE: A Fully Open 16B Model Trained From Scratch on Its Own Instinct GPUs
AMD's Instella-MoE-16B-A3B is a fully open Mixture-of-Experts model trained entirely on Instinct MI300X and MI325X GPUs, with all training weights, data, and configs released.
Moonshot's Kimi Runs on 20,000 Nvidia H200 Chips Leased From Alibaba
Bloomberg reports Moonshot AI trained its Kimi models on a 20,000-chip Nvidia H200 cluster supplied by Alibaba, exposing China's reliance on Western silicon.
Eliyan Becomes a Unicorn With $145 Million Series C to Fix AI's Chip-to-Chip Bottleneck
Silicon Valley startup Eliyan raised $145 million at a $1 billion valuation to build electro-optical interconnects that unclog the data bottleneck slowing giant AI chip clusters.
Brookfield and NextEra Unveil $100 Billion AI Data Center at Former Uranium Site in Kentucky
A coalition led by Brookfield and NextEra Energy plans a privately funded $100 billion AI data center campus at the DOE's Paducah site in Kentucky, backed by up to 4.6 gigawatts of dedicated power generation.
A 28.9-Million-Parameter Language Model Now Runs on an $8 Microcontroller
An open-source project runs a 28.9M-parameter LLM on a cheap ESP32-S3 chip at roughly 9 tokens per second, using a memory trick borrowed from Google's Gemma.
Nvidia Puts $1.5 Billion Behind US AI Chip Packaging With Amkor Deal
Nvidia's $1.5 billion Amkor partnership expands advanced chip packaging and testing in Arizona, targeting a growing bottleneck in AI semiconductor supply chains.
A Single Downed Power Line Exposed the AI Data Center Threat to the U.S. Grid
When a power line fell near Washington, D.C., over 3 gigawatts of AI data centers disconnected at once, spiking voltage across the PJM grid. Experts say it is a warning of worse to come.
NAVER, NVIDIA and Brookfield to Scale Korea's National AI Factory to 200 Megawatts
NAVER, NVIDIA and Brookfield will expand Korea's sovereign AI factory at GAK Sejong to 200 megawatts by 2028, with NVIDIA investing $1 billion and Brookfield funding up to $9 billion.
Nvidia and SK Group Unveil $500 Billion Alliance as Jensen Huang Declares a 'Golden Age of AI' for South Korea
Nvidia CEO Jensen Huang forged a $500 billion alliance with South Korea's SK Group at a US summit with President Lee and Samsung's Jay Y. Lee, declaring a 'golden age of AI' for the country.
Samsung Lands $200 Billion Broadcom Deal to Manufacture AI Chips in Largest Foundry Win to Date
Samsung Electronics has won a $200 billion deal to manufacture AI chips for Broadcom, the largest contract-manufacturing order in its history and a major push to close the gap with TSMC.
Microchip Technology to Acquire Hailo, Expanding Edge AI Chip Portfolio as Inference Market Heats Up
Microchip Technology signs a definitive agreement to acquire edge AI processor maker Hailo, significantly expanding its AI hardware portfolio.
AI Chip Startup Etched Closes $300 Million at $10.3 Billion Valuation, Doubling in Seven Months
Etched, the inference-chip startup founded by three Harvard dropouts, closed a $300 million Series C led by Sequoia at a $10.3 billion valuation after booking $1 billion in orders.
London's Humanoid Becomes Europe's First Humanoid Robotics Unicorn With $152 Million Round
UK startup Humanoid raised $152 million at a $1.35 billion valuation, backed by Bosch and Schaeffler, betting that wheeled robots can outpace bipedal rivals in factories.