AI Hardware & Devices
-
OpenAI MRC Protocol scales Ethernet to 131,000 GPUs — and gets every chip giant to sign on
On May 6, OpenAI, AMD, NVIDIA, Intel, Microsoft and Broadcom co-released MRC (Multipath Reliable Connection) and handed it to the Open Compute Project as an open spec. It’s not a product. It’s a networking protocol — and it’s pointed straight at InfiniBand’s last stronghold. What MRC actually is MRC is an extension of RoCE (RDMA… Continue reading
-
Mojo 1.0 Beta lands after 3 years: one codebase from CPU to GPU, no CUDA
Modular shipped Mojo 1.0.0b1 on May 7 — the first 1.0-track release of the AI-native language Chris Lattner has been building since 2023. HackerNews put it on the front page within hours: 239 points, 164 comments. Lattner is the same person behind LLVM, Clang, and Swift, so the AI infra crowd actually listens when he… Continue reading
-
Khosla-backed Genesis AI ships GENE-26.5 — one model that cracks eggs, pipettes, and plays piano
Khosla-backed Genesis AI just put out GENE-26.5, a foundation model for robotic manipulation — paired with their own human-scale dexterous hand and a tactile data glove. The launch demo runs a 20-step meal prep including one-handed egg cracking. Another shows the same model playing piano. All driven by a single autonomous brain, no task-specific fine-tunes.… Continue reading
-
Anthropic × SpaceX Colossus 1 Deal: 220,000 GPUs from xAI’s Old Datacenter, Live in 30 Days
Anthropic just bought every chip in SpaceX’s Colossus 1 datacenter. Yes — the same site Elon’s xAI built its empire on. Yesterday’s competitor is today’s landlord. The Numbers 220,000+ NVIDIA GPUs — a mix of H100, H200, and GB200. 300 megawatts of power. The kicker: the whole thing comes online within a month. Most hyperscaler… Continue reading
-
OpenAI AI Phone Fast-Tracked to H1 2027: Ming-Chi Kuo report says it ditches the app grid
Ming-Chi Kuo dropped a fresh supply chain note on May 5: OpenAI’s first phone is now slated for H1 2027 mass production, pulled in a full year from the prior 2028 timeline. MediaTek, Qualcomm and Luxshare are the named partners, and the SoC is a 2nm MediaTek Dimensity 9600. This is shaping up as the… Continue reading
-
Chrome Silent Gemini Nano Install puts a 4GB AI model on 3.8 billion devices without asking
A forensic post hit HackerNews #1 last week with 695 points: Chrome silently downloaded a 4GB Gemini Nano weight file in 14 minutes 28 seconds. No prompt, no notification, no opt-in. The file sits at OptGuideOnDeviceModel/weights.bin. Delete it and Chrome pulls it down again. Stopping it requires chrome://flags or enterprise policy — nothing in normal… Continue reading
Top AI Product 