Arm’s Neoverse CSS N4 gives chip designers a validated 128-core building block for AI servers, DPUs, networking silicon, and chiplet-based infrastructure.

OpenAI has unveiled its first custom inference chip, codenamed Jalapeno, developed in partnership with Broadcom. The move signals OpenAI's push to reduce dependence on Nvidia and take greater control of its AI infrastructure stack.

#Apple announced a new multiyear agreement with #Broadcom to design and produce #customsilicon components and wireless connectivity technologies. The agreement, exceeding $30 billion, will lead to the production of over 15 billion U.S.-made chips and support hundreds of American jobs. This partnership is part of Apple’s commitment to invest $600 billion in the U.S. economy over four years. https://www.apple.com/newsroom/2026/07/apple-to-increase-spend-with-broadcom-to-produce-billions-more-us-chips/?eicker.news #tech #media #news

Post by Sandeep Raiza

$100B custom-chip runway and 79% VMware margins clash with licensing exodus, Samsung ambiguity, and hyperscaler concentration.

OpenAI has finalized plans to produce custom AI chips with TSMC by 2026, aiming to reduce its dependence on Nvidia’s GPUs.

Apple is reportedly developing a new generation of custom chips for AI servers (Project Baltra), future Macs (M5, M6, M7, Sotra), and its first smart glasses, signaling a major expansion of its in-house...

Paris-based startup Kog is challenging the assumption that agentic AI requires exotic custom silicon, claiming its inference engine hits 3,000+ tokens per second on standard datacenter GPUs — territory previously reserved for purpose-built chips from the likes of Cerebras and Groq. With $5M raised a

Broadcom posted fiscal third-quarter AI chip revenue of $16.7 billion — a 221 percent increase from a year ago. Behind the figure is a bet Hock Tan made years ago, and three of the world’s most valuab

Nvidia-backed $3.5B deal gives MediaTek NVLink Fusion access as custom silicon TAM hits $80B by 2027

OpenAI’s first Jalapeño benchmarks turn its custom inference ASIC from roadmap item into real silicon, with efficiency claims aimed at the hardest part of AI scaling: serving requests fast.

OpenAI debuts benchmark results for its first custom inference chip, showing 1.5-1.9x more AI work per watt than Nvidia's latest hardware.

South Korean AI chip startup Rebellions has announced a partnership with U.S.-based Marvell Technology to co-develop custom AI infrastructure targeting sovereign-scale deployments. The collaboration a...

Meta has initiated testing of its first in-house AI chip, aiming to cut costs, reduce reliance on Nvidia, and improve AI infrastructure control.