AMD unveiled a next-generation open-source telecom large language model (LLM), OTel 2.0, with AT&T and Microsoft at its Advancing AI 2026 conference, completing a training pipeline of more than 1 trillion tokens. AT&T used AMD Instinct GPUs and Microsoft Azure Foundry compute for post-training, pushing the Open Telco AI initiative from proof of concept toward commercial deployment.
Alongside its expanding AI hardware portfolio, AMD is stepping up efforts to strengthen its AI software ecosystem in a bid to challenge NVIDIA's CUDA dominance. At its Advancing AI 2026 event, the company unveiled ROCm.ai, a new AI-native development platform that integrates AI coding assistants, deployment tools, and automated optimization capabilities to simplify AI application development and accelerate model deployment on AMD hardware.
AMD said the artificial intelligence boom is entering a new phase that will force a major redesign of data centers, as demand shifts from model training toward inference, AI agents, and physical AI. At its Advancing AI 2026 event, the company introduced new server, accelerator, and rack-scale products to power the next generation of AI infrastructure.
Washington is opening a new front in its technology contest with China, moving beyond chips and artificial intelligence models to target humanoid robots and other autonomous systems over national security and data privacy concerns.
Agentic AI is spreading rapidly across personal computers and enterprise endpoints, prompting AMD to expand its local AI strategy with a new platform capable of running significantly larger models on-device.
NAN GreenMet Pvt. Ltd. has launched NAN MagneTech Pvt. Ltd. to develop what it says will be India's first fully integrated manufacturing facility for high-performance Neodymium-Iron-Boron (NdFeB) rare earth magnets. The company plans to invest INR12.5 billion (US$130 million) in the project, which will be located at the Naidupeta Industrial Park in Andhra Pradesh, with commercial production targeted for the first quarter of 2028.
AMD has launched its sixth-generation EPYC Venice server processor as the foundation of its next-generation AI data center platform, while CEO Lisa Su said the company is reshaping its roadmap for a world where AI workloads are spreading from training models to enterprise agents. The shift matters globally because it signals that data centers may need more balanced upgrades to CPU, GPUs, and networking.
India's central government has approved two Electronics Manufacturing Clusters (EMCs) in Tamil Nadu, at Manallur and Pillapaikkam, with a combined project cost of INR10.12 billion (US$104.75 million), in a move aimed at strengthening the state's electronics ecosystem and advancing the country's self-reliance drive. Union Minister for Electronics and Information Technology Ashwini Vaishnaw announced the approvals in a written reply in the Lok Sabha on July 22, 2026, according to Project Monitor and Trade Brains.
Wistron's opening of its D1 AI smart factory in Fort Worth, Texas, marks a new milestone for Taiwan's downstream electronics manufacturers. The plant will support the first US mass production of Nvidia GB300 compute boards, a development with far-reaching implications for Taiwan, the US and the global AI industry.
Stripe is in talks to acquire AI model-routing startup OpenRouter in a deal that could value the company at around US$10 billion, according to The Wall Street Journal, which cited people familiar with the matter. The report said an agreement could be announced soon, although negotiations could still collapse or another bidder could emerge. The acquisition would significantly expand Stripe's ambitions beyond payments into AI infrastructure.
AMD unveiled its sixth-generation EPYC 9006 server processors, widening its push into AI data centers with chips designed for agentic AI, cloud computing, enterprise workloads, and high-performance computing. The launch suggests CPUs may continue to gain importance even as much of the market remains focused on GPUs.
More coverage
