OpenAI and Broadcom have introduced Jalapeño, their first custom AI accelerator designed specifically for large language model (LLM) inference workloads. Early tests show Jalapeño offers significantly better performance per watt compared to current AI accelerators. The chip is optimized to balance compute, memory, and networking to improve efficiency and reduce data movement. This collaboration also involves Celestica and aims to support deployment at gigawatt scale in data centers starting in 2026, marking a strategic move by OpenAI to develop more in-house hardware for AI infrastructure.
In 2026, Marvell Technology's stock surged 218% year-to-date, significantly outperforming Broadcom's 10% rise, despite Broadcom's much larger market value. Marvell's growth is driven by accelerating AI demand in its data center segment, with strong r...
South Korea's Kospi index fell nearly 10% due to foreign investors selling semiconductor stocks, causing circuit breakers to trigger. Analysts suggest the AI investment hype is shifting from major tech spenders like Microsoft, whose stock is down des...
OpenAI and Broadcom have unveiled their first custom-designed chip, Jalapeño, aimed at efficiently running AI technologies like ChatGPT. This milestone supports OpenAI's plan to deploy these chips in data centers worldwide, potentially consuming 10 g...
South Korean equities lost over $900 billion in value after the Kospi dropped nearly 10%, driven by SK Hynix and Samsung shares falling over 12%. The selloff followed SK Hynix surpassing Samsung as Korea's most valuable company and concerns over reco...
Broadcom introduced its first custom chip designed specifically for OpenAI, marking the start of multiple generations aimed at supporting future AI models. This move positions Broadcom as a serious competitor to Nvidia in the AI semiconductor market,...
