Broadcom Launches Next-Generation Low-Power AI Chip to Support High-Speed Optical Networks

Broadcom Corporation recently unveiled its latest 5nm 200 G/channel optical pulse amplitude modulation digital signal processor, the BCM85822, at the 49th European Conference on Optical Communications. This chip integrates a laser driver, operates at a speed of 200 G per channel, employs PAM-4 modulation, and excels in metrics such as bit error rate and power consumption. The BCM85822 chip can achieve a switching capacity of 51.2 terabits, significantly enhancing data center network density. The new chip will also help achieve the goal of both electrical and optical interfaces reaching 200G, facilitating the migration of AI workloads from 800G to 1.6T. Broadcom estimates that the BCM85822 will be officially launched in mid-2024.

Broadcom predicts Anthropic and OpenAI will surpass Google to become the top two customers for AI custom chips

Patrick Chi, president of Broadcom, predicts that as large model vendors intensify their competition for computing power, Anthropic and OpenAI will surpass Google to become the top two core customers for Broadcom's AI ASIC business. Anthropic is expected to become the largest XPU custom chip customer by 2027 to 2028, followed closely by OpenAI in second place.

Samsung and Broadcom Team Up with Over $200 Billion Investment, Fully Betting on Next-Generation AI Chips

Samsung and Broadcom have reached a long-term deep cooperation, covering the entire supply chain of memory chips, foundry, and advanced packaging. It is expected that by 2030, the scale of cooperation will break through $200 billion, marking that AI chips have entered a stage of 'design plus manufacturing' deep integration. The core technology is Broadcom's ASIC custom design capabilities combined with Samsung's manufacturing. The two companies will jointly tackle challenges in sub-2nm process technology and HBM high-bandwidth memory fields.

Meta Joins Forces with Broadcom and TSMC, Self-Developed AI Chip Iris to Begin Mass Production in September

Meta is accelerating its self-developed AI chip initiative, collaborating with Broadcom and TSMC to develop a processor codenamed "Iris," expected to enter mass production in September 2026. The chip is the core of the MTIA project, having passed initial testing, aiming to optimize platform recommendations and generative algorithms, reduce data center costs, and gradually reduce reliance on third-party GPUs.

New Landmark in Space Computing Power: Beijing Space Computing Innovation Center Officially Launched, Opening a New Era of Satellite-AI Collaboration

The Beijing Space Computing Innovation Center was officially unveiled on June 29th in Zhongguancun, marking a new phase of practical collaboration in China's space computing power industry. The center adopts a 'company + alliance' dual-model approach, operated by TianSuan XingLian Technology, with core responsibilities including overcoming common technology challenges, aiming to promote the development of cutting-edge space computing power.

Report: Baidu Kunlun Chip to Go Public in Hong Kong, Target Valuation of $50 Billion

Baidu's AI chip company, Kunlun Chip, is advancing its plan to go public in Hong Kong, with a target valuation of $50 billion. As a core component of Baidu's AI hardware ecosystem, its listing not only marks a milestone in its capitalization but also injects new variables into the global AI computing power competition. Kunlun Chip has launched multiple generations of self-developed cloud AI chips, which have drawn significant attention from the industry.