Alibaba’s cloud arm, Alibaba Cloud, unveiled a new AI‑accelerator chip that the company brands as the most powerful silicon ever produced in China. Built on a cutting‑edge 5 nm process, the chip delivers more than 30 teraflops of AI compute while staying under 150 watts, a combination that outpaces existing domestic offerings and rivals leading foreign designs. Alibaba says the architecture integrates a massive on‑chip memory buffer and a custom interconnect, allowing it to handle the massive tensor operations required by today’s generative‑AI models without bottlenecking data movement. The company will ship the processor to its own data centers first, then make it available to enterprise customers through its Elastic Compute Service, positioning the chip as the backbone for large‑scale AI inference and training workloads.
Alongside the hardware announcement, Alibaba detailed an aggressive roadmap for its next‑generation large language model (LLM). The model, slated to be named “Tongyi‑Qianwen 2,” will exceed 100 billion parameters and be trained on a multilingual corpus that includes Chinese, English and several regional languages. Alibaba intends to open the model through its cloud platform, offering developers API access that competes directly with OpenAI’s GPT‑4 and Baidu’s Ernie series. By coupling the new chip with the upcoming LLM, Alibaba aims to reduce inference latency and cost, thereby making high‑quality generative AI services more affordable for Chinese businesses.
The moves are part of a broader push for technological self‑sufficiency in the face of export controls that have limited Chinese access to advanced semiconductors from the United States and its allies. By designing and manufacturing a home‑grown AI accelerator, Alibaba not only strengthens its own cloud ecosystem but also provides a template for other Chinese firms seeking to bypass foreign supply chains. The announcements signal that China’s AI hardware race is intensifying, and Alibaba’s dual focus on silicon and large‑scale models could reshape the competitive landscape for cloud‑based AI services both domestically and globally.
Artículos relacionados de LaRebelión:
- Alibaba Unveils Massive AI Model, New Cloud Chip
- US and China Align on AI Safety
- US and China to Discuss AI, Trade, and Minerals
- Huawei Accelerates AI Chip Roadmap and Performance
- Microsoft Unveils RTX Spark Desktop and Linux Tools
Fuente Original: NBC News
Artículo generado mediante AI.larebelion.
