Alibaba's AI Ambition: A Dual-Pronged Strategy for Market Leadership

Alibaba accelerates AI with three new open-source Qwen models and its Zhenwu 810E chip to cut costs and reduce reliance on external suppliers like Nvidia, despite near-term stock pressure.

Alibaba is accelerating its artificial intelligence initiatives through a coordinated push on two critical fronts. The Chinese tech giant has unveiled a trio of new open-source AI models while simultaneously advancing its proprietary semiconductor development, a combination analysts view as central to its investment narrative.

Hardware Independence: Custom Chips to Challenge Costs and Nvidia

A key pillar of Alibaba's strategy involves reducing reliance on external suppliers and controlling infrastructure expenses. According to an analysis by CMSI Research, the company is building a "full-stack" ecosystem that integrates base infrastructure, AI models, and product applications. This structure rests on three core components: the Qwen model family, its cloud computing business, and its semiconductor unit, Pingtouge.

Pingtouge has reportedly launched the Zhenwu 810E AI chip. CMSI benchmarks position it as a direct competitor to Nvidia's H20 chip. The strategic goal is clearly defined: in-house semiconductors are intended to lower unit costs within Alibaba's infrastructure. Success in this area would not only impact the profitability of AI services but also enhance scalability, determining how rapidly and cost-effectively new AI products can be deployed.

Software Advancements: A Trio of Efficient Open-Source Models

Complementing its hardware efforts, Alibaba released three new mid-sized, open-source models over two days on February 24 and 25. The releases include Qwen3.5-35B-A3B, Qwen3.5-122B-A10B, and Qwen3.5-27B.

Notably, developer claims suggest the Qwen3.5-35B-A3B variant can outperform older, significantly larger predecessor models. This points to substantial efficiency gains, achieving higher performance with a leaner architectural "weight."

Technical specifications were provided for the Qwen3.5-27B model. It operates with 27 billion parameters and features a native context window of 262,144 tokens, which developers state can be expanded to up to 1.01 million tokens. Its architecture employs a combination of Gated Delta Networks and a Sparse Mixture-of-Experts approach. Furthermore, its language support has been broadened to cover 201 languages. For API-based access, pricing signals from OpenRouter indicate the hosted Qwen3.5-Plus service starts at $0.40 per million input tokens and $2.40 per million output tokens.

Market Performance and Analyst Outlook

Despite these operational developments, Alibaba's share price remains under near-term pressure. In today's trading, the stock is at 129.00 euros, reflecting a decline of 0.92%. Market data from AASTOCKS also highlights notable short-selling activity, with short-sales volume reaching $701.78 million on February 25 and a short-interest ratio of 9.871%.

Against this backdrop, CMSI Research maintains an "Overweight" rating on the stock, with a price target of $204 for its US-listed shares. In the coming weeks, the market is likely to scrutinize whether Alibaba's integrated "software plus proprietary chip" approach delivers tangible improvements in both infrastructure costs and time-to-market for AI products—precisely the metrics central to the current analyst thesis.