Alibaba Bolsters AI Play with Fresh Chip, LLM

Why It Matters

The upgrades position Alibaba to compete directly with Nvidia in AI hardware while shifting its cloud business toward higher‑margin AI services, accelerating revenue growth.

Key Takeaways

• Zhenwu M890 offers 144GB on-chip memory, triple predecessor performance
• Alibaba shipped over 560,000 Zhenwu chips, plans annual upgrades
• Qwen3.7-Max LLM handles multi-step autonomous tasks, integrates Claude Code
• Panjiu AL128 Supernode packs 128 accelerators for bursty AI inference
• AI revenue expected to surpass cloud compute within a year

Pulse Analysis

Alibaba’s introduction of the Zhenwu M890 marks a decisive move to close the performance gap with industry leader Nvidia. By integrating 144 GB of on‑chip memory and delivering three times the throughput of its prior generation, the chip enables both training and inference at scale, a critical capability for enterprises deploying large language models. The annual refresh cadence signals Alibaba’s intent to sustain rapid innovation, leveraging its T‑Head design expertise and the existing base of over half a million units already in the field.

On the software front, the Qwen3.7‑Max LLM reflects a shift from single‑prompt responses toward autonomous, multi‑step reasoning. Its ability to orchestrate external tools—over 1,000 interactions in a 35‑hour benchmark—demonstrates a maturing ecosystem for AI agents that can execute complex workflows without human oversight. Compatibility with frameworks like OpenClaw and Anthropic’s Claude Code broadens its appeal to developers seeking versatile, code‑centric models, positioning Alibaba as a credible alternative to OpenAI and Google’s offerings.

Strategically, these hardware and model advances underpin Alibaba’s forecast that AI‑driven services will eclipse conventional cloud compute revenue within twelve months. The Panjiu AL128 Supernode, housing 128 accelerators, equips the cloud platform to absorb unpredictable inference spikes typical of autonomous agents. This end‑to‑end stack—chip, model, and infrastructure—creates a high‑margin revenue stream and strengthens Alibaba’s competitive moat in the rapidly expanding AI cloud market.