Overview
The 2026 World Internet Conference (WIC) Wuzhen Summit opened on November 3 under the theme Open Source, Open and Shared. The Light of Internet expo previewed on November 2 with 671 enterprises from 54 countries and regions. Ahead of the summit, DeepSeek went beyond open-sourcing compilers and inference operator libraries to fully open its internal inference framework and codec components, exposing low-level capabilities to the community.
On the same day, Huawei’s Ascend 950 SuperNode won the WIC 2026 Leading Scientific and Technological Achievement Award. The supernode supports a 1,024-card cluster with 45% training-compute utilization—1.8x the industry average—and over 40 large models have completed native pre-training on Ascend.
Background & Interpretation
1. Technical Background and Evolution
China’s AI open source is shifting from releasing weights to releasing infrastructure: DeepSeek’s framework and MiniMax H3’s adoption by Nvidia and Pika signal a reordering of technical influence. Domestic compute breaks through via supernodes, cluster utilization and software-hardware co-design, narrowing the gap with overseas flagships.
2. Core Drivers and Underlying Mechanism
Open source lowers the innovation threshold and accelerates the ecosystem flywheel; open inference frameworks let developers reuse low-level capabilities and shorten deployment. Ascend 950’s 1,024-card unified-memory addressing and efficient communication sharply lift training efficiency, easing the chips but no ecosystem weakness.
3. Market Structure and Value-Chain Impact
Dual breakthroughs in open source and domestic compute drive synergy across AI chips, compilers, inference engines and applications. More models trained natively on domestic platforms shift adaptation from porting to native, building an autonomous base.
Implications & Outlook
Near-term, the summit reinforces open-source consensus and attracts global builders; medium-term, domestic ecosystem maturity decides AI autonomy. Watch community adoption of open frameworks, Ascend native-model expansion, and overseas collaboration attitudes.
Takeaways for Industry Participants
Developers should reuse open inference stacks and domestic compute to cut cost; enterprises can leverage the open-source dividend for industry models; investors should watch the open source + domestic compute coupling for supply-chain autonomy.
This column compiles industry information and shares technical perspectives; it does not constitute investment advice.