Huawei Cloud has set out a China-first commercialization timetable for its Ascend 950 Lingqu AI cluster cloud service. At Huawei Connect 2026, Zhou Yuefeng, Huawei Cloud CEO, said commercial operations would begin in China on September 30, 2026, and then extend to global markets on November 30, 2026. Pandaily reported those dates and identified the offering as the Ascend 950 Lingqu AI cluster cloud service. TechNode independently reported the same China date, global-market date, and attribution to Zhou.
The schedule begins with China before a later global-market expansion. The two reports identify September 30 and November 30 as the respective dates, and describe Lingqu as an Ascend 950 AI cluster cloud service. They do not list individual countries or regions within the global-market step.
The two reports agree on the sequence as well as the dates: China commercial operations first, followed by a global-market expansion. In both accounts, the named offering is a cloud service. The reports do not identify a customer deployment, a chip launch, or individual country availability. The timetable therefore identifies the intended order of commercialization without listing users, pricing, capacity, contracts, or regional allocation.
Huawei Cloud’s language sets a clear distinction between the scheduled China operations and the later global-market expansion. September 30 is therefore the stated China date, while November 30 is the stated date for the broader market step. The available reporting does not say that operations had started, that users had gained access, or that the rollout had reached any particular scale when the timetable was announced. Reading the dates as a completed worldwide deployment would go beyond what Huawei Cloud said and what the two reports describe.
The timing also sits within the public record available to the two outlets. Pandaily and TechNode reported the China and global dates as Huawei Cloud’s commercial schedule. Their coverage supports an announced service path, while the separate Huawei release supplies product-status and roadmap context for Atlas and Ascend 960. Each source addresses a different part of Huawei’s infrastructure portfolio.
UnifiedBus anchors Huawei’s 1,024-card system claim
The service announcement is paired with Huawei’s description of the system intended for the Ascend 950 cloud service. Huawei said its UnifiedBus interconnect architecture would be used to build a 1,024-card AI cluster. The company said that configuration can provide up to 1 EFLOPS of FP8 computing power, 2 EFLOPS of FP4 computing power, and 256 TB of globally addressable memory. Pandaily and TechNode each presented those figures as Huawei-stated specifications. Neither report presents independent benchmarks for the performance or memory figures.
Huawei described capacity of up to 1 EFLOPS for FP8 computing power, alongside its 2 EFLOPS FP4 figure and 256 TB memory claim. Pandaily and TechNode present those values as Huawei-stated specifications. The reporting does not provide an independent performance benchmark for the configuration.
The 1,024-card figure belongs to the system description in the two reports. Huawei’s FP8 and FP4 figures are specified with different precision levels, and the up-to qualifier applies to the 1 EFLOPS FP8 claim. The cited coverage also calls the memory globally addressable, preserving the wording used in the reported system specification.
Huawei said the system is intended for full-scale model training on a unified cluster, without extra partitioning or cluster adjustments. That is a stated design objective. UnifiedBus is central to Huawei’s account of how the 1,024-card cluster is assembled, while the cited sources present the feature as a company description.
For separate historical context, EastFrontier has covered domestic AI-chip rules for state-backed data centers and Huawei’s Atlas 950 SuperPoD. Those earlier articles do not establish the Lingqu service schedule or its stated system specifications.
Atlas SuperPoD deployment provides separate context
A separate Huawei statement helps define what the Ascend 950 timetable does and does not cover. In its September 17 material, Huawei said that more than 1,000 Atlas 900 A3 SuperPoDs had been deployed and that the Atlas 950 SuperPoD was seeing large-scale commercial use. Those statements concern Atlas SuperPoD products. Huawei Cloud’s Ascend 950 Lingqu AI cluster cloud service is the announced cloud-service offering with the September 30 China operations date and the November 30 global-market date.
The reported deployment count refers to Atlas 900 A3 SuperPoDs, and Huawei’s commercial-use statement refers to the Atlas 950 SuperPoD. The Ascend 950 Lingqu item is a separate cloud-service timetable. Huawei’s September 17 material therefore places the Atlas product statements and the Lingqu service schedule in different parts of its infrastructure portfolio.
Huawei’s September 17 material also set out future Ascend roadmap points: Ascend 960DT would be available in Q1 2027, and Ascend 960PR would be ready in Q3 2027. Pandaily and TechNode reported the same planned quarters. These roadmap dates, the Atlas statements, and the Ascend 950 Lingqu schedule each describe different parts of Huawei’s stated AI infrastructure plans. For Lingqu, the cited materials state a global-market expansion date of November 30, 2026.
The public materials use closely related Ascend and Atlas labels for different products and schedules. The Atlas 900 A3 deployment figure belongs to Huawei’s statement about that SuperPoD product, while the Atlas 950 commercial-use statement concerns a different SuperPoD. The Ascend 960DT and 960PR points are 2027 roadmap dates. Lingqu is the cloud-service offering for which Huawei Cloud announced the September 30 China date and November 30 global-market date.