Periodic/Sector reports
IT Hardware: Takeaways From 2026 Huawei Connect
OVERWEIGHT (Maintained)
Analyst
Analyst
Highlights
- Huawei hosted its flagship annual Huawei Connect 2026 conference in Shanghai from 17-19 September under the theme Advancing the Agentic World. It outlined a core strategy focused on building AI infrastructure to support long-horizon agentic workloads and massive foundation models.
- Huawei brought forward the release of its flagship Ascend 960 AI accelerators ahead of schedule to 1Q27 (960DT) and 3Q27 (960PR), while unveiling supporting next-generation ICT hardware, including the liquid-cooled Atlas 960E SuperPod, liquid cooled TaiShan 950, NPO-enabled CloudEngine switches, and OceanStor M900 context memory storage.
- Its strategy remains centred on maximising token-per-watt efficiency by addressing interconnect and memory bottlenecks through NPO-enabled high-speed clustering, UnifiedBus scale-up protocols, and 4-level 3D AI data centre designs.
Analysis
- AI accelerators. Release schedules for the Ascend 960 series have been brought forward to 1Q27 for the training-focused 960DT and 3Q27 for the inference-focused 960PR (vs 4Q27 previously). Next-generation 970 and 980 accelerators remain on track for launch in 2028 and 2029, respectively. Set for a 3Q27 launch, the liquid-cooled Atlas 960E SuperPoD multi-rack system leverages Hi-ONE Near-Packaged Optics (NPO, 7.2 Tb/s per engine) over UnifiedBus to deliver an optical inter-rack scale-up domain interconnecting 4,096 NPUs, offering 8 EFLOPS of FP8 (16 EFLOPS of FP4) compute and up to 1 petabyte (PB) of high bandwidth memory capacity, providing the compute capacity needed for next generation 10-trillion parameter models.
Highlights
- Huawei hosted its flagship annual Huawei Connect 2026 conference in Shanghai from 17-19 September under the theme Advancing the Agentic World. It outlined a core strategy focused on building AI infrastructure to support long-horizon agentic workloads and massive foundation models.
- Huawei brought forward the release of its flagship Ascend 960 AI accelerators ahead of schedule to 1Q27 (960DT) and 3Q27 (960PR), while unveiling supporting next-generation ICT hardware, including the liquid-cooled Atlas 960E SuperPod, liquid cooled TaiShan 950, NPO-enabled CloudEngine switches, and OceanStor M900 context memory storage.
- Its strategy remains centred on maximising token-per-watt efficiency by addressing interconnect and memory bottlenecks through NPO-enabled high-speed clustering, UnifiedBus scale-up protocols, and 4-level 3D AI data centre designs.
Analysis
- AI accelerators. Release schedules for the Ascend 960 series have been brought forward to 1Q27 for the training-focused 960DT and 3Q27 for the inference-focused 960PR (vs 4Q27 previously). Next-generation 970 and 980 accelerators remain on track for launch in 2028 and 2029, respectively. Set for a 3Q27 launch, the liquid-cooled Atlas 960E SuperPoD multi-rack system leverages Hi-ONE Near-Packaged Optics (NPO, 7.2 Tb/s per engine) over UnifiedBus to deliver an optical inter-rack scale-up domain interconnecting 4,096 NPUs, offering 8 EFLOPS of FP8 (16 EFLOPS of FP4) compute and up to 1 petabyte (PB) of high bandwidth memory capacity, providing the compute capacity needed for next generation 10-trillion parameter models.
OVERWEIGHT (Maintained)
Analyst
Analyst
IMPORTANT NOTICE - DISCLOSURES AND DISCLAIMERS
This report is provided subject to, and must be read together with, the full Disclosures / Disclaimers available at this link, which are incorporated by reference into this report. In particular, this report is intended for general circulation and informational purposes only and does not constitute personal investment advice or a recommendation to buy or sell any investment product or security. You should independently evaluate the information and, where necessary, seek advice from a qualified financial adviser regarding the suitability of any investment. Analyst certifications required under applicable regulations, including SEC Regulation AC (where relevant), are included in this report. By accessing, receiving or using this report, you acknowledge that you have read, understood and agreed to be bound by the Disclosures / Disclaimers, as may be amended, supplemented or updated from time to time.
Related articles
7 Sept 2026
IT Hardware: 2Q26 Results: With Sector-wide Shortages Now, Buy Execution Over Bottlenecks
14 Sept 2026
IT Hardware - Handset: Apple-linked And Premium Suppliers To Widen Their Lead Amid The Downcycle
4 Sept 2026
Internet - 2Q26 Results Recap: AI Cloud Accelerates; E-Commerce Still Soft; Gaming Resilient
Our latest research
21 Sept 2026
China Internet: Hyperscalers Moving Into AI Chips: Custom Silicon Gains Traction, Compute Strategies Diverge
18 Sept 2026
Automobile: Weekly: Lithium Carbonate Price Drop Reconfirms Our Preference For OEMs; Bottom-fishing Opportunity Emerges For CATL
18 Sept 2026