Huawei Ascend
AI overview
Sign in and the AI will write an overview from our coverage.
Headlines · 11
- Chinese GPU makers race on supernodes as buyers weigh cost and ecosystem
At the 26th China International Industry Expo, where the integrated-circuit zone debuted as a standalone exhibition, domestic GPU vendors focused on rack-scale "supernodes": MetaX showed its 64-GPU fully interconnected Xijing S600 plus a 16-GPU S600 Lite, Enflame showed the Yunsui ESL64-O built with ZTE, and Iluvatar CoreX packed 144 GPUs and 18 CPUs into one 44U rack, saying the TianGai 300 IXL144 reaches 100-nanosecond single-hop latency and a 15% effective bandwidth gain in certain MoE All-to-All scenarios (all vendor claims). Citing an IDC report via Reuters, the article says China's data-center AI accelerator shipments reached 2.008 million units in the first half of 2026, with Nvidia at 678,000 (34%) and Huawei at 572,000 (28%), while China GPU shipments fell 26% year on year to 991,000 units and other AI accelerator types rose nearly 80% to 1.016 million. The report argues that beyond performance and price, stable supply, software ecosystem maturity and the ability to bring real customers and revenue to compute centers will shape commercialization, with vendors ultimately answering how to lower customers' total cost of ownership.
36氪 人工智能 · 🔥 7 - ModelSphere open-source platform targets production-grade AI deployment
On October 16, Paradigm Intelligence (范式智能) and Beijing Ruisizhilian Technology (睿思智联) announced ModelSphere (模域), a one-stop production-grade AI deployment platform that is free and open source for enterprise users under Apache 2.0 and is already available on GitHub. The platform combines compute management, model deployment and runtime optimization in one workflow, offering a graphical console, preset deployments, automatic elastic scaling, prefill-decode disaggregation, three-tier KV caching, high-availability load balancing and automatic service-resumption after stream interruptions, plus AutoTune scheduling intended to speed up first-token response. The vendors claim about 285% higher peak throughput and roughly 4x performance versus a baseline, figures that are company-stated and not independently verified; the platform bundles the EngineX inference engine to schedule heterogeneous Chinese accelerators from Huawei Ascend, Iluvatar CoreX, Moore Threads and Hygon DCU, and says it has optimized support for models such as DeepSeek-V4/V4-Flash, Qwen, Kimi, MiniMax and GLM.
InfoQ 中文 AI&大模型 · 🔥 7 - China's $4T compute buildout faces 63% idle capacity as inference reshapes the math
China's Academy of Information and Communications Technology estimates that of roughly 3.8 billion card-hours of intelligent compute supplied nationwide in 2025, only about 1.4 billion were used — a 36.8% utilization rate — even as the NDRC cites projections of 4 trillion yuan in new direct investment in the compute network during the next five-year plan and Ulanqab touts 89 signed data center projects plus ambitions to become a "Token capital." The piece attributes the idle capacity to investment that outpaces demand, fixed costs at local state-funded AI centers that are hard to amortize, and customers who prefer cloud providers' packaged pay-as-you-go model services over migrating. As agents spread, it cites Gartner's forecast that global inference-related AI infrastructure spending will reach $23.3 billion in 2026, overtaking $19 billion for training, while DeepSeek's September time-of-day pricing and falling industry token prices make scheduling and per-task cost the new yardstick.
36氪 人工智能 · 🔥 5 - Microsoft details Azure Maia 300, targeting large-scale deployment in 2027
Microsoft chairman and CEO Satya Nadella announced on the evening of Oct 12 that the company plans to deploy Azure Maia 300 at scale in 2027, according to a TMTPost analysis. Microsoft says the chip uses a software-defined dataflow architecture and a chiplet design with its compute die on TSMC's N2P process, offering 18TB/s peak memory bandwidth per accelerator at 950W thermal design power, while a rack-level system of 96 accelerators provides 27.6TB of HBM capacity and 1.7PB/s of aggregate bandwidth; the company claims 40% more AI tokens per megawatt than leading GPU systems. The Maia line has moved from Maia 100 (November 2023) to Maia 200, released in January and now in production carrying Copilot workloads, to Maia 300 for 2027, though the article notes Microsoft still relies heavily on Nvidia GPUs and CUDA and says its custom silicon will not replace third-party supply.
钛媒体 · 🔥 0 - IDC: Nvidia's China data-center AI chip share falls to 34% as Huawei climbs to 28%
An IDC report seen by Reuters says Nvidia's share of China's data-center AI chip market fell to 34% in H1 2026, with AI accelerator shipments of about 678,000 units in January–June, down 44% year over year. Huawei shipped 572,000 accelerators, up from just over 300,000 a year earlier, lifting its share to 28% and narrowing the gap to 6 percentage points; nine of the 11 Chinese chipmakers tracked by IDC grew shipments, with Alibaba's T-Head at 7%, Cambricon 5%, Baidu's Kunlunxin 4% and Hygon 3%. IDC did not detail which products make up Nvidia's 34% share or whether restricted high-end parts such as B200 and B300 are included, and Nvidia did not respond to a request for comment; CEO Jensen Huang has repeatedly criticized export controls, saying they hand roughly $50 billion of China AI chip business to Huawei. Total China AI accelerator shipments reached 2.008 million units, with GPUs down 26% to 991,000 and other accelerators up nearly 80% to 1.016 million, surpassing GPUs in unit terms for the first time.
Techmeme · 🔥 0 - AI infra spending boom meets idle GPUs at Beijing salon
A closed-door salon on AI infrastructure hosted by Touzhongwang with Zhongguancun partners, republished by 36Kr, examined the gap between huge AI infrastructure spending and weak efficiency. It cites ByteDance's $29.6 billion syndicated loan, Alibaba's HK$80 billion placement, more than $80 billion raised earlier by Google and Amazon, and NVIDIA's $500 billion plan, while noting China's daily token calls topped 140 trillion in March, over a thousand times early-2024 levels. Participants cited a Carnegie Mellon test of 756 GPUs running 31 days with the cluster waiting nearly 20% of the time, claimed over half of OpenAI's inference energy goes to idle work, and said average GPU utilization at Chinese smart-computing centers is below 30%. Several investors called lowering token cost and raising utilization the new deciding factor; SGLang's RadixArk reportedly raised a $100 million seed at a $400 million valuation, while one dissenting view argued inference revenue will fall sharply as algorithms compress token use by orders of magnitude.
36氪 人工智能 · 🔥 0 - Huawei says Ascend tops Nvidia in China share as supply stays tight
According to a media briefing summary published by Jiemian, Huawei rotating chairman Xu Zhijun said Ascend is still far from meeting domestic demand and that Huawei has no plan to expand broadly overseas, urging the supply chain to "go faster, produce more"; he said Huawei's own data suggests Ascend's China market share has surpassed Nvidia's, though the note gave no figure or methodology. Nvidia said in its quarterly report that H200 shipments to some Chinese customers permitted since February accounted for less than 1% of data center revenue in the quarter, and it excluded China data center compute revenue from next-quarter guidance. The article also reports intent prices for Ascend 950DT accelerator cards up 20%-50% in two months to over 250,000 yuan and 950PR up from about 60,000 yuan early this year to over 80,000 yuan (intent quotes, not final prices), and that DeepSeek published Ascend-targeted software components including TileKernels and DeepGEMM-Ascend on September 30.
虎嗅 AI · 🔥 0 - Qualcomm to pay Huawei under patent cross-license, says analysis
An analysis published by TMTPost (authored by Shi Tianhao) says Huawei and Qualcomm reached a multi-year, broad patent licensing agreement covering cross-licenses in 5G, computing, AI and networking, with Qualcomm also acquiring some of Huawei's US patents in computing, AI and networking. Citing CCTV Finance, it says Qualcomm will pay to use more than 680 Huawei patents, and that Huawei's cumulative licensing revenue across all agreements is expected to exceed $6.9 billion (the value of this deal was not disclosed); Qualcomm has denied being a "net payer" and denied the transaction involves "logic folding." The piece sets the deal in a 25-year timeline in which Huawei, once the payer, has become a payee, contrasts the Kirin 9050 Pro's "logic folding" and Taifa with Xiaomi's XRING O1, and notes claims that Ascend's China market share has passed Nvidia, that DeepSeek is co-developing open-source software tooling for Ascend, and that AMD is buying World Labs for about $8.2 billion.
钛媒体 · 🔥 0 - vLLM climbs GitHub Trending among Python projects
vLLM is a high-throughput, memory-efficient inference and serving engine for large language models, originally developed in the Sky Computing Lab at UC Berkeley and now maintained by a community of more than 2,000 contributors. It manages attention key-value memory with PagedAttention and supports continuous batching, chunked prefill, prefix caching, speculative decoding and quantization formats including FP8, INT8/INT4, GPTQ/AWQ and GGUF, alongside an OpenAI-compatible API server. The engine runs on NVIDIA, AMD and Intel GPUs plus x86/ARM/PowerPC CPUs, and supports 200+ model architectures on Hugging Face.
GitHub Trending · Python · 🔥 0 - AMD to buy World Labs for $8.2B in all-stock deal
An analysis republished by 36Kr says AMD announced on Sept. 28 an all-stock acquisition of World Labs valued at $8.2 billion, expected to close by the end of 2026 pending regulatory approval, with Fei-Fei Li set to become an AMD executive vice president reporting to Lisa Su. The piece cites NVIDIA's roughly $12.9 billion purchase of Hugging Face, Qualcomm's acquisition of Modular and SpaceX's acquisition of Anysphere as signs chip giants are moving into the model layer. For Chinese chip makers such as Moore Threads and Huawei's Ascend, it argues that partnerships with world-model teams — rather than acquisitions — are the more realistic route, currently progressing from proving models can run to sharing in hardware definition.
36氪 人工智能 · 🔥 0 - DeepSeek open-sources Ascend ports of DeepGEMM and DeepEP, with Huawei backing
DeepSeek has open-sourced two new repositories for Huawei's Ascend accelerators: DeepGEMM-Ascend for matrix multiplication kernels and DeepEP-Ascend for the all-to-all communication that mixture-of-experts models rely on, with Huawei's support. TechRadar notes that the other components named in the announcement — TileKernels, DeepSelect and FlashMLA — are Ascend-targeted updates to existing projects, and that the headline TileLang is neither new nor DeepSeek's own. Requirements are narrow: DeepGEMM-Ascend needs Ascend 950 series silicon and CANN 9.20, DeepEP-Ascend needs Ascend 950 with UBMEM connectivity, and the published benchmarks ran on a proof-of-concept hardware kit DeepSeek says is not a publicly distributed release.
GitHub Trending（每日） · 🔥 0
Experience and discussion from the community
Share my Huawei Ascend experienceAsk about Huawei Ascend
Nobody has shared their experience with Huawei Ascend yet.