Observed Signal · Sep 25, 2026 · Technical Release · Source: SemiAnalysis · Impact: 4/5 · Sentiment: Positive
China's AI Datacenter Boom: 24GW Capacity, $20B BAT Capex
SemiAnalysis introduces its China Datacenter Model, revealing China's datacenter capacity to be over 24GW, larger than EMEA and rest of Asia combined. This is driven by massive capex from Alibaba, Tencent, Baidu (BAT), and ByteDance, with BAT's combined Q2 2026 capex reaching $20B, more than doubling YoY. The report highlights the shift from telecom-dominated retail to wholesale AI-driven buildouts, the impact of the 'Eastern Data, Western Compute' policy, and rapid construction speeds enabled by prefabrication. Despite high vacancy rates in legacy facilities, AI demand fills new wholesale capacity quickly. The model tracks over 1,000 facilities and over 60 operators, providing building-level data on capacity, tenants, and growth trajectories through 2032.
This report provides a comprehensive building-level model of China's datacenter industry, revealing over 24GW of capacity and a massive capex surge from Chinese hyperscalers. This is critical for understanding global AI infrastructure supply, which directly impacts the AdTech ecosystem through AI-driven advertising tools, personalization, and data processing. The scale and speed of China's buildout could shift the balance of AI compute availability and cost globally.
Track Huawei Signals & Market Shifts in Real-Time
Polaris7 autonomous intelligence agents track regulatory filings, primary sources, executive changes, and deal flow 24/7. Create your free Explorer workspace to monitor these entities.
Key Takeaways & Evidence Grounding
- China's datacenter capacity reaches over 24GW as of 2026, exceeding EMEA and APAC ex-China.
- Combined capex of Alibaba, Tencent, and Baidu (BAT) reached $20B in Q2 2026, more than doubling YoY, with all three posting negative free cash flow.
- ByteDance occupies roughly one-fifth of delivered datacenter capacity in China, making it the largest tenant.
- GDS and VNET signed 1.3GW of wholesale orders in 1H26, but captured only a third of ByteDance and Alibaba orders in 2024-2026YTD.
- The 'Eastern Data, Western Compute' policy has shifted buildout to western hubs like Inner Mongolia, with 100MW facilities delivered in under 12 months.
Connected Companies & Entities
7 Entities mapped“Huawei self-builds almost everything; has self-build clusters in Gui'an....”
“Baidu's 2025 revenue fell 3%, online marketing down double digits; part of BAT combined capex....”
“Tencent wrote largest external ticket in DeepSeek's Series A....”
“Tencent deployed its third-gen modular design in 2014; Tencent's self-built campuses in Qingyuan inspected for being over scale; Tencent bac...”
“Combined BAT (Alibaba, Tencent, Baidu) capex reached $20B in Q2 2026; Alibaba Cloud revenue compounded at ~110% CAGR 2015-2019....”
“ByteDance's Doubao serves 345M monthly users; occupies ~20% of delivered datacenter capacity; operates three self-build clusters....”
“Alibaba's relationship with Moonshot involves training/serving on Alibaba's cloud....”
Ontology Mapping & Concepts
Related Market Signals & Shifts
Recent verified developments and strategic activity across this market segment.
Samsung to invest $1B in AI infrastructure firm Helix
Samsung Electronics and five affiliates will invest a combined $1 billion in Helix Digital Infrastructure, an AI infrastructure company launched by KKR and backed by Nvidia. Samsung Electronics contributes $500 million, with the rest from Samsung C&T, Samsung SDS, Samsung SDI, Samsung Life Insurance, and Samsung Fire & Marine Insurance. Helix, led by former AWS CEO Adam Selipsky, focuses on hyperscale data centers, power generation, transmission, and fiber-optic networks. The investment adds to over $10 billion already committed by other investors including KKR, Kuwait Investment Authority, Nvidia, and Vistra. The move allows Samsung to leverage its semiconductor, cooling, data center construction, and battery capabilities to expand in the AI infrastructure market.
Modal Labs closing in on $750M round at $15.75B valuation
AI inference infrastructure provider Modal Labs is nearing a $750 million funding round led by Accel at a $15.75 billion valuation, according to a source. This would more than triple its valuation from $4.65 billion in May. The round comes amid surging demand for inference services, with other startups like Baseten, Fireworks, and Fal also raising at higher valuations. Modal Labs, founded in 2021 by Erik Bernhardsson and Akshat Bubna, provides infrastructure for training and running AI models without managing servers. The company has surpassed $300 million in annualized revenue as of May. The funding talks follow a security incident in July where a customer's data was compromised, but Modal's platform was not breached.
GLM-5.3 Sparse Attention Impact on DRAM Memory TAM
This article analyzes the impact of sparse attention mechanisms, specifically DeepSeek Sparse Attention (DSA) used in Z.ai's GLM-5.3 model, on the total addressable market (TAM) for DRAM memory, including HBM and NAND. It explains that while sparse attention reduces KV cache memory and bandwidth during the attention operation, it does not reduce overall memory capacity requirements because the top-k selection still requires full context in HBM. The article discusses system optimizations like HiSparse, which offloads KV cache to host DRAM to overcome capacity bottlenecks. It also provides detailed performance and cost comparisons for serving GLM-5.3 on different hardware (GB200, GB300, MI355X) using inference engines like Dynamo-SGLang, Dynamo-TRT-LLM, and ATOM, highlighting cost-efficiency and interactivity trade-offs. The analysis includes a deep dive into GLM-5's architecture, including the lightning indexer, MLA configuration, and post-training pipeline.
Track Real-Time Market Signals & Shifts
Set up custom watchlists to receive automated, evidence-grounded executive digests whenever material signals or shifts occur across your tracked landscape.